Add Batch c35bf38d-613a-4e90-9fda-1abe3e537e0e

radarenhancingradiologyreportgenerationwithsupplementaryknowledgeinjection/8b840976-be89-429b-a54d-844857c7238e_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd941ac1bc6926525bbc66a560ca7c01a6ff614879d2e998eb66dec17086f61b
+size 106101

radarenhancingradiologyreportgenerationwithsupplementaryknowledgeinjection/8b840976-be89-429b-a54d-844857c7238e_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1c98ee926ea5cad02166e5c8db1e59b98628bf5056598e17081d6f0afbcd98d
+size 129114

radarenhancingradiologyreportgenerationwithsupplementaryknowledgeinjection/8b840976-be89-429b-a54d-844857c7238e_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07e9f3617dd90b813a763aa7d49182c1f14c8106153d4764db143ade1a4a4975
+size 2168671

radarenhancingradiologyreportgenerationwithsupplementaryknowledgeinjection/full.md

@@ -0,0 +1,394 @@
+# RADAR: Enhancing Radiology Report Generation with Supplementary Knowledge Injection
+
+Wenjun Hou $^{1,2}$ , Yi Cheng $^{1*}$ , Kaishuai Xu $^{1*}$ , Heng Li $^{2}$ , Yan Hu $^{2\dagger}$ , Wenjie Li $^{1}$ , Jiang Liu $^{2,3\dagger}$
+
+$^{1}$ Department of Computing, The Hong Kong Polytechnic University
+
+$^{2}$ Research Institute of Trustworthy Autonomous Systems and Department of Computer Science and Engineering,
+
+Southern University of Science and Technology
+
+$^{3}$ School of Computer Science, University of Nottingham Ningbo China
+
+houwenjun060@gmail.com
+
+# Abstract
+
+Large language models (LLMs) have demonstrated remarkable capabilities in various domains, including radiology report generation. Previous approaches have attempted to utilize multimodal LLMs for this task, enhancing their performance through the integration of domain-specific knowledge retrieval. However, these approaches often overlook the knowledge already embedded within the LLMs, leading to redundant information integration. To address this limitation, we propose RADAR, a framework for enhancing radiology report generation with supplementary knowledge injection. RADAR improves report generation by systematically leveraging both the internal knowledge of an LLM and externally retrieved information. Specifically, it first extracts the model's acquired knowledge that aligns with expert image-based classification outputs. It then retrieves relevant supplementary knowledge to further enrich this information. Finally, by aggregating both sources, RADAR generates more accurate and informative radiology reports. Extensive experiments on MIMIC-CXR, CHEXPERT-PLUS, and IU X-RAY demonstrate that our model outperforms state-of-the-art LLMs in both language quality and clinical accuracy $^1$ .
+
+# 1 Introduction
+
+Radiology report generation (Chen et al., 2020, 2021) plays a crucial role in chest X-ray interpretation, requiring highly specialized domain knowledge (Jain et al., 2021; Irvin et al., 2019). Recent advances in foundation models (Pellegrini et al., 2023; Chen et al., 2024; Hyland et al., 2024), which leverage large language models (LLMs) for enhanced medical image analysis, have demonstrated
+
+![](images/1e03665be410c1bd28210f651b3b82a0ce8cf67b82a48c978341661866e2a185.jpg)  
+Figure 1: A motivating example. The report directly generated by the multimodal LLM showcases its knowledge regarding several findings  $(O_R)$  but can contain hallucinations and overlook some other findings. To address this, we regard the part that aligns with another expert model  $(O_R \cap O_I)$  as trustworthy and we incorporate supplementary knowledge for the remaining part  $(\mathcal{O} - O_R \cap O_I)$  to enhance the report generation.
+
+remarkable potential in generating fluent and cohesive clinical text, aiding radiologists in their diagnostic workflow.
+
+Despite their ability to generate highly readable and clinically plausible report content, LLMs still face persistent challenges in ensuring clinical accuracy. One major challenge lies in the knowledge gap between the medical and general domains. Many studies have attempted to bridge this disparity by augmenting models with retrieved domain-specific knowledge (Yang et al., 2021; Liu et al.,
+
+2021; Li et al., 2023c; Ranjit et al., 2023; Sun et al., 2025). However, these approaches often overlook the knowledge LLMs have already acquired. That is, much of the retrieved information is often duplicate knowledge already encoded within the model's parameters, leading to redundant information retrieval. Moreover, the knowledge learned by LLMs (Liu et al., 2024) is not always trustworthy, as hallucinations frequently occur (Huang et al., 2025). For instance, in Figure 1, the LLM correctly identifies Cardiomegaly, making the retrieval of additional knowledge about this observation unnecessary. Additionally, the generated Pleural Effusion is highly credible, as it aligns with the expert model, whereas Edema remains uncertain. Thus, balancing learned and retrieved knowledge in radiology report generation is crucial to address these challenges.
+
+In this paper, we propose RADAR, a framework for RADlology report generation that integrates both the internal knowledge of LLMs and external supplement ARy knowledge. Our framework primarily consists of two stages: preliminary findings generation and supplementary findings augmentation. In the first stage, RADAR generates an initial report from the input images. Subsequently, an expert model processes the images for observation classification. The overlapping information between the generated report and the classified observations is identified as high-confidence internal knowledge. In the second stage, RADAR additionally retrieves new knowledge to supplement the internal knowledge. Finally, both internal and supplementary knowledge sources are aggregated to enhance the report generation process. Our main contributions can be summarized as follows:
+
+- We propose RADAR, a novel framework that enhances the clinical accuracy of radiology report generation by effectively integrating both the internal knowledge of LLMs and externally retrieved domain-specific knowledge.  
+- To optimize knowledge utilization, we introduce a knowledge extraction method that identifies and retains non-overlapping information from the model's learned knowledge, reducing redundancy and bridging the knowledge gap.  
+- We conduct extensive experiments on three benchmark datasets: MIMIC-CXR, CHEXPERT-PLUS, and IU X-RAY, demonstrating the effectiveness of RADAR.
+
+# 2 Preliminary
+
+# 2.1 Problem Formulation
+
+A multimodal LLM (MLLM) generally consists of a vision encoder, a vision connector that transforms visual signals into the language space (e.g., MLP (Liu et al., 2023), Q-Former (Li et al., 2023b), or Perceiver Resampler (Xue et al., 2024)), and an LLM, as illustrated in the left part of Figure 2. For radiology report generation $^2$ , the MLLM takes a radiograph  $X$ , its prior  $X_{p}$  (if available), and the clinical context  $C$  (e.g., Indication or Prior Findings) as input and generates the report  $Y = \{y_{1}, \dots, y_{T}\}$ . The probability of the  $t$ -th token is computed as follows:
+
+$$
+p (y _ {t}) = \operatorname {M L L M} (X, X _ {p}, C, y _ {<   t}),
+$$
+
+where the MLLM is optimized using the negative log-likelihood loss:
+
+$$
+\mathcal {L} = - \sum_ {t = 1} ^ {T} \log p (y _ {t}).
+$$
+
+# 2.2 Semi-Structured Report as Knowledge
+
+In this paper, the training set of MIMIC-CXR serves as the knowledge source for radiology report generation. To effectively leverage the knowledge encoded in each report, we convert it into semi-structured data. Specifically, given a report consisting of  $N$  sentences,  $Y = \{S_{1},\dots ,S_{N}\}$ , we annotate each sentence using the 14-category CheXpert observations (Irvin et al., 2019) with the CheXbert model (Smit et al., 2020). Each observation falls into one of four classes: Positive, Negative, Uncertain, or Blank. To ensure conciseness, we retain only sentences annotated with Positive observations. These selected sentences collectively represent the knowledge extracted from the report, as illustrated in the top-right part of Figure 2. Note that we annotate and process Preliminary Findings ( $\S 3.1$ ) and Supplementary Findings ( $\S 3.2$ ) in the same manner.
+
+# 3 The RADAR Framework
+
+# 3.1 Stage I: Preliminary Findings Generation
+
+We illustrate the Stage I process in the left part of Figure 2. To assess the learned knowledge of an
+
+![](images/c91634318f55cef70831c90ed0cc87e4ac1c7cd97fc9f3368b35f7c5b000482a.jpg)  
+Stage I: Preliminary Findings Generation  
+Stage II: Supplement Findings Augmentation  
+Figure 2: Overview of the RADAR. In Preliminary Findings, only sentences that reach agreement are retained, whereas in Supplementary Findings, only sentences that supplement the Preliminary Findings are preserved.
+
+LLM, we first feed the input  $(X,X_{p},$  and  $C)$  into RADAR to generate a report  $\hat{Y}$  ..
+
+$$
+\hat {Y} = \underset {\hat {Y} \in \mathcal {Y}} {\operatorname {a r g m a x}} \prod_ {t = 1} ^ {T} \mathrm {M L L M} (X, X _ {p}, C, \hat {y} _ {<   t}),
+$$
+
+where  $\mathcal{V}$  represents the set of possible reports. Note that exact maximization is intractable and we employ an approximate decoding algorithm for generation. Next, we convert the findings into semistructured knowledge, as described in §2.2, and denote the observations of  $\hat{Y}$  as  $O_{R}$ .
+
+To extract credible knowledge from  $\hat{Y}$  while filtering out untrustworthy information, we train an expert model that predicts observations for the image. Unlike previous works (Hou et al., 2023b; Pellegrini et al., 2023), which consider only the image as input, we incorporate the clinical context to enhance performance. Specifically, the expert model  $f(X)$  encodes  $X$  and  $C$  using an image encoder  $\mathrm{Encoder}_v$  and a text encoder  $\mathrm{Encoder}_t$ , respectively, and then processes their outputs through an MLP for observation classification:
+
+$$
+\mathbf {h} _ {v} = \operatorname {E n c o d e r} _ {v} (X), \quad \mathbf {h} _ {t} = \operatorname {E n c o d e r} _ {t} (C),
+$$
+
+$$
+p (O _ {i}) = \sigma (\mathrm {M L P} ([ \mathbf {h} _ {v}; \mathbf {h} _ {t} ])),
+$$
+
+where  $[;]$  is the concatenation function,  $\mathbf{h}_v$  and  $\mathbf{h}_t$  are the pooled outputs of the image and text encoders, respectively, and  $p(O_i)$  represents the probability of the  $i$ -th observation. We denote the observations derived from  $f(X)$  as  $O_I$ , and the credible and high-confidence observations,  $O_{\checkmark}$ , are then obtained by intersecting  $O_I$  and  $O_R$ , as follows:
+
+$$
+O _ {\checkmark} = O _ {I} \cap O _ {R}.
+$$
+
+Finally, we refine  $\hat{Y}$  by removing sentences that do not correspond to  $O_{\checkmark}$ , yielding the Preliminary Findings (PF).
+
+To train the expert model, we collect observations from each report as image annotations and optimize the expert model using binary cross-entropy loss. Following Pellegrini et al. (2023), we address data imbalance by re-weighting the positive observations with a log-scale weight, defined as  $\alpha_{i} = \log \left(1 + \frac{|\mathcal{D}_{\mathrm{train}}|}{w_{i}}\right)$ , where  $|\mathcal{D}_{\mathrm{train}}|$  is the total number of training samples and  $w_{i}$  denotes the frequency of observation  $O_{i}$ .
+
+# 3.2 Stage II: Supplementary Findings Augmentation
+
+Supplementary Knowledge Retrieval. We follow the retrieval process of Yang et al. (2021) to search
+
+for domain knowledge. Specifically, the expert model described in §3.1 produces probabilities for 14 observations, and we compute the similarity between different samples using KL-divergence:
+
+$$
+\hat {z} = \mathrm {N o r m a l i z e} (f (X)),
+$$
+
+$$
+\operatorname {S i m} (X, X _ {i}) = - \sum_ {j = 1} ^ {| \mathcal {O} |} \hat {z} _ {j} \log \frac {\hat {z} _ {j}}{\hat {z} _ {i , j}},
+$$
+
+where  $\mathrm{Normalize}(\cdot)$  normalizes  $f(X)$  to 1,  $\hat{z}$  represents the normalized scores for  $(X)$ , and  $\hat{z}_{i,j}$  denotes the score of the  $j$ -th observation in the  $i$ -th sample from the database (i.e., the training set of the MIMIC-CXR dataset). We then rank the samples based on their similarity scores,  $\mathrm{Sim}(X,X_i)$ , and retrieve the top- $K$  reports, denoted as  $\mathcal{Y}^S = \{Y_1^S,\dots ,Y_K^S\}$ .
+
+Supplementary Knowledge Extraction. Since the retrieved information may overlap with the knowledge learned by LLMs, we extract only supplementary knowledge based on two principles: (1) it should be concise and relevant, and (2) it should complement, rather than duplicate, the preliminary findings. Thus, for each supplementary report  $Y_{i}^{S}$  with its corresponding observations  $O^{S}$ , we retain only the following observations:
+
+$$
+O _ {\delta} = \mathcal {O} - O _ {\checkmark}.
+$$
+
+Next, we convert  $Y_{i}^{S}$  into semi-structured knowledge and remove sentences that do not correspond to  $O_{\delta}$ , referring to these findings as Supplementary Findings (SF). Notably, all sentences corresponding to negative observations are removed, ensuring that SF remains concise and clinically relevant.
+
+# 3.3 Enhanced Radiology Report Generation
+
+We integrate both PF and SF into the clinical context  $C$  to form the augmented context  $C^A$ , from which the final report  $Y$  is generated as:
+
+$$
+Y = \operatorname *{argmax}_{Y\in \mathcal{Y}}\prod_{t = 1}^{T}\mathrm{MLM}(X,X_{p},C^{A},y_{<  t}).
+$$
+
+Since PF and SF contain information from various studies, summarizing high-level information before generating the report is necessary. Thus, we include the observations of  $Y$  as part of the training targets. Specifically, during training,  $Y$  is converted into a structured format:
+
+$$
+Y ^ {O} = \left\{O _ {1}, \dots , O _ {N}, y _ {1}, \dots , y _ {L} \right\},
+$$
+
+where  $\{O_1,\ldots ,O_N\}$  represents the observations in  $Y$  , and  $\{y_{1},\dots ,y_{L}\}$  corresponds to the tokens of the report. We refer to this process as Observation Identification (OI). During inference, we extract the final report from the generated output for evaluation.
+
+# 4 Experiments
+
+# 4.1 Datasets
+
+We evaluate our model using three publicly available radiology report generation datasets: MIMIC-CXR $^3$  (Johnson et al., 2019), CHEXPERT PLUS $^4$  (Chambon et al., 2024), and IU X-RAY $^5$  (Demner-Fushman et al., 2016):
+
+- MIMIC-CXR contains 377,110 chest radiographs and 227,827 reports. We use this dataset for fine-tuning, and we include only frontal images in our experiments. The number of samples in the train, validation, and test sets is 162,955, 1,286, and 2,461, respectively.  
+- CHEXPERT PLUS comprises 223,462 unique radiology reports and chest X-ray pairs from 187,711 studies. We evaluate our model using only frontal images from the validation set, which includes 62 samples.  
+- IU X-RAY is a dataset collected by Indiana University. Following Bannur et al. (2024), we use all frontal images for evaluation, totaling 3,199 studies.
+
+# 4.2 Evaluation Metrics
+
+Lexical Metrics. Following previous research (Chen et al., 2020; Li et al., 2023c), BLEU-1/4 (Papineni et al., 2002), ROUGE-L (Lin, 2004), and METEOR (Banerjee and Lavie, 2005) are adopted for evaluating the languages of generated outputs.
+
+Clinical Metrics. We evaluate the factual accuracy using several metrics. Specifically,  $\mathrm{RG - F_1}$  and  $\mathrm{RG}_{\overline{\mathrm{ER}} (\mathrm{ER})}$  (Jain et al., 2021) evaluate the entity-level factuality and RadCliQ0 (Yu et al., 2022), denoted as  $\mathrm{CliQ_0}$ , aligns with the preference of radiologists. For observation evaluation,  $^{14}\mathrm{MacroF_1}$ $(^{14}\mathrm{Ma - F_1})$  and  $^{14}\mathrm{Micro - F_1}$ $(^{14}\mathrm{Mi - F_1})$  evaluate the macro and micro  $\mathbf{F}_1$  of 14 observations (refers to Table 7), respectively. In addition,  $^{5}\mathrm{Macro - F_1}$
+
+Dataset: MIMIC-CXR (Training and Evaluation)  
+
+<table><tr><td rowspan="2">Model</td><td colspan="4">Lexical Metrics</td><td colspan="7">Clinical Metrics (CheXpert: Uncertain as Negative / Positive)</td></tr><tr><td>B-1</td><td>B-4</td><td>MTR</td><td>R-L</td><td>RG-F1</td><td>RGER</td><td>ClicQ0(↓)</td><td>14Ma-F1</td><td>5Ma-F1</td><td>14Mi-F1</td><td>5Mi-F1</td></tr><tr><td>RadFM</td><td>-</td><td>0.128</td><td>-</td><td>0.182</td><td>-</td><td>-</td><td>-</td><td>-</td><td>-</td><td>-</td><td>-</td></tr><tr><td>XrayGPT</td><td>0.128</td><td>0.004</td><td>0.079</td><td>0.111</td><td>-</td><td>-</td><td>-</td><td>-</td><td>-</td><td>-</td><td>-</td></tr><tr><td>R2GenGPT</td><td>0.411</td><td>0.134</td><td>0.160</td><td>0.297</td><td>-</td><td>-</td><td>-</td><td>0.389</td><td>-</td><td>-</td><td>-</td></tr><tr><td>R2-LLM</td><td>0.402</td><td>0.128</td><td>0.175</td><td>0.291</td><td>-</td><td>-</td><td>-</td><td>-</td><td>-</td><td>-</td><td>-</td></tr><tr><td>RaDialog</td><td>0.346</td><td>0.095</td><td>0.140</td><td>0.271</td><td>-</td><td>-</td><td>-</td><td>0.394</td><td>-</td><td>-</td><td>-</td></tr><tr><td>LLaVA-Med</td><td>0.354</td><td>0.149</td><td>0.353</td><td>0.276</td><td>0.191</td><td>0.238</td><td>3.30</td><td>0.269</td><td>0.363</td><td>0.427</td><td>0.439</td></tr><tr><td>CheXagent</td><td>0.169</td><td>0.047</td><td>-</td><td>0.215</td><td>-</td><td>0.205</td><td>-</td><td>0.247</td><td>0.345</td><td>0.393</td><td>0.412</td></tr><tr><td>GPT-4V</td><td>0.164</td><td>0.178</td><td>-</td><td>0.132</td><td>-</td><td>0.132</td><td>-</td><td>0.204</td><td>0.196</td><td>0.355</td><td>0.258</td></tr><tr><td>Med-PaLM</td><td>0.323</td><td>0.115</td><td>-</td><td>0.275</td><td>0.267</td><td>-</td><td>-</td><td>0.398</td><td>0.516</td><td>0.536</td><td>0.579</td></tr><tr><td>LLaVA-Rad</td><td>0.381</td><td>0.154</td><td>-</td><td>0.306</td><td>-</td><td>0.294</td><td>-</td><td>0.395</td><td>0.477</td><td>0.573</td><td>0.574</td></tr><tr><td rowspan="2">MAIRA-1</td><td rowspan="2">0.392</td><td rowspan="2">0.142</td><td rowspan="2">0.333</td><td rowspan="2">0.289</td><td rowspan="2">0.243</td><td rowspan="2">0.296</td><td rowspan="2">3.10</td><td>0.386</td><td>0.477</td><td>0.557</td><td>0.560</td></tr><tr><td>0.423</td><td>0.517</td><td>0.553</td><td>0.588</td></tr><tr><td>MAIRA-2</td><td>0.465</td><td>0.234</td><td>0.420</td><td>0.384</td><td>0.346</td><td>0.396</td><td>2.64</td><td>0.416</td><td>0.504</td><td>0.581</td><td>0.591</td></tr><tr><td>MedVerse</td><td>-</td><td>0.178</td><td>-</td><td>-</td><td>0.280</td><td>-</td><td>2.71</td><td>-</td><td>-</td><td>-</td><td>-</td></tr><tr><td>M4CXR</td><td>0.339</td><td>0.103</td><td>-</td><td>-</td><td>0.218</td><td>0.285</td><td>-</td><td>0.400</td><td>0.495</td><td>0.606</td><td>0.618</td></tr><tr><td>Libra</td><td>0.513</td><td>0.245</td><td>0.489</td><td>0.367</td><td>0.329</td><td>0.376</td><td>2.70</td><td>0.404</td><td>0.538</td><td>0.559</td><td>0.601</td></tr><tr><td rowspan="2">RADAR (Ours)</td><td rowspan="2">0.509</td><td rowspan="2">0.262</td><td rowspan="2">0.450</td><td rowspan="2">0.397</td><td rowspan="2">0.346</td><td rowspan="2">0.393</td><td rowspan="2">2.61</td><td>0.460</td><td>0.567</td><td>0.627</td><td>0.653</td></tr><tr><td>0.497</td><td>0.602</td><td>0.627</td><td>0.674</td></tr></table>
+
+Table 1: Evaluation results of our model and baseline methods on the MIMIC-CXR dataset. Baseline results are cited from their respective literature. The best results are shown in bold, while underlined values indicate the second-best results.  $\downarrow$  denotes that lower values are better. Results of CheXpert treat Uncertain labels as Positive when compared with MAIRA-1. Comparisons with SOTA specialists are provided in Table 8.  
+
+<table><tr><td colspan="7">Dataset: IU X-RAY (Evaluation Only)</td></tr><tr><td rowspan="2">Model</td><td colspan="2">Lexical</td><td colspan="4">Clinical</td></tr><tr><td>B-4</td><td>R-L</td><td>RG-F1</td><td>ClicQ0(↓)</td><td>14Ma-F1</td><td>14Mi-F1</td></tr><tr><td>LLaVA-Rad</td><td>-</td><td>0.253</td><td>-</td><td>-</td><td>-</td><td>0.535</td></tr><tr><td>MAIRA-2</td><td>0.117</td><td>0.274</td><td>0.271</td><td>2.68</td><td>0.319</td><td>0.525</td></tr><tr><td>RADAR (Ours)</td><td>0.116</td><td>0.276</td><td>0.237</td><td>2.78</td><td>0.325</td><td>0.546</td></tr><tr><td>BACKBONE</td><td>0.112</td><td>0.275</td><td>0.236</td><td>2.79</td><td>0.269</td><td>0.514</td></tr></table>
+
+$(^{5}\mathrm{Ma - F_{1}})$  and  ${}^{5}\mathrm{Micro - F_{1}}$ $(^{5}\mathrm{Mi - F_{1}})$  measure the performance of 5 common observations (Atelectasis, Cardiomegaly, Consolidation, Edema, and Pleural Effusion). Two lines of CheXpert results are reported, i.e., Uncertain as Negative and Uncertain as Positive.
+
+# 4.3 Baselines
+
+On the MIMIC-CXR dataset, we compare our models with the state-of-the-art (SOTA) MLLMs, including RadFM (Wu et al., 2023), XrayGPT (Thawakar et al., 2024), LLaVA-Med (Li et al., 2023a), R2GenGPT (Wang et al., 2023b), R2-LLM (Liu et al., 2024), RaDialog (Pellegrini et al., 2023), CheXagent (Chen et al., 2024), GPT-4V (OpenAI, 2023), LLaVA-Rad (Chaves et al., 2024), MedPaLM (Singhal et al., 2022), MAIRA-1 (Hyland
+
+Table 2: Evaluation on the IU X-RAY dataset. Results of LLaVA-Rad and MAIRA-2 are cited from Bannur et al. (2024).  
+
+<table><tr><td colspan="7">Dataset: CHEXPERT PLUS (Evaluation Only)</td></tr><tr><td rowspan="2">Model</td><td rowspan="2">Train</td><td colspan="2">Lexical</td><td rowspan="2">RGER(ER)</td><td rowspan="2">Clinical14(5)Ma-F1</td><td rowspan="2">14(5)Mi-F1</td></tr><tr><td>B-4</td><td>R-L</td></tr><tr><td rowspan="3">SWINv2-BERT</td><td>M*</td><td>0.034</td><td>0.191</td><td>0.136 (0.198)</td><td>0.268 (0.383)</td><td>0.410 (0.423)</td></tr><tr><td>C</td><td>0.057</td><td>0.228</td><td>0.183 (0.250)</td><td>0.331 (0.401)</td><td>0.508 (0.432)</td></tr><tr><td>M&amp;C</td><td>0.056</td><td>0.234</td><td>0.201 (0.277)</td><td>0.366 (0.495)</td><td>0.560 (0.532)</td></tr><tr><td rowspan="2">RADAR (Ours)</td><td rowspan="2">M</td><td rowspan="2">0.076</td><td rowspan="2">0.203</td><td rowspan="2">0.143 (0.216)</td><td>0.362 (0.417)</td><td>0.541 (0.524)</td></tr><tr><td>0.401 (0.540)</td><td>0.554 (0.608)</td></tr><tr><td rowspan="2">BACKBONE</td><td rowspan="2">M</td><td rowspan="2">0.073</td><td rowspan="2">0.203</td><td rowspan="2">0.143 (0.206)</td><td>0.282 (0.437)</td><td>0.477 (0.466)</td></tr><tr><td>0.317 (0.502)</td><td>0.492 (0.552)</td></tr></table>
+
+Table 3: Evaluation on the CHEXPERT PLUS dataset. The results for  $\mathrm{SWIN}_{\mathrm{V2}}$  -BERT are cited from Chambon et al. (2024), and we primarily compare RADAR with its  $\star$  variant. The "Train" column indicates the training datasets, where M and C denote the MIMIC-CXR and CHEXPERT PLUS datasets, respectively.
+
+et al., 2024), MAIRA-2 (Bannur et al., 2024), MedVerse (Zhou et al., 2024), M4CXR (Park et al., 2024), and Libra (Zhang et al., 2024). Other SOTA specialists are in the Appendix A.1. We also compare RADAR with LLaVA-Rad and MAIRA-2 on the IU X-RAY dataset. On the CHEXPERT-PLUS dataset, we compare RADAR with the baseline  $\mathrm{SWIN}_{\mathrm{V2}}$ -BERT (Chambon et al., 2024) consisting of a Swin Transformer V2 (Liu et al., 2022) and a BERT decoder (Devlin et al., 2019). The  $\mathrm{SWIN}_{\mathrm{V2}}$ -BERT model has three variants, each trained on a different dataset: the MIMIC-CXR dataset, the CHEXPERT PLUS dataset, and a combined version
+
+<table><tr><td rowspan="2">Model</td><td colspan="3">Modules</td><td colspan="4">Lexical Metrics</td><td colspan="7">Clinical Metrics (CheXpert: Uncertain as Negative)</td></tr><tr><td>PF</td><td>SF</td><td>OI</td><td>B-1</td><td>B-4</td><td>MTR</td><td>R-L</td><td>RG-F1</td><td>\(RG_{ER}\)</td><td>\(C_iQ_0\)(↓)</td><td>\(^{14}Ma-F_1\)</td><td>\(^{5}Ma-F_1\)</td><td>\(^{14}Mi-F_1\)</td><td>\(^{5}Mi-F_1\)</td></tr><tr><td>RADAR</td><td>✓</td><td>✓</td><td>✓</td><td>0.509</td><td>0.262</td><td>0.450</td><td>0.397</td><td>0.346</td><td>0.393</td><td>2.61</td><td>0.460</td><td>0.567</td><td>0.627</td><td>0.653</td></tr><tr><td>BACKBONE</td><td>×</td><td>×</td><td>×</td><td>0.497</td><td>0.259</td><td>0.444</td><td>0.396</td><td>0.343</td><td>0.387</td><td>2.67</td><td>0.402</td><td>0.495</td><td>0.565</td><td>0.581</td></tr><tr><td>\(RADAR_{w/o F}\)</td><td>×</td><td>×</td><td>✓</td><td>0.506</td><td>0.260</td><td>0.448</td><td>0.396</td><td>0.343</td><td>0.391</td><td>2.63</td><td>0.442</td><td>0.545</td><td>0.624</td><td>0.651</td></tr><tr><td>\(RADAR_{w/o SF}\)</td><td>✓</td><td>×</td><td>✓</td><td>0.508</td><td>0.262</td><td>0.451</td><td>0.398</td><td>0.346</td><td>0.394</td><td>2.62</td><td>0.447</td><td>0.543</td><td>0.626</td><td>0.650</td></tr><tr><td>\(RADAR_{w/o PF}\)</td><td>×</td><td>✓</td><td>✓</td><td>0.508</td><td>0.261</td><td>0.450</td><td>0.396</td><td>0.344</td><td>0.389</td><td>2.63</td><td>0.456</td><td>0.559</td><td>0.623</td><td>0.652</td></tr></table>
+
+Table 4: Ablation results of RADAR with different modules. Per-observation results of BACKBONE, RADARw/o F, RADARw/o SF, RADARw/o PF, and RADAR are provided in Appendix, Table 7.  
+
+<table><tr><td rowspan="2">Model</td><td rowspan="2">Vision</td><td rowspan="2">Modules Resampler</td><td rowspan="2">LLM</td><td colspan="4">Lexical Metrics</td><td colspan="6">Clinical Metrics (CheXpert: Uncertain as Negative)</td></tr><tr><td>B-1</td><td>B-4</td><td>MTR</td><td>R-L</td><td>RG-F1</td><td>RGER</td><td>ClioQ0(↓)</td><td>14Ma-F1</td><td>5Ma-F1</td><td>14Mi-F1</td></tr><tr><td>BACKBONE</td><td>✓</td><td>✓</td><td>✓</td><td>0.497</td><td>0.259</td><td>0.444</td><td>0.396</td><td>0.343</td><td>0.387</td><td>2.67</td><td>0.402</td><td>0.495</td><td>0.565</td></tr><tr><td>BACKBONE-V1</td><td>X</td><td>✓</td><td>X</td><td>0.430</td><td>0.183</td><td>0.359</td><td>0.318</td><td>0.245</td><td>0.296</td><td>3.15</td><td>0.284</td><td>0.415</td><td>0.476</td></tr><tr><td>BACKBONE-V2</td><td>X</td><td>✓</td><td>✓</td><td>0.483</td><td>0.246</td><td>0.428</td><td>0.381</td><td>0.321</td><td>0.368</td><td>2.78</td><td>0.361</td><td>0.465</td><td>0.532</td></tr></table>
+
+Table 5: Ablation results of fine-tuning different modules of BACKBONE.  
+
+<table><tr><td>Hyperparameters</td><td>Stage I</td><td>Stage II</td></tr><tr><td>Trainable Module</td><td>Vision Encoder (LoRA) 
+Perceiver Resampler (Full) 
+LLM (LoRA)</td><td>LLM (LoRA)</td></tr><tr><td>Training Epoch</td><td>3</td><td>2</td></tr><tr><td>Learning Rate</td><td colspan="2">1e-4</td></tr><tr><td>Optimizer</td><td colspan="2">AdamW</td></tr><tr><td>LR Scheduler</td><td colspan="2">Cosine</td></tr><tr><td>Warmup Ratio</td><td colspan="2">0.03</td></tr><tr><td>LoRA Config</td><td colspan="2">r=64,α=128</td></tr><tr><td>Batch Size</td><td colspan="2">32</td></tr></table>
+
+Table 6: Detailed hyperparameters for training RADAR. LoRA is used to fine-tune both the vision encoder and the LLM, while the Perceiver Resampler is fully finetuned.
+
+of both.
+
+# 4.4 Implementation Details
+
+Training and Inference. We implement RADAR using BLIP-  $3^{6}$  (Xue et al., 2024) as the backbone, which comprises a SigLIP (Zhai et al., 2023) vision encoder, a Perceiver Resampler, and a Phi3-mini $_{3.8\mathrm{B}}$  (Abdin et al., 2024) language model. Our implementation is based on Hugging Face's Transformers library (Wolf et al., 2020). The expert model consists of a Swin Transformer V $^{7}$  (Liu et al., 2022) and a BioClinicalBERT $^{8}$  (Alsentzer et al., 2019). Top-2 reports are selected as knowledge. The hyperparameters used for training RADAR are provided in Table 6. During inference, we employ beam search with a beam width
+
+of 5 for report generation and set the length penalty to 2.0. As proposed by Xue et al. (2024), BLIP-3 samples vision tokens using a Perceiver Resampler with learned queries and supports images of any resolution, resulting in significant performance gains across multiple tasks. In this paper, we use only the base resolution  $(384 \times 384)$  with 128 learned query tokens to ensure a fair comparison with other baselines. For training, in Stage I, we fine-tune all three components (i.e., the vision encoder, the Perceiver Resampler, and the LLM) in BLIP-3 since the model is not specifically designed for medical tasks. In Stage II, we further fine-tune only the LoRA of the LLM to enhance performance.
+
+Data Preprocessing. Following previous research (Hyland et al., 2024; Bannur et al., 2024; Zhang et al., 2024), we incorporate Indication, History, Comparison, Technique, and Prior Findings as clinical context for the MIMIC-CXR and CHEXPERT PLUS datasets, when available. Since the IU XRAY dataset does not include follow-up studies, we extract only Indication, Comparison, and Technique as clinical context. For a better illustration, we provide the prompt template in Table 9.
+
+# 5 Results and Analyses
+
+# 5.1 Quantitative Analysis
+
+Comparison with MLLMs. As shown in Table 1, RADAR achieves SOTA performance compared to other MLLM baselines. In terms of lexical metrics, RADAR outperforms the best baselines (i.e., Libra and MAIRA-2) with absolute improvements of  $1.7\%$  in BLEU-4 and  $1.3\%$  in ROUGE-L, while maintaining competitive performance of 0.509 in BLEU-1 and 0.450 in METEOR. Regarding entity
+
+level clinical metrics, our model achieves the best performance on  $\mathrm{RG - F_1}$  and RadCliQ0, attaining scores of 0.346 and 2.61, respectively. Additionally, RADAR surpasses the top baselines, achieving improvements across multiple observation-level clinical metrics, with  $^{14}\mathrm{Macro - F_1}$  increasing to 0.460,  $^{5}\mathrm{Macro - F_1}$  to 0.567,  $^{14}\mathrm{Micro - F_1}$  to 0.627, and  $^{5}\mathrm{Micro - F_1}$  to 0.653, respectively. Notably, the smallest gain over the second-best model is  $2.1\%$ , underscoring RADAR's effectiveness. Furthermore, we provide an additional set of CheXpert results using the Uncertain as Positive policy and compare RADAR with MAIRA-1. We observe that the improvements under this setting follow a similar trend to those obtained with the Uncertain as Negative policy. These results collectively demonstrate the effectiveness of RADAR in generating coherent and clinically accurate radiology reports.
+
+Comparison with SOTA Specialists. The results of other specialists are shown in Table 8. We find that models incorporating clinical context (e.g., Indication) as input generally achieve better performance than others. For example, the Controllable model significantly outperforms other baselines across both lexical and clinical metrics. This trend also holds for MLLMs, as shown in Table 1. Moreover, benefiting from the strong contextual comprehension and language generation capabilities of LLMs, RADAR further improves linguistic quality, which requires models to integrate diverse information sources. However, we observe that the  $^{14}\mathrm{Macro - F_1}$  score of our model still lags behind that of the Controllable baseline (0.497 vs. 0.553). This discrepancy may stem from differences in learning objectives, as this baseline treats Uncertain cases as Positive.
+
+Model Generalization. Following prior research (Bannur et al., 2024), we further evaluate RADAR on the CHEXPERT PLUS and IU X-RAY datasets to assess its generalization capability. The results are presented in Table 2 and Table 3. On the IU X-RAY dataset, RADAR outperforms MAIRA-2 in terms of CheXpert metrics, achieving a  $^{14}$  Macro- $\mathrm{F_1}$  of 0.325 and a  $^{14}$  Micro- $\mathrm{F_1}$  of 0.546. However, a performance gap remains in RG- $\mathrm{F_1}$  and RadCliQ0, which may be attributed to differences in training data, as MAIRA-2 is trained with the additional USMix dataset. Meanwhile, RADAR demonstrates comparable performance to the baselines in terms of lexical metrics. On the CHEXPERT PLUS dataset, our model significantly outperforms  $\mathrm{SWIN_{V2}}$ -BERT trained on the MIMIC
+
+![](images/930cc964c5c8f14b2ef5b5eef089a4af82fe11fe6051dde26eaa2155b9ede6ee.jpg)  
+Figure 3: Comparisons among BACKBONE+RAG, BACKBONE+FP+SF, and RADAR on six clinical metrics.
+
+CXR dataset, across both lexical and clinical metrics. Furthermore, RADAR surpasses the baseline that is trained on CHEXPERT PLUS alone as well as the one trained on a combination of both datasets. These results demonstrate the strong generalization ability of RADAR across different datasets. Additionally, RADAR significantly outperforms the BACKBONE, underscoring the effectiveness of the integrated knowledge.
+
+Analysis of PF, SF, and OI. We analyze the impact of PF, SF, and OI on the performance of RADAR, with results summarized in Table 4.  $\mathrm{RADAR}_{\mathrm{w/oF}}$  which first identifies observations before report generation without incorporating knowledge, significantly improves the CheXpert metrics, particularly  $^{14}\mathrm{Macro - F_1}$  and  $^{5}\mathrm{Macro - F_1}$ , as observation information captures high-level abstractions of reports and aligns closely with the objectives of these metrics. This highlights the crucial role of OI in enhancing clinical accuracy, independent of other components. When PF and SF are introduced individually with OI, introducing PF alone helps preserve the knowledge embedded in the LLM, resulting in comparable performance across both lexical and clinical metrics. In contrast, introducing SF alone substantially improves  $^{14/5}\mathrm{Macro - F_1}$ , but negatively impacts  $\mathrm{RG_{ER}}$  and  $\mathrm{RadCliQ_0}$ . Moreover, combining both PF and SF leverages the strengths of each, leading to further improvements in the clinical metrics while maintaining comparable performance across the other metrics. We notice that BACKBONE tends to retain easily acquired knowledge (i.e., PF) and that selectively supplementing it with external information (i.e., SF) is crucial for bridging the remaining knowledge gaps.
+
+![](images/e90f341160186615965f72e2f0f07c55c447fa49d1d8f27c9fb55f8452a3e432.jpg)  
+Preliminary Findings:  $O_{R}$  [Atelectasis] Heart size is normal....[Atelectasis] Mild atelectasis is noted in the left lung base. No focal consolidation, pleural effusion or pneumothorax is present..... Expert Observations:  $O_{I} = \emptyset$  Agreement:  $O_{\vee} = \emptyset$  Overall Findings:  $O = \emptyset$  The cardiac, mediastinal and hilar contours are normal. Lungs are clear.... Reference:  $O = \emptyset$  Cardiac, mediastinal and hilar contours are normal....
+
+# Preliminary Findings:  $O_{R}$  [Edema, Cardiomegaly]
+
+[Edema] .... there is still evidence of mild-to-moderate pulmonary edema. [Cardiomegaly] Moderate cardiomegaly persists.
+
+# Supplementary Findings:
+
+[Atelectasis] There are areas of left mid-to-lower lung linear atelectasis/scarring. [Atelectasis] Mild areas of atelectasis at the left lung bases.
+
+Expert Observations:  $O_I$  [Cardiomegaly, Atelectasis, Edema] Agreement:  $O_{\checkmark}$  [Edema, Cardiomegaly] Overall Findings:  $O$
+
+[Edema] Moderate pulmonary edema with [Cardiomegaly] moderate cardiomegaly and [Atelectasis] bilateral areas of atelectasis.
+
+# Reference:
+
+[Edema] There are still signs indicative of mild pulmonary edema. [Atelectasis] In addition, there is a small right medial basal atelectasis. [Cardiomegaly] Moderate cardiomegaly. ....
+
+![](images/ad00f7031a948baf988ec75d730f575c44af8b338742f6a07b7d32e24ba5d214.jpg)  
+Case B
+
+![](images/2c270e39d2f72554e93fac32f0bae089a9148d7cb90b3e83e446cd4b8e93cb86.jpg)  
+Figure 4: Two cases generated by RADAR, where false positive observation appears in the PF of case A and false negative observation shows in the PF of case B.
+
+Analysis of RADAR versus RAG. To evaluate the effectiveness of knowledge integration in RADAR, we conduct experiments comparing our model against three baselines: (1) BACKBONE PLUS, (2) BACKBONE+RAG, and (3) BACKBONE+PF+SF. The results are presented in Figure 3. Note that these baselines do not include the OI. Since RADAR undergoes two-stage training (i.e., two additional epochs), we apply the same extended training to BACKBONE, referring to this variant as BACKBONE PLUS. In addition, we introduce a standard RAG baseline (BACKBONE+RAG), which utilizes the same retrieved findings as RADAR. Building upon this baseline, BACKBONE+PF+SF further includes PF as context. Our findings reveal that while all four models achieve comparable performance on lexical metrics (e.g.,  $50\% / 26\%$  B-1/4), they differ in clinical metrics. Specifically, BACKBONE+RAG and BACKBONE PLUS show similar performance, and BACKBONE+FP+SF outperforms these two baselines on CheXpert metrics and exhibits similar performance on RadGraph metrics. This demonstrates that incorporating credible knowledge can effectively enhance report generation even without OI. Moreover, RADAR demonstrates a relative improvement of over  $6\%$  across four key CheXpert metrics. This suggests that structured integration of internal and external knowledge contributes to its enhanced clinical accuracy.
+
+Analysis of Fine-tuning Different Modules in BACKBONE. To assess the contributions of different components in the base model (i.e., BLIP-3), we conduct an ablation study on the impact of fine-tuning the vision encoder, the Resampler, and the LLM. The results are summarized in Table 5. By comparing BACKBONE and BACKBONE-V2, we find that fine-tuning the vision encoder to incor
+
+porate domain-specific knowledge is crucial for achieving high clinical accuracy, even though both configurations exhibit strong language coverage in lexical metrics. Furthermore, fine-tuning the LLM (i.e., Phi-3) results in substantial improvements in both lexical and clinical metrics, as evidenced by the comparison between V1 and V2. This highlights the importance of adapting the LLM to the clinical domain for optimal performance. Notably, RADAR utilizes a 3.8B LLM as the decoder and outperforms many larger models (e.g., LLaVA-Med and MAIRA-1).
+
+# 5.2 Qualitative Analysis
+
+Case Study. We conduct a case study to illustrate the advantages of incorporating both internal knowledge and retrieved information, as shown in Figure 4. In Case A, RADAR initially generates a report that includes the finding Atelectasis. However, expert assessment indicates the image shows no positive findings. As a result, their intersection is  $\emptyset$ , and by removing this incorrect observation, RADAR ultimately produces an accurate report. This example highlights the model's ability to refine its predictions when guided by expert constraints, effectively eliminating unnecessary or incorrect findings. Another more complex case is presented on the right side of this Figure. Specifically, RADAR initially identifies findings related to Edema and Cardiomegaly, which the expert model also notes. However, the observation of Atelectasis is omitted from the preliminary findings. By incorporating retrieved evidence such as "... linear atelectasis ..." and "Mild areas of atelectasis ..., RADAR successfully corrects the omission and generates a complete and accurate report. This case demonstrates the model's capability to lever
+
+Indication: History of chest pain x20 minutes, please evaluate for pneumonia or other acute process.
+
+![](images/724da2b10707f9346054e146f7a594f04a1a7d6c2a73506e65479f6d8094851d.jpg)  
+Prrr
+
+![](images/5e9aceed92eb7da758273ff537c48601393f925bc3e690344b0a5c6475b85156.jpg)  
+Curren
+
+![](images/cb741168ac45eb82599d88baf3e399ce9414ba9e25c83d544e18bbdd275895f2.jpg)  
+Figure 5: Error case generated by RADAR, where spans and spans indicate incorrect and correct observations.
+
+age external knowledge to recover missing findings, thereby improving factual completeness.
+
+Error Analysis. We conduct an error analysis to gain deeper insights, as shown in Figure 5. RADAR initially generates a report containing the observation Cardiomegaly, which is also present in the expert model's output. In this case, the observation reflects credible knowledge possessed by the LLM and should be preserved. Subsequently, RADAR produces a false positive finding, Edema, which aligns with the retrieved supplementary findings. This error may result from the model's overreliance on external knowledge. Moreover, since Edema is clinically associated with Cardiomegaly, it is possible that RADAR has learned only superficial correlations between them. To address these issues, potential solutions include refining the expert model and expanding the training dataset.
+
+# 6 Related Works
+
+Radiology report generation (Jing et al., 2018; Li et al., 2018) is a valuable yet challenging task. Numerous research efforts have been dedicated to improving clinical accuracy, employing diverse approaches such as memory-based neural models (Chen et al., 2020, 2021), planning-based methods (Nishino et al., 2022; Hou et al., 2023b), and reinforcement learning-optimized techniques (Lovelace and Mortazavi, 2020; Miura et al., 2021; Qin and Song, 2022). Additionally, several studies (Ramesh et al., 2022; Bannur et al., 2023; Hou
+
+et al., 2023a; Dalla Serra et al., 2023; Hou et al., 2024) have addressed the issue of hallucination, particularly in the absence of prior studies. Given the critical role of domain knowledge in this field, researchers have leveraged knowledge graphs to enhance report generation (Yang et al., 2021; Li et al., 2023c; Huang et al., 2023; Yan et al., 2023).
+
+With the emergence of MLLMs (Li et al., 2023b; Liu et al., 2023), which demonstrate exceptional capabilities in image understanding and captioning, many studies (Singhal et al., 2022; Wu et al., 2023; Thawakar et al., 2024; Li et al., 2023a) have explored their application in the medical domain. Chen et al. (2024) introduced a foundation model for chest X-ray interpretation, while Chaves et al. (2024) developed a lightweight MLLM tailored for radiology. Park et al. (2024) investigated the multitask potential of LLMs, and Zhang et al. (2024) incorporated temporal information to enhance chest X-ray analysis.
+
+# 7 Conclusion
+
+In this paper, we introduce RADAR, a novel approach designed to enhance radiology report generation by leveraging both the internal knowledge of an LLM and externally retrieved information. Our model first generates a report and subsequently classifies the image based on observations, with their shared components regarded as internal knowledge. It then retrieves supplementary information to further refine and complement this knowledge. Extensive experiments on three public datasets demonstrate that RADAR achieves SOTA performance in both language quality and clinical accuracy, highlighting the effectiveness of integrating internal and external knowledge for more accurate and coherent radiology report generation.
+
+# Limitations
+
+Our experiments are conducted using a single backbone architecture. While this choice provides a controlled evaluation, the performance of alternative architectures remains unexplored. Future work should investigate whether different model architectures can achieve comparable or better results. In addition, our study focuses exclusively on a single imaging modality (e.g., Chest X-ray). The model's effectiveness in other imaging modalities, such as CT scans or MRI, has not been evaluated. Extending our approach to multiple imaging modalities would be an important direction for future research
+
+to enhance its clinical utility and generalizability.
+
+# Ethical Considerations
+
+This study utilizes the MIMIC-CXR (Johnson et al., 2019), IU X-RAY (Demner-Fushman et al., 2016), and CHEXPERT PLUS (Chambon et al., 2024) datasets, all of which are publicly available and have been automatically de-identified to mitigate privacy risks. Our primary objective is to improve the clinical accuracy of reports generated by LLMs in medical imaging. However, despite our efforts, the generated reports may contain inaccuracies or omissions. Therefore, these outputs should not be used as a substitute for expert medical judgment. We strongly advocate for thorough validation by qualified radiologists or healthcare professionals before any clinical or diagnostic application.
+
+# Acknowledgments
+
+This work was supported in part by the National Key R&D Program of China (2024YFE0198100, 2024YFC2510800), the Natural Science Foundation of China (82272086), and the Research Grants Council of Hong Kong (15207920, 15207821, 15207122).
+
+# References
+
+Marah Abdin, Jyoti Aneja, Hany Awadalla, Ahmed Awadallah, Ammar Ahmad Awan, Nguyen Bach, Amit Bahree, Arash Bakhtiari, Jianmin Bao, Harkirat Behl, Alon Benhaim, Misha Bilenko, Johan Bjorck, Sebastien Bubeck, Martin Cai, Qin Cai, Vishrav Chaudhary, Dong Chen, Dongdong Chen, Weizhu Chen, Yen-Chun Chen, Yi-Ling Chen, Hao Cheng, Parul Chopra, Xiyang Dai, Matthew Dixon, Ronen Eldan, Victor Fragoso, Jianfeng Gao, Mei Gao, Min Gao, Amit Garg, Allie Del Giorno, Abhishek Goswami, Suriya Gunasekar, Emman Haider, Junheng Hao, Russell J. Hewett, Wenxiang Hu, Jamie Huynh, Dan Iter, Sam Ade Jacobs, Mojan Javaheripi, Xin Jin, Nikos Karampatziakis, Piero Kauffmann, Mahoud Khademi, Dongwoo Kim, Young Jin Kim, Lev Kurilenko, James R. Lee, Yin Tat Lee, Yuanzhi Li, Yunsheng Li, Chen Liang, Lars Liden, Xihui Lin, Zeqi Lin, Ce Liu, Liyuan Liu, Mengchen Liu, Weishung Liu, Xiaodong Liu, Chong Luo, Piyush Madan, Ali Mahmoudzadeh, David Majercak, Matt Mazzola, Caio Cesar Teodoro Mendes, Arindam Mitra, Hardik Modi, Anh Nguyen, Brandon Norick, Barun Patra, Daniel Perez-Becker, Thomas Portet, Reid Pryzant, Heyang Qin, Marko Radmilac, Liliang Ren, Gustavo de Rosa, Corby Rosset, Sambudha Roy Olatunji Ruwase, Olli Saarikivi, Amin Saied, Adil Salim, Michael Santacroce, Shital Shah, Ning Shang Hiteshi Sharma, Yelong Shen, Swadheen Shukla, Xia
+
+Song, Masahiro Tanaka, Andrea Tupini, Praneetha Vaddamanu, Chunyu Wang, Guanhua Wang, Lijuan Wang, Shuohang Wang, Xin Wang, Yu Wang, Rachel Ward, Wen Wen, Philipp Witte, Haiping Wu, Xiaoxia Wu, Michael Wyatt, Bin Xiao, Can Xu, Jiahang Xu, Weijian Xu, Jilong Xue, Sonali Yadav, Fan Yang, Jianwei Yang, Yifan Yang, Ziyi Yang, Donghan Yu, Lu Yuan, Chenruidong Zhang, Cyril Zhang, Jianwen Zhang, Li Lyna Zhang, Yi Zhang, Yue Zhang, Yunan Zhang, and Xiren Zhou. 2024. Phi-3 technical report: A highly capable language model locally on your phone. Preprint, arXiv:2404.14219.
+
+Emily Alsentzer, John R. Murphy, Willie Boag, WeiHung Weng, Di Jin, Tristan Naumann, and Matthew B. A. McDermott. 2019. Publicly available clinical bert embeddings. Preprint, arXiv:1904.03323.
+
+Satanjeev Banerjee and Alon Lavie. 2005. ME-TEOR: An automatic metric for MT evaluation with improved correlation with human judgments. In Proceedings of the ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and/or Summarization, pages 65-72, Ann Arbor, Michigan. Association for Computational Linguistics.
+
+Shruthi Bannur, Kenya Bouzid, Daniel C. Castro, Anton Schwaighofer, Anja Thieme, Sam Bond-Taylor, Maximilian Ilse, Fernando Pérez-García, Valentina Salvatelli, Harshita Sharma, Felix Meissen, Mercy Ranjit, Shaury Srivastav, Julia Gong, Noel C. F. Codella, Fabian Falck, Ozan Oktay, Matthew P. Lungren, Maria Teodora Wetscherek, Javier Alvarez-Valle, and Stephanie L. Hyland. 2024. Maira-2: Grounded radiology report generation. Preprint, arXiv:2406.04449.
+
+Shruthi Bannur, Stephanie Hyland, Qianchu Liu, Fernando Pérez-García, Maximilian Ilse, Daniel C. Castro, Benedikt Boecking, Harshita Sharma, Kenya Bouzid, Anja Thieme, Anton Schwaighofer, Maria Wetscherek, Matthew P. Lungren, Aditya Nori, Javier Alvarez-Valle, and Ozan Oktay. 2023. Learning to exploit temporal structure for biomedical vision-language processing. Preprint, arXiv:2301.04558.
+
+Pierre Chambon, Jean-Benoit Delbrouck, Thomas Sounack, Shih-Cheng Huang, Zhihong Chen, Maya Varma, Steven QH Truong, Chu The Chuong, and Curtis P. Langlotz. 2024. Chexpert plus: Augmenting a large chest x-ray dataset with text radiology reports, patient demographics and additional image formats. Preprint, arXiv:2405.19538.
+
+Juan Manuel Zambrano Chaves, Shih-Cheng Huang, Yanbo Xu, Hanwen Xu, Naoto Usuyama, Sheng Zhang, Fei Wang, Yujia Xie, Mahmoud Khademi, Ziyi Yang, Hany Awadalla, Julia Gong, Houdong Hu, Jianwei Yang, Chunyuan Li, Jianfeng Gao, Yu Gu, Cliff Wong, Mu Wei, Tristan Naumann, Muhao Chen, Matthew P. Lungren, Akshay Chaudhari, Serena Yeung-Levy, Curtis P. Langlotz, Sheng Wang, and Hoifung Poon. 2024. Towards a clinically accessible radiology foundation model: open-access and lightweight, with automated evaluation. Preprint, arXiv:2403.08002.
+
+Zhihong Chen, Yaling Shen, Yan Song, and Xiang Wan. 2021. Cross-modal memory networks for radiology report generation. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, ACL/IJCNLP 2021, (Volume 1: Long Papers), Virtual Event, August 1-6, 2021, pages 5904-5914. Association for Computational Linguistics.  
+Zhihong Chen, Yan Song, Tsung-Hui Chang, and Xi-ang Wan. 2020. Generating radiology reports via memory-driven transformer. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1439-1449, Online. Association for Computational Linguistics.  
+Zhihong Chen, Maya Varma, Jean-Benoit Delbrouck, Magdalini Paschali, Louis Blankemeier, Dave Van Veen, Jeya Maria Jose Valanarasu, Alaa Youssef, Joseph Paul Cohen, Eduardo Pontes Reis, Emily B. Tsai, Andrew Johnston, Cameron Olsen, Tanishq Mathew Abraham, Sergios Gatidis, Akshay S. Chaudhari, and Curtis Langlotz. 2024. Chexagent: Towards a foundation model for chest x-ray interpretation. Preprint, arXiv:2401.12208.  
+Francesco Dalla Serra, Chaoyang Wang, Fani Deligianni, Jeff Dalton, and Alison O'Neil. 2023. Controllable chest X-ray report generation from longitudinal representations. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 4891-4904, Singapore. Association for Computational Linguistics.  
+Dina Demner-Fushman, Marc D Kohli, Marc B Rosenman, Sonya E Shooshan, Laritza Rodriguez, Sameer Antani, George R Thoma, and Clement J McDonald. 2016. Preparing a collection of radiology examinations for distribution and retrieval. Journal of the American Medical Informatics Association, 23(2):304-310.  
+Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4171-4186, Minneapolis, Minnesota. Association for Computational Linguistics.  
+Wenjun Hou, Yi Cheng, Kaishuai Xu, Yan Hu, Wenjie Li, and Jiang Liu. 2024. ICON: Improving interreport consistency in radiology report generation via lesion-aware mixup augmentation. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 9043-9056, Miami, Florida, USA. Association for Computational Linguistics.  
+Wenjun Hou, Yi Cheng, Kaishuai Xu, Wenjie Li, and Jiang Liu. 2023a. RECAP: Towards precise radiology report generation via dynamic disease progression reasoning. In Findings of the Association
+
+for Computational Linguistics: EMNLP 2023, pages 2134-2147, Singapore. Association for Computational Linguistics.  
+Wenjun Hou, Kaishuai Xu, Yi Cheng, Wenjie Li, and Jiang Liu. 2023b. ORGAN: Observation-guided radiology report generation via tree reasoning. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 8108-8122, Toronto, Canada. Association for Computational Linguistics.  
+Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qianglong Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, and Ting Liu. 2025. A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. ACM Transactions on Information Systems, 43(2):1-55.  
+Zhongzhen Huang, Xiaofan Zhang, and Shaoting Zhang. 2023. Kiut: Knowledge-injected u-transformer for radiology report generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 19809-19818.  
+Stephanie L. Hyland, Shruthi Bannur, Kenya Bouzid, Daniel C. Castro, Mercy Ranjit, Anton Schwaighofer, Fernando Pérez-García, Valentina Salvatelli, Shaury Srivastav, Anja Thieme, Noel Codella, Matthew P. Lungren, Maria Teodora Wetscherek, Ozan Oktay, and Javier Alvarez-Valle. 2024. Maira-1: A specialised large multimodal model for radiology report generation. Preprint, arXiv:2311.13668.  
+Jeremy Irvin, Pranav Rajpurkar, Michael Ko, Yifan Yu, Silviana Ciurea-Ilcus, Chris Chute, Henrik Marklund, Behzad Haghgoo, Robyn L. Ball, Katie S. Shpanskaya, Jayne Seekins, David A. Mong, Safwan S. Halabi, Jesse K. Sandberg, Ricky Jones, David B. Larson, Curtis P. Langlotz, Bhavik N. Patel, Matthew P. Lungren, and Andrew Y. Ng. 2019. Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison. In The Thirty-Third AAAI Conference on Artificial Intelligence, pages 590-597.  
+Saahil Jain, Ashwin Agrawal, Adriel Saporta, Steven Q. H. Truong, Du Nguyen Duong, Tan Bui, Pierre Chambon, Yuhao Zhang, Matthew P. Lungren, Andrew Y. Ng, Curtis P. Langlotz, and Pranav Rajpurkar. 2021. Radgraph: Extracting clinical entities and relations from radiology reports. CoRR, abs/2106.14463.  
+Haibo Jin, Haoxuan Che, Yi Lin, and Hao Chen. 2024. Promptmrg: Diagnosis-driven prompts for medical report generation. Preprint, arXiv:2308.12604.  
+Baoyu Jing, Pengtao Xie, and Eric P. Xing. 2018. On the automatic generation of medical imaging reports. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics, ACL 2018, Melbourne, Australia, July 15-20, 2018, Volume 1: Long Papers, pages 2577-2586. Association for Computational Linguistics.
+
+Alistair EW Johnson, Tom J Pollard, Nathaniel R Greenbaum, Matthew P Lungren, Chih-ying Deng, Yifan Peng, Zhiyong Lu, Roger G Mark, Seth J Berkowitz, and Steven Horng. 2019. Mimic-cxr-jpg, a large publicly available database of labeled chest radiographs. arXiv preprint arXiv:1901.07042.  
+Chunyuan Li, Cliff Wong, Sheng Zhang, Naoto Usuyama, Haotian Liu, Jianwei Yang, Tristan Naumann, Hoifung Poon, and Jianfeng Gao. 2023a. Llava-med: Training a large language-and-vision assistant for biomedicine in one day. Preprint, arXiv:2306.00890.  
+Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. 2023b. Blip-2: Bootstrapping language-image pretraining with frozen image encoders and large language models. Preprint, arXiv:2301.12597.  
+Mingjie Li, Bingqian Lin, Zicong Chen, Haokun Lin, Xiaodan Liang, and Xiaojun Chang. 2023c. Dynamic graph enhanced contrastive learning for chest x-ray report generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3334-3343.  
+Yuan Li, Xiaodan Liang, Zhiting Hu, and Eric P. Xing. 2018. Hybrid retrieval-generation reinforced agent for medical image report generation. In Advances in Neural Information Processing Systems 31: Annual Conference on Neural Information Processing Systems 2018, NeurIPS 2018, December 3-8, 2018, Montréal, Canada, pages 1537-1547.  
+Chin-Yew Lin. 2004. ROUGE: A package for automatic evaluation of summaries. In Text Summarization Branches Out, pages 74-81, Barcelona, Spain. Association for Computational Linguistics.  
+Chang Liu, Yuanhe Tian, Weidong Chen, Yan Song, and Yongdong Zhang. 2024. Bootstrapping large language models for radiology report generation. In AAAI, pages 18635-18643.  
+Fenglin Liu, Xian Wu, Shen Ge, Wei Fan, and Yuexian Zou. 2021. Exploring and distilling posterior and prior knowledge for radiology report generation. In IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021, pages 13753-13762. Computer Vision Foundation / IEEE.  
+Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. 2023. Visual instruction tuning. Preprint, arXiv:2304.08485.  
+Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, and Baining Guo. 2022. Swin transformer v2: Scaling up capacity and resolution. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12009-12019.  
+Justin Lovelace and Bobak Mortazavi. 2020. Learning to generate clinically coherent chest X-ray reports.
+
+In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 1235-1243, Online. Association for Computational Linguistics.  
+Yasuhide Miura, Yuhao Zhang, Emily Tsai, Curtis Langlotz, and Dan Jurafsky. 2021. Improving factual completeness and consistency of image-to-text radiology report generation. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 5288-5304, Online. Association for Computational Linguistics.  
+Toru Nishino, Yasuhide Miura, Tomoki Taniguchi, Tomoko Ohkuma, Yuki Suzuki, Shoji Kido, and Noriyuki Tomiyama. 2022. Factual accuracy is not enough: Planning consistent description order for radiology report generation. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, Online. Association for Computational Linguistics.  
+Farhad Nooralahzadeh, Nicolas Perez Gonzalez, Thomas Frauenfelder, Koji Fujimoto, and Michael Krauthammer. 2021. Progressive transformer-based generation of radiology reports. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 2824-2832, Punta Cana, Dominican Republic. Association for Computational Linguistics.  
+OpenAI. 2023. Gpt-4v(ision) system card.  
+Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 311-318, Philadelphia, Pennsylvania, USA. Association for Computational Linguistics.  
+Jonggwon Park, Soobum Kim, Byungmu Yoon, Jihun Hyun, and Kyoyun Choi. 2024. M4cxxr: Exploring multi-task potentials of multi-modal large language models for chest x-ray interpretation. Preprint, arXiv:2408.16213.  
+Chantal Pellegrini, Ege Özsoy, Benjamin Busam, Nassir Navab, and Matthias Keicher. 2023. Radialog: A large vision-language model for radiology report generation and conversational assistance. Preprint, arXiv:2311.18681.  
+Han Qin and Yan Song. 2022. Reinforced cross-modal alignment for radiology report generation. In Findings of the Association for Computational Linguistics: ACL 2022, Dublin, Ireland, May 22-27, 2022, pages 448-458. Association for Computational Linguistics.  
+Vignav Ramesh, Nathan Andrew Chi, and Pranav Rajpurkar. 2022. Improving radiology report generation systems by removing hallucinated references to nonexistent priors. Preprint, arXiv:2210.06340.  
+Mercy Ranjit, Gopinath Ganapathy, Ranjit Manuel, and Tanuja Ganu. 2023. Retrieval augmented chest x-ray
+
+report generation using openai gpt models. Preprint, arXiv:2305.03660.  
+Karan Singhal, Shekoofeh Azizi, Tao Tu, S. Sara Mahdavi, Jason Wei, Hyung Won Chung, Nathan Scales, Ajay Tanwani, Heather Cole-Lewis, Stephen Pfohl, Perry Payne, Martin Seneviratne, Paul Gamble, Chris Kelly, Nathaneal Scharli, Aakanksha Chowdhery, Philip Mansfield, Blaise Aguera y Arcas, Dale Webster, Greg S. Corrado, Yossi Matias, Katherine Chou, Juraj Gottweis, Nenad Tomasev, Yun Liu, Alvin Rajkomar, Joelle Barral, Christopher Semturs, Alan Karthikesalingam, and Vivek Natarajan. 2022. Large language models encode clinical knowledge. Preprint, arXiv:2212.13138.  
+Akshay Smit, Saahil Jain, Pranav Rajpurkar, Anuj Parek, Andrew Ng, and Matthew Lungren. 2020. Combining automatic labelers and expert annotations for accurate radiology report labeling using BERT. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1500-1519, Online. Association for Computational Linguistics.  
+Xiao Song, Xiaodan Zhang, Junzhong Ji, Ying Liu, and Pengxu Wei. 2022. Cross-modal contrastive attention model for medical report generation. In Proceedings of the 29th International Conference on Computational Linguistics, pages 2388-2397, Gyeongju, Republic of Korea. International Committee on Computational Linguistics.  
+Liwen Sun, James Jialun Zhao, Wenjing Han, and Chenyan Xiong. 2025. Fact-aware multimodal retrieval augmentation for accurate medical radiology report generation. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 643-655, Albuquerque, New Mexico. Association for Computational Linguistics.  
+Tim Tanida, Philip Müller, Georgios Kaissis, and Daniel Rueckert. 2023. Interactive and explainable region-guided radiology report generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 7433-7442.  
+Omkar Chakradhar Thawakar, Abdelrahman M. Shaker, Sahal Shaji Mullappilly, Hisham Cholakkal, Rao Muhammad Anwer, Salman Khan, Jorma Laaksonen, and Fahad Khan. 2024. XrayGPT: Chest radiographs summarization using large medical vision-language models. In Proceedings of the 23rd Workshop on Biomedical Natural Language Processing, pages 440-448, Bangkok, Thailand. Association for Computational Linguistics.  
+Zhanyu Wang, Lingqiao Liu, Lei Wang, and Luping Zhou. 2023a. Metransformer: Radiology report generation by transformer with multiple learnable expert tokens. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 11558-11567.
+
+Zhanyu Wang, Lingqiao Liu, Lei Wang, and Luping Zhou. 2023b. R2gengpt: Radiology report generation with frozen llms. Preprint, arXiv:2309.09812.  
+Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumont, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mariama Drame, Quentin Lhoest, and Alexander Rush. 2020. Transformers: State-of-the-art natural language processing. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 38-45, Online. Association for Computational Linguistics.  
+Chaoyi Wu, Xiaoman Zhang, Ya Zhang, Yanfeng Wang, and Weidi Xie. 2023. Towards generalist foundation model for radiology by leveraging web-scale 2d&3d medical data. Preprint, arXiv:2308.02463.  
+Le Xue, Manli Shu, Anas Awadalla, Jun Wang, An Yan, Senthil Purushwalkam, Honglu Zhou, Viraj Prabhu, Yutong Dai, Michael S Ryoo, Shrikant Kendre, Jieyu Zhang, Can Qin, Shu Zhang, Chia-Chih Chen, Ning Yu, Juntao Tan, Tulika Manoj Awalgaonkar, Shelby Heinecke, Huan Wang, Yejin Choi, Ludwig Schmidt, Zeyuan Chen, Silvio Savarese, Juan Carlos Niebles, Caiming Xiong, and Ran Xu. 2024. xgen-mm (blip-3): A family of open large multimodal models. Preprint, arXiv:2408.08872.  
+Benjamin Yan, Ruochen Liu, David Kuo, Subathra Adithan, Eduardo Reis, Stephen Kwak, Vasantha Venugopal, Chloe O'Connell, Agustina Saenz, Pranav Rajpurkar, and Michael Moor. 2023. Style-aware radiology report generation with RadGraph and few-shot prompting. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 14676-14688, Singapore. Association for Computational Linguistics.  
+Shuxin Yang, Xian Wu, Shen Ge, Shaohua Kevin Zhou, and Li Xiao. 2021. Knowledge matters: Radiology report generation with general and specific knowledge. CoRR, abs/2112.15009.  
+Feiyang Yu, Mark Endo, Rayan Krishnan, Ian Pan, Andy Tsai, Eduardo Pontes Reis, Eduardo Kaiser Ururahy Nunes Fonseca, Henrique Min Ho Lee, Zahra Shakeri Hossein Abad, Andrew Y. Ng, Curtis P. Langlotz, Vasantha Kumar Venugopal, and Pranav Rajpurkar. 2022. Evaluating progress in automatic chest x-ray radiology report generation. medRxiv.  
+Xiaohua Zhai, Basil Mustafa, Alexander Kolesnikov, and Lucas Beyer. 2023. Sigmoid loss for language image pre-training. Preprint, arXiv:2303.15343.  
+Xi Zhang, Zaiqiao Meng, Jake Lever, and Edmond S. L. Ho. 2024. Libra: Leveraging temporal images for biomedical radiology analysis. Preprint, arXiv:2411.19378.
+
+Hong-Yu Zhou, Subathra Adithan, Julián Nicolas Acosta, Eric J. Topol, and Pranav Rajpurkar. 2024. A generalist learner for multifaceted medical image interpretation. Preprint, arXiv:2405.07988.
+
+# A Appendix
+
+# A.1 Full List of Specialists
+
+In addition to specialist baselines in Table 1, the following baselines are included: R2GEN (Chen et al., 2020), R2GENCMN (Chen et al., 2021),  $\mathcal{M}^2\mathrm{TR}$  (Nooralahzadeh et al., 2021), KNOWMAT (Yang et al., 2021), CMM-RL (Qin and Song, 2022), CMCA (Song et al., 2022), KiUT (Huang et al., 2023), DCL (Li et al., 2023c), METrans (Wang et al., 2023a), RGRG (Tanida et al., 2023), RECAP (Hou et al., 2023a), Controllable (Dalla Serra et al., 2023), PromptMRG (Jin et al., 2024), and ICON (Hou et al., 2024).
+
+<table><tr><td>Observation</td><td>P</td><td>R</td><td>F1</td></tr><tr><td>Atelectasis</td><td>0.518</td><td>0.645</td><td>0.574</td></tr><tr><td>Cardiomegaly</td><td>0.656</td><td>0.783</td><td>0.713</td></tr><tr><td>Consolidation</td><td>0.370</td><td>0.174</td><td>0.237</td></tr><tr><td>Edema</td><td>0.518</td><td>0.610</td><td>0.560</td></tr><tr><td>Pleural Effusion</td><td>0.695</td><td>0.800</td><td>0.744</td></tr><tr><td>5Macro Average</td><td>0.551</td><td>0.602</td><td>0.567</td></tr><tr><td>5Micro Average</td><td>0.607</td><td>0.707</td><td>0.653</td></tr><tr><td>Enlarged Card.</td><td>0.277</td><td>0.204</td><td>0.235</td></tr><tr><td>Lung Opacity</td><td>0.644</td><td>0.496</td><td>0.561</td></tr><tr><td>Lung Lesion</td><td>0.492</td><td>0.207</td><td>0.291</td></tr><tr><td>Pneumonia</td><td>0.283</td><td>0.232</td><td>0.255</td></tr><tr><td>Pneumothorax</td><td>0.407</td><td>0.636</td><td>0.496</td></tr><tr><td>Pleural Other</td><td>0.333</td><td>0.173</td><td>0.228</td></tr><tr><td>Fracture</td><td>0.421</td><td>0.244</td><td>0.309</td></tr><tr><td>Support Devices</td><td>0.823</td><td>0.866</td><td>0.844</td></tr><tr><td>No Finding</td><td>0.302</td><td>0.569</td><td>0.395</td></tr><tr><td>14Macro Average</td><td>0.481</td><td>0.474</td><td>0.460</td></tr><tr><td>14Micro Average</td><td>0.614</td><td>0.640</td><td>0.627</td></tr></table>
+
+Table 7: Experimental results of RADAR for each observation on the MIMIC-CXR dataset.
+
+Dataset: MIMIC-CXR (Compared with SOTA Specialists)  
+
+<table><tr><td rowspan="2">Model</td><td colspan="6">Lexical Metrics</td><td colspan="3">CE (14Macro) Metrics</td></tr><tr><td>B-1</td><td>B-2</td><td>B-3</td><td>B-4</td><td>MTR</td><td>R-L</td><td>P</td><td>R</td><td>F1</td></tr><tr><td>R2GEN</td><td>0.353</td><td>0.218</td><td>0.145</td><td>0.103</td><td>0.142</td><td>0.270</td><td>0.333</td><td>0.273</td><td>0.276</td></tr><tr><td>R2GENCMN</td><td>0.353</td><td>0.218</td><td>0.148</td><td>0.106</td><td>0.142</td><td>0.278</td><td>0.344</td><td>0.275</td><td>0.278</td></tr><tr><td>M2TR</td><td>0.378</td><td>0.232</td><td>0.154</td><td>0.107</td><td>0.145</td><td>0.272</td><td>0.240</td><td>0.428</td><td>0.308</td></tr><tr><td>KNOWMAT</td><td>0.363</td><td>0.228</td><td>0.156</td><td>0.115</td><td>-</td><td>0.284</td><td>0.458</td><td>0.348</td><td>0.371</td></tr><tr><td>CMM-RL</td><td>0.381</td><td>0.232</td><td>0.155</td><td>0.109</td><td>0.151</td><td>0.287</td><td>0.342</td><td>0.294</td><td>0.292</td></tr><tr><td>CMCA</td><td>0.360</td><td>0.227</td><td>0.156</td><td>0.117</td><td>0.148</td><td>0.287</td><td>0.444</td><td>0.297</td><td>0.356</td></tr><tr><td>KiUT</td><td>0.393</td><td>0.243</td><td>0.159</td><td>0.113</td><td>0.160</td><td>0.285</td><td>0.371</td><td>0.318</td><td>0.321</td></tr><tr><td>DCL</td><td>-</td><td>-</td><td>-</td><td>0.109</td><td>0.150</td><td>0.284</td><td>0.471</td><td>0.352</td><td>0.373</td></tr><tr><td>METrans</td><td>0.386</td><td>0.250</td><td>0.169</td><td>0.124</td><td>0.152</td><td>0.291</td><td>0.364</td><td>0.309</td><td>0.311</td></tr><tr><td>RGRG</td><td>0.373</td><td>0.249</td><td>0.175</td><td>0.126</td><td>0.168</td><td>0.264</td><td>0.380</td><td>0.319</td><td>0.305</td></tr><tr><td>ORGAN</td><td>0.407</td><td>0.256</td><td>0.172</td><td>0.123</td><td>0.162</td><td>0.293</td><td>0.416</td><td>0.418</td><td>0.385</td></tr><tr><td>RECAP</td><td>0.429</td><td>0.267</td><td>0.177</td><td>0.125</td><td>0.168</td><td>0.288</td><td>0.389</td><td>0.443</td><td>0.393</td></tr><tr><td>Controllable</td><td>0.486</td><td>0.366</td><td>0.295</td><td>0.246</td><td>0.216</td><td>0.423</td><td>0.597</td><td>0.516</td><td>0.553</td></tr><tr><td>PromptMRG</td><td>0.398</td><td>-</td><td>-</td><td>0.112</td><td>0.157</td><td>0.268</td><td>0.396</td><td>0.393</td><td>0.381</td></tr><tr><td>ICON</td><td>0.429</td><td>0.266</td><td>0.178</td><td>0.126</td><td>0.170</td><td>0.287</td><td>0.445</td><td>0.505</td><td>0.464</td></tr><tr><td>RADAR (Ours)</td><td>0.509</td><td>0.390</td><td>0.315</td><td>0.262</td><td>0.450</td><td>0.397</td><td>0.481</td><td>0.474</td><td>0.460</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td>0.523</td><td>0.500</td><td>0.497</td></tr></table>
+
+Table 8: Experimental results of our model and SoTA specialists on the MIMIC-CXR dataset. Results denotes Uncertain as Positive.
+
+<table><tr><td>Role</td><td>Prompt</td></tr><tr><td>SYSTEM</td><td>&lt;|system|&gt;You are an assistant in radiology, responsible for analyzing medical imaging studies and generating detailed, structured, and accurate radiology reports.&lt;|end|&gt;</td></tr><tr><td>USER</td><td>&lt;|user|&gt;&lt;prior image&gt;(If prior available)&lt;current image&gt;Indication: ....History: ....Comparison: ....Technique: ....Prior Findings: ....(If prior available)Preliminary Findings: ....(If available)Supplementary Findings: ....(If available)Generate a comprehensive and detailed description of findings based on this chest X-ray image. Include a thorough comparison with a prior chest X-ray, emphasizing any significant changes, progression, or improvement. (If prior available) Before this, systematically identify all observations.&lt;|end|&gt;</td></tr><tr><td>ASSISTANT</td><td>&lt;|assisitant|&gt;Identified Observations:......Overall Findings: (e.g., the target)......&lt;|end|&gt;</td></tr></table>
+
+Table 9: The prompt template for RADAR and its variants, consisting of three roles: System, User, and Assistant.

radarenhancingradiologyreportgenerationwithsupplementaryknowledgeinjection/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71c8c856b90c9ef07c7e1dd1b71b3f2cb3f926a03df0bf68fce62aa4f87d8079
+size 951586

radarenhancingradiologyreportgenerationwithsupplementaryknowledgeinjection/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97ab6ce0a0813dd257e114ccd25cdae5b16ed7127fac64201750ece042819b80
+size 502989

raemollmretrievalaugmentedllmsforcrossdomainmisinformationdetectionusingincontextlearningbasedonemotionalinformation/3e2a23cb-d0ef-471f-b18b-26c80a6d621e_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b5067ba681d75d61cdd2ea2e797a2491c31fbab0f40e7b46ee5b97f0bcf77c4
+size 124143

raemollmretrievalaugmentedllmsforcrossdomainmisinformationdetectionusingincontextlearningbasedonemotionalinformation/3e2a23cb-d0ef-471f-b18b-26c80a6d621e_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63acfd6945df36199576a6e309e2b261b742d940e2fd231aa38be80ef52c6827
+size 145651

raemollmretrievalaugmentedllmsforcrossdomainmisinformationdetectionusingincontextlearningbasedonemotionalinformation/3e2a23cb-d0ef-471f-b18b-26c80a6d621e_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e3d834392540d9fee8d1d267958e1b5c19c13f9752a70b450bb9313d300b24c
+size 5659088

raemollmretrievalaugmentedllmsforcrossdomainmisinformationdetectionusingincontextlearningbasedonemotionalinformation/full.md

@@ -0,0 +1,413 @@
+# RAEmoLLM: Retrieval Augmented LLMs for Cross-Domain Misinformation Detection Using In-Context Learning Based on Emotional Information
+
+Zhiwei Liu $^{1}$  Kailai Yang $^{1}$  Qianqian Xie $^{2}$  Christine de Kock $^{3}$  Sophia Ananiadou $^{1*}$  Eduard Hovy $^{3,4}$
+
+<sup>1</sup> The University of Manchester <sup>2</sup> School of Artificial Intelligence, Wuhan University <sup>3</sup> The University of Melbourne <sup>4</sup> Carnegie Mellon University {zhiwei.liu,kailai.yang,sophia.ananiadou}@manchester.ac.uk xqq.sincere@gmail.com, {christine.dekock,eduard.hovy}@unimelb.edu.au
+
+# Abstract
+
+Misinformation is prevalent in various fields such as education, politics, health, etc., causing significant harm to society. However, current methods for cross-domain misinformation detection rely on effort- and resource-intensive fine-tuning and complex model structures. With the outstanding performance of LLMs, many studies have employed them for misinformation detection. Unfortunately, they focus on in-domain tasks and do not incorporate significant sentiment and emotion features (which we jointly call affect). In this paper, we propose RAEmoLLM, the first retrieval augmented (RAG) LLMs framework to address cross-domain misinformation detection using in-context learning based on affective information. RAEmoLLM includes three modules. (1) In the index construction module, we apply an emotional LLM to obtain affective embeddings from all domains to construct a retrieval database. (2) The retrieval module uses the database to recommend top K examples (text-label pairs) from source domain data for target domain contents. (3) These examples are adopted as few-shot demonstrations for the inference module to process the target domain content. The RAEmoLLM can effectively enhance the general performance of LLMs in cross-domain misinformation detection tasks through affect-based retrieval, without fine-tuning. We evaluate our framework on three misinformation benchmarks. Results show that RAEmoLLM achieves significant improvements compared to the other few-shot methods on three datasets, with the highest increases of  $15.64\%$ ,  $31.18\%$ , and  $15.73\%$  respectively. This project is available at https://github.com/lzw108/RAEmoLLM.
+
+# 1 Introduction
+
+The internet is flooded with misinformation (Scheuftele and Krause, 2019), which has a sig-
+
+nificant impact on people's lives and societal stability (Della Giustina, 2023). Misinformation is pervasive across various domains such as education, health, technology, and especially on the internet, which requires people to invest significant time and effort in discerning the truth (Pérez-Rosas et al., 2018). However, models trained in specific known domains are often fragile and prone to making incorrect predictions when presented with samples from new domains (Saikh et al., 2020). As a result, detecting cross-domain misinformation has become an urgent global issue and poses greater challenges and difficulties.
+
+Although some studies address cross-domain misinformation detection (Comito et al., 2023; Tang et al., 2023; Shi et al., 2023), they require effort-intensive fine-tuning, and apply only traditional machine learning methods or complex deep learning methods. Recently, LLMs have achieved impressive results in various tasks through zero-shot, few-shot (Li, 2023), or instruction tuning (Zhang et al., 2023a). Many researchers have applied LLMs to identify misinformation (Li et al., 2023; Hu et al., 2024; Cheung and Lam, 2023). However, these methods perform only in-domain misinformation detection. Moreover, emotions and sentiments (which we jointly call affect) are important characteristics of human expression and communication (Hakak et al., 2017). When authors publish misinformation, they often consciously choose specific emotions to capture the attention and resonance of readers to encourage rapid spread (Keen, 2006; Liu et al., 2024d). Unfortunately, there are few LLMs that utilize affective information to detect misinformation, and the only ConspEmoLLM (Liu et al., 2024b) are developed based on an emotional LLM, which does not make full use of affective information, has no cross-domain ability, and also needs time-consuming fine-tuning.
+
+In-context learning (ICL) needs only task instructions and few-shot examples (input-label pairs),
+
+eliminating fine-tuning on specific task labels (Dong et al., 2022b). A few studies have used ICL to address cross-domain problems (Long et al., 2023; Wu et al., 2024). To the best of our knowledge, there is currently no application of ICL for cross-domain misinformation detection based on affective information retrieval.
+
+To address these issues, we propose the first retrieval augmented (RAG) LLMs framework based on emotional information (RAEmoLLM), to address cross-domain misinformation detection using in-context learning based on affective information. RAEmoLLM contains three modules: (1) In the index construction module, we apply EmoLLaMAchat-7B (Liu et al., 2024c) to encode all domain corpora, obtaining implicit affective embeddings to construct the retrieval database as well as explicit affective labels. We also conduct a comprehensive affective analysis to demonstrate the effectiveness of affective information for discriminating between true and misinformation. (2) The retrieval module recommends the top K affect-related examples (text-label pairs) from the source domain corpus according to the target domain content, obtained from the retrieval database. (3) These examples are utilized as the few-shot demonstrations in the inference module, which is driven by a prompt template to guide the LLM to verify the target content for misinformation. The template helps combine implicit and explicit affective information. This framework effectively enhances the capabilities of LLMs in multiple cross-domain misinformation detection tasks through leveraging affective information, without the need for fine-tuning. In this work, we make three main contributions:
+
+- We conduct affective analysis on different kinds of misinformation datasets and construct the retrieval database according to the implicit affective information for misinformation datasets.  
+- We propose RAEmoLLM, the first framework for cross-domain misinformation detection using ICL based on affective information, which does not require fine-tuning. Experimental results show that RAEmoLLM outperforms the zero-shot method and other few-shot methods.  
+- We evaluate RAEmoLLM on a variety of misinformation benchmarks, including fake news, rumours, and conspiracy theory datasets. Results show that RAEmoLLM achieves signifi
+
+cant improvements compared to the other few-shot methods on three datasets, with the highest increases of  $15.64\%$ ,  $31.18\%$ , and  $15.73\%$  respectively, which illustrate the effectiveness of RAEmoLLM framework.
+
+# 2 Methodology
+
+This section introduces our method of cross-domain misinformation detection, using the index construction module, retrieval module and inference module. The overall architecture of RAEMoLLM is shown in Figure 1. In the index construction module (Sec. 2.1), we collect domain datasets, and employ an emotional LLM to obtain affective embeddings as well as affective labels to conduct a comprehensive affective analysis on them to detect the affective differences between real and false information. The implicit embeddings are adopted to construct the retrieval database, which will be used by the retrieval module (Sec. 2.2) to obtain source-domain examples. These results are used as the few-shot examples for inference module's (Sec. 2.3) in-context learning to detect target domain misinformation.
+
+# 2.1 Index Construction Module
+
+In this section, we first introduce the original datasets and the processing procedure at Sec. 2.1.1. We subsequently conduct affective analysis on these datasets and present how and why to obtain implicit and explicit affective information at Sec. 2.1.2. Finally, we apply the implicit affective information to construct the retrieval database (Sec. 2.1.3).
+
+# 2.1.1 Datasets
+
+We collect FakeNewsAMT (Pérez-Rosas et al., 2018), Celebrity (Pérez-Rosas et al., 2018), PHEME (Kochkina et al., 2018), and COCO (Langguth et al., 2023) datasets. The statistics of these datasets are presented in Table 1. FakeNewsAMT is a cross-domain dataset, including six domains. The legitimate news in FakeNewsAMT was obtained from various mainstream news websites. The authors adopted crowdsourcing via Amazon Mechanical Turk (AMT) to generate fake versions of legitimate news items. The Celebrity dataset was derived from online magazines. We combine FakeNewsAMT and Celebrity as AMTCele. PHEME contains a collection of Twitter rumours and non-rumours posted during nine breaking news events. COCO dataset consists
+
+![](images/021c28d9ef99563d5c79a3b328b66555b9637c2952062a8a3d5bfdc7f6c7c7fa.jpg)  
+Figure 1: The architecture of RAEmoLLM. D: Domain. T: Target domain. S: Source domain. C: Corpus. L: Label. Aff: Affective information. M: Number of source domain data. Index Construction Module: Apply an emotional LLM to obtain affective embeddings to construct a retrieval database. Retrieval Module: Recommend top K examples (text-label pairs) from source domain data. Inference Module: Adopt the recommended examples as demonstrations for inference.
+
+<table><tr><td colspan="3">AMTCele</td><td colspan="3">PHEME</td><td colspan="3">COCO</td></tr><tr><td>Domain</td><td>Legit</td><td>Fake</td><td>Events</td><td>Rumours</td><td>Non-rumours</td><td>Topics</td><td>Related</td><td>Conspiracy</td></tr><tr><td>Technology</td><td>40</td><td>40</td><td>Charlie Hebdo</td><td>458</td><td>1621</td><td>Fake Virus</td><td></td><td></td></tr><tr><td>Education</td><td>40</td><td>40</td><td>Sydney siege</td><td>522</td><td>699</td><td>Harmful Radiation</td><td>248</td><td>612</td></tr><tr><td>Business</td><td>40</td><td>40</td><td>Ferguson</td><td>284</td><td>859</td><td>Depopulation</td><td></td><td></td></tr><tr><td>Sports</td><td>40</td><td>40</td><td>Ottawa shooting</td><td>470</td><td>420</td><td>Other 9 domains</td><td>540</td><td>1181</td></tr><tr><td>Politics</td><td>40</td><td>40</td><td>Germanwings-crash</td><td>238</td><td>231</td><td>Total</td><td>788</td><td>1793</td></tr><tr><td>Entertainment</td><td>40</td><td>40</td><td>Putin missing</td><td>126</td><td>112</td><td></td><td></td><td></td></tr><tr><td>Celebrities</td><td>250</td><td>250</td><td>Prince Toronto</td><td>229</td><td>4</td><td></td><td></td><td></td></tr><tr><td>Total</td><td>490</td><td>490</td><td>Gurlitt</td><td>61</td><td>77</td><td></td><td></td><td></td></tr><tr><td></td><td></td><td></td><td>Ebola Essien</td><td>14</td><td>0</td><td></td><td></td><td></td></tr><tr><td></td><td></td><td></td><td>Total</td><td>2402</td><td>4023</td><td></td><td></td><td></td></tr></table>
+
+Table 1: Statistics of datasets. AMTCele includes 7 domains. PHEME contains 9 domains (events). COCO has 12 domains (topics). For AMTCele and PHEME, we apply leave-one-domain-out strategy for evaluation. For COCO, we select 3 domains as test set.  
+
+<table><tr><td rowspan="2">Datasets</td><td rowspan="2">Affective</td><td rowspan="2">sub-emotion</td><td colspan="2">legit/non-rumours/related</td><td colspan="2">fake/rumours/conspiracy</td><td colspan="2">t-test</td></tr><tr><td>mean</td><td>var</td><td>mean</td><td>var</td><td>t</td><td>p</td></tr><tr><td rowspan="4">AMTCele</td><td rowspan="4">EIreg</td><td>Anger</td><td>0.3584</td><td>0.0064</td><td>0.4055</td><td>0.0060</td><td>-9.3294</td><td>6.91E-20</td></tr><tr><td>Fear</td><td>0.3587</td><td>0.0137</td><td>0.4047</td><td>0.0124</td><td>-6.2861</td><td>4.90E-10</td></tr><tr><td>Joy</td><td>0.3392</td><td>0.0180</td><td>0.2897</td><td>0.0142</td><td>6.1054</td><td>1.48E-09</td></tr><tr><td>Sadness</td><td>0.3341</td><td>0.0109</td><td>0.3697</td><td>0.0106</td><td>-5.3726</td><td>9.70E-08</td></tr><tr><td rowspan="3">PHEME COCO</td><td>Vreg</td><td>-</td><td>0.5471</td><td>0.0204</td><td>0.4940</td><td>0.0170</td><td>6.0656</td><td>1.88E-09</td></tr><tr><td>Elreg</td><td>Sadness</td><td>0.5215</td><td>0.0152</td><td>0.5177</td><td>0.0182</td><td>1.1442</td><td>0.2526</td></tr><tr><td>Vreg</td><td>-</td><td>0.3961</td><td>0.0095</td><td>0.3973</td><td>0.0066</td><td>-0.3325</td><td>0.7395</td></tr></table>
+
+Table 2: Statistics values of EReg and Vreg on different datasets. The t-test is conducted between legit/non-rumours/related and fake/rumours/conspiracy. The complete statistics on PHEME and COCO can be found in Table 13 in the Appendix G.
+
+of 12 conspiracy theory categories<sup>1</sup>. Each tweet in COCO is assigned an overall intention label, as follows: **Conspiracy** is assigned to tweets for which the tweet is related to at least one of the 12 categories and is actively spreading conspiracy theories. Otherwise, if the tweet is related to the specific category, but it does not propagate misinformation or conspiracy theories, then the overall label of Re
+
+lated is used. The overall label of Unrelated is only used for tweets that are unrelated to all 12 conspiracy categories. We remove the Unrelated text since the aim of the cross-domain test.
+
+For AMTCele and PHEME, we apply leave-one-domain-out strategy² to evaluate the model. For COCO dataset, due to one text data may involve
+
+one or multiple topics, we select all data points involving the Fake Virus, Harmful Radiation, and Depopulation topics as the test set, and the other topics as the retrieval dataset.
+
+# 2.1.2 Affective Analysis
+
+We firstly conduct a comprehensive affective analysis after collecting datasets. EmoLLaMA-chat-7B, which has the best overall performance among the EmoLLMs (Liu et al., 2024c), is used for affective analysis. EmoLLaMA-chat-7B can be used to extract five kinds of affective dimensions (which we jointly call affect), including Emotion intensity (EIreg), Emotion intensity classification (Eloc), Sentiment (valence) strength (Vreg), Sentiment (valence) classification (Voc) and Emotion detection (Ec). The detailed introduction can be found in Appendix G.1.
+
+Obtain implicit and explicit affective information: Following the guidelines of EmoLLMs (Liu et al., 2024c), we add prompts provided by EmoLLMs for each data point in order to obtain vectors from the last hidden layer (i.e., 4096d) for each affective dimension, as well as final labels using EmoLLaMA-chat-7B. We subsequently determine the distribution of affective information in different categories in each dataset.
+
+Explicit affective analysis: Table 2 and Table 13 show regression information (i.e., EReg and Vreg) of final labels. We use the t-test<sup>3</sup> to measure the difference in emotional intensity between two sets of data. The t-value and p-value calculated between legit/non-rumours/related and fake/rumours/conspiracy demonstrate that there are statistically significant affective differences between the different categories. Figure 3 to Figure 8 and the chi-squared test in Appendix G.2 confirm that other classifications using affective information are also related to misinformation. However, Table 2 also presents some special cases that cannot effectively distinguish real and false information (e.g. EReg-sadness in PHEME, Vreg in COCO). Liu et al. (2024b) also conducted some experiments that demonstrated that simply utilizing explicit affective information does not enhance the model's capability. Therefore, we introduce implicit affective information.
+
+Implicit affective analysis: Table 14 shows statistics of different affective embeddings (i.e. last hidden layer of EmoLLaMA-chat-7B). We perform t-tests on the top- $K$  cosine similarity within categories and across categories. The results indicate that the similarity within categories is significantly higher than across categories, confirming that similar top- $K$  data points are likely to belong to the same category (further analysis can be found in Appendix G.2). We also visualize the data distribution reduced to 3 dimensions using PCA in Figures 9 and 10 in Appendix. It can be observed that different categories are clearly separated in the latent space. All the above demonstrate the close relationship between affective information and misinformation.
+
+# 2.1.3 Retrieval Database Construction
+
+After obtaining the implicit affective embeddings in the previous step, we proceed to construct a comprehensive retrieval database. This database consists of vectors that encapsulate rich affective information, enabling efficient retrieval and analysis.
+
+# 2.2 Retrieval Module
+
+Algorithm 1 Retrieval process  
+Require: Target domain corpus  $D_{T}$ , source domain corpus  $D_{S}$ , retrieval database  $R$ .  
+Ensure: Target domain corpus with top K retrieval examples  $D_{retri}$ .  
+1:  $E_{T} \gets R(D_{T})$   
+2:  $E_{S} \gets R(D_{S})$   
+3: for  $e_t$  in  $E_{T}$  do  
+4: for  $e_s$  in  $E_{S}$  do  
+5: score = cosine  $(e_t, e_s)$   
+6: Sco ← score  
+7: end for  
+8:  $D_{retri} \gets$  select top k examples in  $R(D_{S})$  according to Sco  
+9: end for
+
+The retrieval database constructed in Sec 2.1 is represented as R. Algorithm 1 shows the retrieval process. In this module, we first process the multi-domain datasets into text-label pairs to obtain the target domain data  $D_{T} = \{[c_{t1},l_{t1}],\{c_{t2},l_{t2}\},\dots,\{c_{tN},l_{tN}\} \}$  and source domain data  $D_{S} = \{[c_{s1},l_{s1}],\{c_{s2},l_{s2}\},\dots,\{c_{sM},l_{sM}\} \}$  (c denotes corpus text, and  $l$  is the label.  $N$  and  $M$  are the numbers of target domain data and source domain data respectively). Following that, we obtain the target domain affective embedding  $E_{T} = [e_{t1},e_{t2},\dots,e_{tN}]$  and source domain
+
+affective embedding  $E_S = [e_{s1}, e_{s2}, \dots, e_{sM}]$  through the embedding retrieval database R based on the corpus texts in  $D_T$  and  $D_S$ . Subsequently, we traverse the target domain embedding  $(e_t)$  in  $E_T$  and calculate the similarity values with each source domain embedding  $e_s$  from  $E_S$  using the cosine method. Finally, we select the top k examples from source domain for each target domain data based on Sco to  $D_{retri}$ , which will be the few-shot examples for LLM inference.
+
+# 2.3 Inference Module
+
+We apply template 1 to construct the instruction datasets for inference once we get the top examples for each target domain data. [task prompt] denotes the instruction for the task (The different [task prompts] for each datasets can be found in Appendix B). [input text] is a data item from the target domain data. [examples] are the retrieval examples from source domain data (i.e.  $D_{\text{retri}}$ ) and the [output] is the output from LLM.
+
+# Template 1
+
+Task: [task prompt]
+
+Target text: [input text]
+
+Here are a few examples: [examples]
+
+According to the above information, the label of target
+
+text: [output]
+
+We also apply template 2 to add explicit affective information. [affective information] contains five dimensions described in Section 2.1.2. The format of [examples] is "Text: [text]. [Affective info]: [value]. The label of text: [label)". Table 6 shows one complete example.
+
+# Template 2
+
+Task: [task prompt]
+
+Target text: [input text] + [affective info]
+
+Here are a few examples retrieved by [affective info]: [examples]
+
+According to the above information, the label of target text: [output]
+
+# 3 Experiments
+
+# 3.1 Base Models
+
+- LLMs: We apply ChatGPT (gpt-3.5-turbo-0125), GPT-4o $^4$ , Llama3-8b-Instruct, Llama3.2-(1b,3b)-Instruct $^5$ , Gemma-instruct-(2b, 7b) (Team et al., 2024), Mistral-7b-Instruct (Jiang et al., 2023) and Vicuna-(7b, 13b, 33b) (Chiang et al., 2023) as base models to test our methods.  
+- PLMs: We select BERT (Devlin et al., 2018) and RoBERTa (Liu et al., 2019) as fine-tuning baselines. Specifically, one domain is selected as the target domain, other domains are used as the training dataset to fine-tune.
+
+<sup>4</sup>https://openai.com/  
+<sup>5</sup>https://llama.meta.com/llama3/
+
+- Domain generalization methods (DGMs): MOSE (Qin et al., 2020) is a multi-domain mixture-of-experts (MoE) model, and each domain has its specific head. EDDFN (Silva et al., 2021) preserves domain-specific and domain-shared knowledge. MDFEND (Nan et al., 2021) utilizes a Domain Gate to select useful experts of MoE. CANMD (Yue et al., 2022) performs label shift correction and contrastive learning. MetaAdapt (Yue et al., 2023) adopts a meta-learning approach for domain-adaptive few-shot misinformation detection.  
+- Retrieval method according to other types of embeddings: We use the last Hidden state of RoBERTa and another popular sentiment model (i.e. Sentibert (Yin et al., 2020)) as semantic and another kind of sentiment representation of each sentence respectively, then apply the same process of RAEmoLLM to deploy the ablation experiment.  
+- Zero-shot and few-shot methods: We also develop experiments of zero-shot method (LLMs-zs), randomly sample examples without using affective information (LLMs-random), and randomly sample examples with explicit Vreg information (LLMs-random-addexpl) for baselines.
+
+# 3.2 Evaluation Metric
+
+Misinformation detection is typically regarded as a classification task, therefore we employ a variety of metrics—Accuracy, Precision, Recall, and F1 for evaluation (Su et al., 2020) (All metrics use the weighted variant).
+
+# 3.3 Results
+
+We evaluate RAEmoLLM framework on one Nvidia Tesla A100 GPU with 80GB of memory. The max length of new tokens is 256 and do_sample is False. Others all use the default setting in the "model_generate" package. We firstly select the instruction data based on Vreg to test the effectiveness of the RAEmoLLM framework on different LLMs. The result is the overall performance, which means that in AMTCele and PHEME, every domain is considered as the target domain test set, and the overall result is the performance of the combination of each domain test set. For Gemma series, Llama series and Vicuna series, we only show the best overall performing one in the table. In this section, we will be discussing results exclusively based on the F1 score. We firstly compare the RAEmoLLM framework with various baseline methods (e.g. PLMs, domain generalization methods, zero-shot, and few-shot methods) at Sec. 3.3.1. The ablation study of each module is conducted at Sec. 3.3.2. We subsequently compare the results on the data retrieved based on different affective information at Sec. 3.3.3.
+
+<table><tr><td rowspan="2">Model</td><td colspan="4">AMTCele</td><td colspan="4">PHEME</td><td colspan="4">COCO</td></tr><tr><td>Acc</td><td>Pre</td><td>Rec</td><td>F1</td><td>Acc</td><td>Pre</td><td>Rec</td><td>F1</td><td>Acc</td><td>Pre</td><td>Rec</td><td>F1</td></tr><tr><td>BERT</td><td>0.5414</td><td>0.5453</td><td>0.5414</td><td>0.5322</td><td>0.7214</td><td>0.7203</td><td>0.7214</td><td>0.7208</td><td>0.7288</td><td>0.7510</td><td>0.7288</td><td>0.6356</td></tr><tr><td>RoBERTa</td><td>0.5678</td><td>0.7228</td><td>0.5678</td><td>0.4730</td><td>0.7199</td><td>0.7213</td><td>0.7199</td><td>0.7204</td><td>0.7328</td><td>0.7851</td><td>0.7328</td><td>0.6388</td></tr><tr><td>MDFEND</td><td>0.5878</td><td>0.5934</td><td>0.5878</td><td>0.5815</td><td>0.5796</td><td>0.6425</td><td>0.5796</td><td>0.5829</td><td>0.7988</td><td>0.7939</td><td>0.7988</td><td>0.7793</td></tr><tr><td>EDDFN</td><td>0.7041</td><td>0.7313</td><td>0.7041</td><td>0.6951</td><td>0.7004</td><td>0.6925</td><td>0.7004</td><td>0.6816</td><td>0.7116</td><td>0.5064</td><td>0.7116</td><td>0.5917</td></tr><tr><td>MOSE</td><td>0.5031</td><td>0.5051</td><td>0.5031</td><td>0.4482</td><td>0.7135</td><td>0.7130</td><td>0.7135</td><td>0.6890</td><td>0.7198</td><td>0.7335</td><td>0.7198</td><td>0.6162</td></tr><tr><td>CANMD</td><td>0.6296</td><td>0.6650</td><td>0.6296</td><td>0.6086</td><td>0.7382</td><td>0.7338</td><td>0.7382</td><td>0.7346</td><td>0.7291</td><td>0.7324</td><td>0.7291</td><td>0.6441</td></tr><tr><td>MetaAdapt</td><td>0.6429</td><td>0.6564</td><td>0.6429</td><td>0.6350</td><td>0.6193</td><td>0.6804</td><td>0.6193</td><td>0.6230</td><td>0.5186</td><td>0.7267</td><td>0.5186</td><td>0.5222</td></tr><tr><td>Mistral-7b-zs</td><td>0.7020</td><td>0.7346</td><td>0.7020</td><td>0.6926</td><td>0.5897</td><td>0.6491</td><td>0.5897</td><td>0.5936</td><td>0.3686</td><td>0.7050</td><td>0.3686</td><td>0.4673</td></tr><tr><td>Mistral-7b-random</td><td>0.7082</td><td>0.7768</td><td>0.7082</td><td>0.6889</td><td>0.6177</td><td>0.6334</td><td>0.6177</td><td>0.6227</td><td>0.7128</td><td>0.7455</td><td>0.7128</td><td>0.7287</td></tr><tr><td>Mistral-7b-random-addexpl</td><td>0.6337</td><td>0.7050</td><td>0.6337</td><td>0.5988</td><td>0.5804</td><td>0.6177</td><td>0.5804</td><td>0.5870</td><td>0.6802</td><td>0.7245</td><td>0.6802</td><td>0.7010</td></tr><tr><td>Mistral-7b-Vreg</td><td>0.7469</td><td>0.7748</td><td>0.7469</td><td>0.7404</td><td>0.6760</td><td>0.6837</td><td>0.6760</td><td>0.6788</td><td>0.7779</td><td>0.8031</td><td>0.7779</td><td>0.7898</td></tr><tr><td>Mistral-7b-Vreg-addexpl</td><td>0.7735</td><td>0.7822</td><td>0.7735</td><td>0.7717</td><td>0.6921</td><td>0.6919</td><td>0.6921</td><td>0.6920</td><td>0.7814</td><td>0.8053</td><td>0.7814</td><td>0.7931</td></tr><tr><td>Gemma-2b-zs</td><td>0.4153</td><td>0.4568</td><td>0.4153</td><td>0.3815</td><td>0.3606</td><td>0.5113</td><td>0.3606</td><td>0.2303</td><td>0.3302</td><td>0.4572</td><td>0.3302</td><td>0.3835</td></tr><tr><td>Gemma-2b-random</td><td>0.4980</td><td>0.4997</td><td>0.4980</td><td>0.4649</td><td>0.4269</td><td>0.5799</td><td>0.4269</td><td>0.3575</td><td>0.4477</td><td>0.6336</td><td>0.4477</td><td>0.4816</td></tr><tr><td>Gemma-2b-random-addexpl</td><td>0.4929</td><td>0.4928</td><td>0.4929</td><td>0.4927</td><td>0.5914</td><td>0.5777</td><td>0.5914</td><td>0.5820</td><td>0.6221</td><td>0.6164</td><td>0.6221</td><td>0.5587</td></tr><tr><td>Gemma-2b-Vreg</td><td>0.6235</td><td>0.6298</td><td>0.6235</td><td>0.6213</td><td>0.4361</td><td>0.5953</td><td>0.4361</td><td>0.3708</td><td>0.5302</td><td>0.7326</td><td>0.5302</td><td>0.5814</td></tr><tr><td>Gemma-2b-Vreg-addexpl</td><td>0.5847</td><td>0.6190</td><td>0.5847</td><td>0.5525</td><td>0.5875</td><td>0.5846</td><td>0.5875</td><td>0.5859</td><td>0.6767</td><td>0.6932</td><td>0.6767</td><td>0.5990</td></tr><tr><td>Llama3.2-1b-zs</td><td>0.4796</td><td>0.4841</td><td>0.4796</td><td>0.4801</td><td>0.5549</td><td>0.4480</td><td>0.5549</td><td>0.4712</td><td>0.5826</td><td>0.5997</td><td>0.5826</td><td>0.5385</td></tr><tr><td>Llama3.2-1b-random</td><td>0.5398</td><td>0.5483</td><td>0.5398</td><td>0.5222</td><td>0.3949</td><td>0.4831</td><td>0.3949</td><td>0.3417</td><td>0.7116</td><td>0.5064</td><td>0.7116</td><td>0.5917</td></tr><tr><td>Llama3.2-1b-random-addexpl</td><td>0.4867</td><td>0.4868</td><td>0.4867</td><td>0.4782</td><td>0.4118</td><td>0.4821</td><td>0.4118</td><td>0.3996</td><td>0.7116</td><td>0.5064</td><td>0.7116</td><td>0.5917</td></tr><tr><td>Llama3.2-1b-Vreg</td><td>0.6173</td><td>0.6360</td><td>0.6173</td><td>0.6065</td><td>0.6254</td><td>0.6432</td><td>0.6254</td><td>0.6307</td><td>0.7233</td><td>0.7640</td><td>0.7233</td><td>0.6242</td></tr><tr><td>Llama3.2-1b-Vreg-addexpl</td><td>0.6429</td><td>0.6460</td><td>0.6429</td><td>0.6438</td><td>0.6473</td><td>0.6831</td><td>0.6473</td><td>0.6535</td><td>0.7372</td><td>0.7718</td><td>0.7372</td><td>0.6545</td></tr><tr><td>ChatGPT-zs</td><td>0.7265</td><td>0.7420</td><td>0.7265</td><td>0.7221</td><td>0.5236</td><td>0.6551</td><td>0.5236</td><td>0.5032</td><td>0.7860</td><td>0.7920</td><td>0.7860</td><td>0.7551</td></tr><tr><td>ChatGPT-random</td><td>0.6990</td><td>0.7475</td><td>0.6990</td><td>0.6835</td><td>0.6173</td><td>0.6539</td><td>0.6173</td><td>0.6234</td><td>0.7616</td><td>0.7782</td><td>0.7616</td><td>0.7079</td></tr><tr><td>ChatGPT-random-addexpl</td><td>0.6959</td><td>0.7193</td><td>0.6959</td><td>0.6876</td><td>0.6092</td><td>0.6584</td><td>0.6092</td><td>0.6144</td><td>0.7651</td><td>0.7824</td><td>0.7651</td><td>0.7174</td></tr><tr><td>ChatGPT-Vreg</td><td>0.6745</td><td>0.7366</td><td>0.6745</td><td>0.6516</td><td>0.6370</td><td>0.6681</td><td>0.6370</td><td>0.6429</td><td>0.8151</td><td>0.8249</td><td>0.8151</td><td>0.7925</td></tr><tr><td>ChatGPT-Vreg-addexpl</td><td>0.7163</td><td>0.7628</td><td>0.7163</td><td>0.7032</td><td>0.6318</td><td>0.6762</td><td>0.6318</td><td>0.6372</td><td>0.8012</td><td>0.8068</td><td>0.8012</td><td>0.7772</td></tr><tr><td>GPT4o-zs</td><td>0.8816</td><td>0.8856</td><td>0.8816</td><td>0.8813</td><td>0.6170</td><td>0.6398</td><td>0.6170</td><td>0.6228</td><td>0.7837</td><td>0.8150</td><td>0.7837</td><td>0.7396</td></tr><tr><td>GPT4o-random</td><td>0.8776</td><td>0.8850</td><td>0.8776</td><td>0.8770</td><td>0.6739</td><td>0.6830</td><td>0.6739</td><td>0.6771</td><td>0.8291</td><td>0.8526</td><td>0.8291</td><td>0.8090</td></tr><tr><td>GPT4o-random-addexpl</td><td>0.8724</td><td>0.8824</td><td>0.8724</td><td>0.8716</td><td>0.6559</td><td>0.6693</td><td>0.6559</td><td>0.6601</td><td>0.8337</td><td>0.8527</td><td>0.8337</td><td>0.8158</td></tr><tr><td>GPT4o-Vreg</td><td>0.8888</td><td>0.8934</td><td>0.8888</td><td>0.8884</td><td>0.7004</td><td>0.6983</td><td>0.7004</td><td>0.6992</td><td>0.8477</td><td>0.8627</td><td>0.8477</td><td>0.8326</td></tr><tr><td>GPT4o-Vreg-addexpl</td><td>0.8847</td><td>0.8912</td><td>0.8847</td><td>0.8842</td><td>0.7155</td><td>0.7170</td><td>0.7155</td><td>0.7162</td><td>0.8419</td><td>0.8605</td><td>0.8419</td><td>0.8242</td></tr><tr><td>Vicuna-7b-zs</td><td>0.5490</td><td>0.5545</td><td>0.5490</td><td>0.5384</td><td>0.4378</td><td>0.6502</td><td>0.4378</td><td>0.3542</td><td>0.2942</td><td>0.7054</td><td>0.2942</td><td>0.1592</td></tr><tr><td>Vicuna-7b-random</td><td>0.5837</td><td>0.5872</td><td>0.5837</td><td>0.5806</td><td>0.4073</td><td>0.6116</td><td>0.4073</td><td>0.3017</td><td>0.7070</td><td>0.6037</td><td>0.7070</td><td>0.5928</td></tr><tr><td>Vicuna-7b-random-addexpl</td><td>0.5622</td><td>0.6040</td><td>0.5622</td><td>0.5206</td><td>0.5334</td><td>0.5849</td><td>0.5334</td><td>0.5423</td><td>0.7023</td><td>0.5063</td><td>0.7023</td><td>0.5884</td></tr><tr><td>Vicuna-7b-Vreg</td><td>0.6000</td><td>0.6069</td><td>0.6000</td><td>0.6023</td><td>0.4512</td><td>0.6549</td><td>0.4512</td><td>0.3821</td><td>0.7837</td><td>0.7999</td><td>0.7837</td><td>0.7471</td></tr><tr><td>Vicuna-7b-Vreg-addexpl</td><td>0.6316</td><td>0.6680</td><td>0.6316</td><td>0.6248</td><td>0.6065</td><td>0.6145</td><td>0.6065</td><td>0.6105</td><td>0.7756</td><td>0.7956</td><td>0.7756</td><td>0.7501</td></tr></table>
+
+Table 3: Overall results on three datasets. "zs" denotes the zero-shot method. "random" denotes randomly sample four examples without using affective information. "random-addexpl" denotes adding explicit Vreg information for the random sample examples. "Vreg" denotes retrieving four examples based on Vreg information using Template 1. "Vreg-addexpl" denotes adding explicit Vreg information using Template 2.
+
+# 3.3.1 Comparison with baselines
+
+(1) Comparison with PLMs and other domain generalization methods: We can observe that most LLMs with RAEmoLLM framework outperform fine-tuned RoBERTa, BERT, and DGMs on AMTCele and COCO datasets, but they slightly underperform fine-tuned models and some DGMs in the PHEME dataset. One possible reason is that in cross-domain misinformation detection tasks, the fine-tuning method may perform better for simple short-text discrimination problems in the largescale dataset (e.g. PHEME). However, they may not be suitable for long texts (e.g. AMTCele) or complex tasks (e.g. intent recognition in COCO), especially in small datasets. We can see that the current DGMs do not have stable performance on different datasets, although they have complex structures. And their results are lower than the
+
+best performance of LLMs with the RAEmoLLM framework in most cases.
+
+(2) Comparison with zero-shot method (LLMs-zs), random few-shot methods (LLMs-random, LLMs-random-addexpl): From Table 3, we can observe that the RAEmoLLM framework largely increases the LLMs with zero-shot method in most cases and performs better than the random few-shot methods (For random few-shot, the largest increase in AMTCele is Gemma-2b (+15.64%), in PHEME is Llama3.2-1b (+31.18%), and in COCO is Vicuna-7b (+15.73%)). The results of LLMs-random-addexpl show that simply applying explicit information has little effect in most cases<sup>7</sup>. A special case is that in the AMTCele dataset, GPT-4o and ChatGPT perform well in zero
+
+shot settings, with ChatGPT even surpassing other few-shot methods. One possible reason is that the AMTCele dataset is collected from fact-checking websites, and ChatGPT's and GPT-4o's training set includes the data and can effectively utilize this information. One example is shown in Table 10.
+
+Table 7 and Table 8 in Appendix C present the performance of Mistral-7b on each domain on AMTCele and PHEME separately. It can be observed that Mistral-7b with RAEmoLLM framework overtakes Mistral with zero-shot and few-shot methods in most domains except for the prince domain in PHEME, which has significantly imbalanced data. Additionally, we also conduct some special cases analysis in Appendix D.
+
+# 3.3.2 Ablation analysis of each module
+
+(1) Index Construction Module (retrieval based on different information): From Table 3, we can observe retrieval based on affective information (LLMs-Vreg, LLMs-Vreg-addexpl) overtake non-retrieval methods (i.e. random few-shot methods (LLMs-random, LLMs-random-addexpl)). From Table 4, we can observe that the RAEmoLLM framework achieves the best results compared to other types of embeddings, which indicates the effectiveness of Vreg embedding.
+
+<table><tr><td></td><td>AMT</td><td>PHEME</td><td>COCO</td></tr><tr><td>Mistral-7b-Vreg</td><td>0.7404</td><td>0.6788</td><td>0.7898</td></tr><tr><td>Mistral-7b-Vreg-addexpl</td><td>0.7717</td><td>0.6920</td><td>0.7931</td></tr><tr><td>Mistral-7b-semantic</td><td>0.6904</td><td>0.6718</td><td>0.7771</td></tr><tr><td>Mistral-7b-sentibert</td><td>0.6984</td><td>0.6663</td><td>0.7687</td></tr></table>
+
+(2) Retrieval Module (different numbers of retrieval examples): Table 5 presents the F1 score of retrieval of different numbers of examples based on Vreg (we only tested 16 examples in the AMTCele dataset due to its long text). From the table, it can be observed that increasing the retrieval examples does not consistently improve the model's performance, and it may even lead to a decline in its performance (e.g. Vreg-addexpl in COCO). One possible reason is that when the model has multiple examples as references, it needs to consider a large amount of information comprehensively, which depends on the model's capability. Another reason we can infer from Table 14. For the three datasets, the p-values in retrieval top 4 examples are all zero. However, as the number of retrieval examples in
+
+Table 4: F1 score of retrieval using different kinds of embeddings. "semantic" denotes retrieval based on RoBERTa.  
+
+<table><tr><td>Datasets</td><td>methods</td><td>4</td><td>8</td><td>16</td><td>32</td><td>64</td></tr><tr><td rowspan="3">AMTCele</td><td>Random</td><td>0.6889</td><td>0.7006</td><td>0.6287</td><td>-</td><td>-</td></tr><tr><td>Vreg</td><td>0.7404</td><td>0.7395</td><td>0.7271</td><td>-</td><td>-</td></tr><tr><td>Vreg-addexpl</td><td>0.7717</td><td>0.7611</td><td>0.7710</td><td>-</td><td>-</td></tr><tr><td rowspan="3">PHEME</td><td>Random</td><td>0.6227</td><td>0.6253</td><td>0.6268</td><td>0.6400</td><td>0.6353</td></tr><tr><td>Vreg</td><td>0.6788</td><td>0.6856</td><td>0.6830</td><td>0.6910</td><td>0.7031</td></tr><tr><td>Vreg-addexpl</td><td>0.6920</td><td>0.6949</td><td>0.6979</td><td>0.6979</td><td>0.6990</td></tr><tr><td rowspan="3">COCO</td><td>Random</td><td>0.7287</td><td>0.7534</td><td>0.7442</td><td>0.7628</td><td>0.7541</td></tr><tr><td>Vreg</td><td>0.7898</td><td>0.7842</td><td>0.7854</td><td>0.8172</td><td>0.7993</td></tr><tr><td>Vreg-addexpl</td><td>0.7931</td><td>0.7208</td><td>0.7499</td><td>0.7600</td><td>0.7475</td></tr></table>
+
+Table 5: F1 score of Mistral-7b with retrieval of different numbers of examples based on Vreg.
+
+creases, the second p-values in AMTCele and the first p-value in COCO dataset also gradually increase. This indicates that the retrieved content may come from another category or unrelated examples, thereby affecting the model's judgment ability. Therefore, when employing retrieval augmentation techniques, it is not just about blindly increasing the number of examples, but rather selectively choosing the most useful examples.
+
+(3) Inference Module (different templates and different base LLMs): We can see LLMs with explicit affective information based on Template 2 (i.e. LLMs-Vreg-expl) exceed LLM without explicit affective information based on Template 1 (i.e. LLMs-Vreg) in most cases. For LLMs-zs and LLMs-random, different base models show significant performance differences. GPT-4o performs the best, followed by ChatGPT and Mistral-7b, while the Gemma-2b model has the lowest score. After using RAEmoLLM framework, the difference between different modules becomes narrowing (e.g. Mistral-7b has achieved or even surpassed the performance of ChatGPT.)
+
+Based on the analysis above, we can conclude that retrieval based on implicit affective information and adding explicit affective information through Template 2 is the most effective way to enhance the LLMs' performance in using Vreg affective cases. The number of retrieval examples seems to have little impact. The LLMs focus on the most relevant examples.
+
+Table 3 shows that Mistral-7b has the best performance among open-sourced LLMs. We choose Mistral-7b to conduct the following experiments.
+
+# 3.3.3 Results on the data retrieved based on different affective information
+
+Figure 2 presents the results of retrieval with different affective embeddings. For retrieval using affective regression information (i.e. Vreg, EReg), it is evident that adding explicit affective information (affect-addexpl) method can improve the per
+
+![](images/c69bb4756b3055acd0f2d57b58f125d781e932619dcaa123967dc20778e90d16.jpg)  
+Figure 2: Results of Mistral-7b based on different affective information on three datasets. "affect" denotes retrieving four examples based on one affective information using Template 1. "affect-addexpl" denotes adding explicit affective information using Template 2.
+
+![](images/b32620ee7c56c86c5c0fac721f27bd6a870b2b55b91646be40f9baedc038d21c.jpg)
+
+![](images/022f790d62deb0fb9fc99302283ec3fe1777be9c8eb24ca7150a867c61a880ca.jpg)
+
+formance compared to solely relying on implicit affective information retrieval (affect). However, when using affective classification information (e.g. ELoc in AMTCele and PHEME), adding explicit affective information may confuse the model. In the COCO dataset, all the affect-addexpl method outperforms affect except for EReg-fear. Regarding the affect-addexpl method, in AMTcele, we can see the results retrieval based on Vreg are best, followed by EReg-sadness and EReg-joy. And the final three rankings are retrieved based on ELocanger, fear, and sadness. It seems that affective intensity and strength are more suitable for cross-domain fake news detection tasks. In PHEME, retrieval based on Ec exhibits the highest performance, with the Vreg and EReg series closely trailing behind. While the last few are the ELoc series, which may suggest that a coarse-grained emotional intensity classification is not suitable for rumour detection. However, it is the opposite in the conspiracy theory dataset. In COCO, the performance of retrieval based on the ELoc series is better than that based on the EReg series. This could be attributed to the dataset's focus on COVID-19 topic, which may elicit more consistent emotional expressions among individuals.
+
+# 4 Conclusion and Future Work
+
+In this paper, we propose RAEmoLLM, the first RAG framework to address cross-domain misinformation detection using in-context learning based on affective information. We introduce the three modules of RAEmoLLM. We also conduct a comprehensive affective analysis for three public misinformation datasets. We evaluate the performance of RAEmoLLM on the three misinformation bench
+
+marks based on various LLMs. The results show that RAEmoLLM can significantly improve LLMs compared to the zero-shot method and other few-shot methods, which illustrates the effectiveness of RAEmoLLM. We also conduct an ablation analysis of each module and analyze the performance of retrieval based on different affective information, which provide a foundation for further improvements in the future.
+
+In the future, we will explore the application of multimodal affective information in the task of detecting misinformation. We will also evaluate the application of the RAEmoLLM framework in other fields (e.g. mental health and finance). In addition to affective information, there are many other influencing factors in misinformation, such as stance and topic. We will combine sentiments and emotions with other features to construct a more robust retrieval database. Furthermore, the retrieval process can be slowed down by a large amount of data. In the future, we will also explore more efficient retrieval methods.
+
+# 5 Limitations
+
+Due to restricted computational resources, we only carried out inference of 1B, 2B, 7B, 8B, 13B, and 33B open-sourced LLMs. As such, we have not considered how the use of larger or different model architectures may potentially impact upon performance in cross-domain misinformation detection tasks.
+
+Though achieving outstanding performance, RAEmoLLM still bears limitations. Firstly, for domain data with imbalanced distribution, RAEmoLLM performs worse compared to zero-shot methods (e.g. prince domain in PHEME). The
+
+special cases analysis in Appendix D also illustrates that in the imbalanced datasets, the retrieval in RAEmoLLM will be influenced for some special cases. Therefore, further exploration is needed to address such issues. Secondly, in the PHEME dataset, RAEmoLLM performs worse than finetuning methods without emotional information. This indicates that for simple tasks with shorter texts, the model still struggles to effectively balance textual features and emotional information.
+
+# Acknowledgments
+
+This work is supported by the Turing Scheme Fund, the Manchester-Melbourne-Toronto Research Fund, the Centre for Digital Trust and Society at the University of Manchester, and the New Energy and Industrial Technology Development Organization. This work is also supported by the scholar award from the Department of Computer Science at the University of Manchester.
+
+# References
+
+Tsun-Hin Cheung and Kin-Man Lam. 2023. Factllama: Optimizing instruction-following language models with external knowledge for automated fact-checking. In 2023 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), pages 846-853. IEEE.  
+Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E Gonzalez, et al. 2023. Vicuna: An open-source chatbot impressing gpt-4 with  $90\%$  chatgpt quality. See https://vicuna.lmsys.org (accessed 14 April 2023), 2(3):6.  
+Arjun Choudhry, Inder Khatri, Arkajyoti Chakraborty, Dinesh Vishwakarma, and Mukesh Prasad. 2022. Emotion-guided cross-domain fake news detection using adversarial domain adaptation. In Proceedings of the 19th International Conference on Natural Language Processing (ICON), pages 75-79.  
+Carmela Comito, Francesco Sergio Pisani, Erica Coppolillo, Angelica Liguori, Massimo Guarascio, and Giuseppe Manco. 2023. Towards self-supervised cross-domain fake news detection.  
+Nicholas Della Giustina. 2023. Misinformation and its effects on individuals and society from 2015-2023: A mixed methods review study.  
+Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805.
+
+Diwen Dong, Fuqiang Lin, Guowei Li, and Bo Liu. 2022a. Sentiment-aware fake news detection on social media with hypergraph attention networks. In 2022 IEEE International Conference on Systems, Man, and Cybernetics (SMC), pages 2174-2180. IEEE.  
+Qingxiu Dong, Lei Li, Damai Dai, Ce Zheng, Zhiyong Wu, Baobao Chang, Xu Sun, Jingjing Xu, and Zhifang Sui. 2022b. A survey on in-context learning. arXiv preprint arXiv:2301.00234.  
+Nida Manzoor Hakak, Mohsin Mohd, Mahira Kirmani, and Mudasir Mohd. 2017. Emotion analysis: A survey. In 2017 international conference on computer, communications and electronics (COMPTELIX), pages 397-402. IEEE.  
+Beizhe Hu, Qiang Sheng, Juan Cao, Yuhui Shi, Yang Li, Danding Wang, and Peng Qi. 2024. Bad actor, good advisor: Exploring the role of large language models in fake news detection. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 22105-22113.  
+Albert Q Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, et al. 2023. Mistral 7b. arXiv preprint arXiv:2310.06825.  
+Suzanne Keen. 2006. A theory of narrative empathy. Narrative, 14(3):207-236.  
+Elena Kochkina, Maria Liakata, and Arkaitz Zubiaga. 2018. All-in-one: Multi-task learning for rumour verification. In Proceedings of the 27th International Conference on Computational Linguistics, pages 3402-3413.  
+Johannes Langguth, Daniel Thilo Schroeder, Petra Filkukova, Stefan Brenner, Jesper Phillips, and Konstantin Pogorelov. 2023. Coco: an annotated twitter dataset of Covid-19 conspiracy theories. Journal of Computational Social Science, pages 1-42.  
+Guanghua Li, Wensheng Lu, Wei Zhang, Defu Lian, Kezhong Lu, Rui Mao, Kai Shu, and Hao Liao. 2024. Re-search for the truth: Multi-round retrieval-augmented large language models are strong fake news detectors. arXiv preprint arXiv:2403.09747.  
+Miaoran Li, Baolin Peng, and Zhu Zhang. 2023. Self-checker: Plug-and-play modules for fact-checking with large language models. arXiv preprint arXiv:2305.14623.  
+Yinheng Li. 2023. A practical survey on zero-shot prompt design for in-context learning. arXiv preprint arXiv:2309.13205.  
+Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692.
+
+Yinpeng Liu, Jiawei Liu, Xiang Shi, Qikai Cheng, and Wei Lu. 2024a. Let's learn step by step: Enhancing in-context learning ability with curriculum learning. arXiv preprint arXiv:2402.10738.  
+Zhiwei Liu, Boyang Liu, Paul Thompson, Kailai Yang, and Sophia Ananiadou. 2024b. Conspemollm: Conspiracy theory detection using an emotion-based large language model. In ECAI 2024, pages 4649-4656. IOS Press.  
+Zhiwei Liu, Kailai Yang, Tianlin Zhang, Qianqian Xie, Zeping Yu, and Sophia Ananiadou. 2024c. Emollms: A series of emotional large language models and annotation tools for comprehensive affective analysis. arXiv preprint arXiv:2401.08508.  
+Zhiwei Liu, Tianlin Zhang, Kailai Yang, Paul Thompson, Zeping Yu, and Sophia Ananiadou. 2024d. Emotion detection for misinformation: A review. Information Fusion, page 102300.  
+Quanyu Long, Wenya Wang, and Sinno Jialin Pan. 2023. Adapt in contexts: Retrieval-augmented domain adaptation via in-context learning. arXiv preprint arXiv:2311.11551.  
+Qiong Nan, Juan Cao, Yongchun Zhu, Yanyan Wang, and Jintao Li. 2021. Mdfend: Multi-domain fake news detection. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management, pages 3343-3347.  
+Verónica Pérez-Rosas, Bennett Kleinberg, Alexandra Lefevre, and Rada Mihalcea. 2018. Automatic detection of fake news. In Proceedings of the 27th International Conference on Computational Linguistics, pages 3391-3401.  
+Zhen Qin, Yicheng Cheng, Zhe Zhao, Zhe Chen, Donald Metzler, and Jingzheng Qin. 2020. Multitask mixture of sequential experts for user activity streams. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 3083-3091.  
+Tanik Saikh, Arkadipta De, Asif Ekbal, and Pushpak Bhattacharyya. 2020. A deep learning approach for automatic detection of fake news. arXiv preprint arXiv:2005.04938.  
+Dietram A Scheufele and Nicole M Krause. 2019. Science audiences, misinformation, and fake news. Proceedings of the National Academy of Sciences, 116(16):7662-7669.  
+Jiao Shi, Xin Zhao, Nan Zhang, Yu Lei, and Lingtong Min. 2023. Rough-fuzzy graph learning domain adaptation for fake news detection. IEEE Transactions on Computational Social Systems.  
+Amila Silva, Ling Luo, Shanika Karunasekera, and Christopher Leckie. 2021. Embracing domain differences in fake news: Cross-domain fake news detection using multi-modal data. In Proceedings of the AAAI conference on artificial intelligence, volume 35, pages 557-565.
+
+Qi Su, Mingyu Wan, Xiaoqian Liu, and Chu-Ren Huang. 2020. Motivations, methods and metrics of misinformation detection: an nlp perspective. Natural Language Processing Research, 1(1-2):1-13.  
+Wei Tang, Zuyao Ma, Haifeng Sun, and Jingyu Wang. 2023. Learning sparse alignments via optimal transport for cross-domain fake news detection. In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1-5. IEEE.  
+Gemma Team, Thomas Mesnard, Cassidy Hardin, Robert Dadashi, Surya Bhupatiraju, Shreya Pathak, Laurent Sifre, Morgane Riviere, Mihir Sanjay Kale, Juliette Love, et al. 2024. Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295.  
+Yu Tong, Weihai Lu, Zhe Zhao, Song Lai, and Tong Shi. 2024. Mmdfnd: Multi-modal multi-domain fake news detection. In ACM Multimedia 2024.  
+Liang Wang, Nan Yang, and Furu Wei. 2023. Learning to retrieve in-context examples for large language models. arXiv preprint arXiv:2307.07164.  
+Xinyi Wang, Wanrong Zhu, Michael Saxon, Mark Steyvers, and William Yang Wang. 2024. Large language models are latent variable models: Explaining and finding good demonstrations for in-context learning. Advances in Neural Information Processing Systems, 36.  
+Yi Wu, Ziqiang Li, Chaoyue Wang, Heliang Zheng, Shanshan Zhao, Bin Li, and Dacheng Tao. 2024. Domain re-modulation for few-shot generative domain adaptation. Advances in Neural Information Processing Systems, 36.  
+Shangqing Xu and Chao Zhang. 2024. Misconfidence-based demonstration selection for llm in-context learning. arXiv preprint arXiv:2401.06301.  
+Xin Xu, Yue Liu, Panupong Pasupat, Mehran Kazemi, et al. 2024. In-context learning with retrieved demonstrations for language models: A survey. arXiv preprint arXiv:2401.11624.  
+Keyang Xuan, Li Yi, Fan Yang, Ruochen Wu, Yi R Fung, and Heng Ji. 2024. Lemma: Towards lvm-enhanced multimodal misinformation detection with external knowledge augmentation. arXiv preprint arXiv:2402.11943.  
+Da Yin, Tao Meng, and Kai-Wei Chang. 2020. Sentibert: A transferable transformer-based architecture for compositional sentiment semantics. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 3695–3706.  
+Zhenrui Yue, Huimin Zeng, Ziyi Kou, Lanyu Shang, and Dong Wang. 2022. Contrastive domain adaptation for early misinformation detection: A case study on COVID-19. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management, pages 2423-2433.
+
+Zhenrui Yue, Huimin Zeng, Yimeng Lu, Lanyu Shang, Yang Zhang, and Dong Wang. 2024. Evidence-driven retrieval augmented response generation for online misinformation. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 5628-5643.
+
+Zhenrui Yue, Huimin Zeng, Yang Zhang, Lanyu Shang, and Dong Wang. 2023. Metaadapt: Domain adaptive few-shot misinformation detection via meta learning. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5223-5239.
+
+Shengyu Zhang, Linfeng Dong, Xiaoya Li, Sen Zhang, Xiaofei Sun, Shuhe Wang, Jiwei Li, Runyi Hu, Tianwei Zhang, Fei Wu, et al. 2023a. Instruction tuning for large language models: A survey. arXiv preprint arXiv:2308.10792.
+
+Xuewen Zhang, Yaxiong Pan, Xiao Gu, and Gang Liang. 2023b. Sentiment analysis-based social network rumor detection model with bi-directional graph convolutional networks. In International Conference on Computer Application and Information Security (ICCAIS 2022), volume 12609, pages 463-469. SPIE.
+
+# A Related Work
+
+# A.1 Minformation detection
+
+Cross-domain misinformation detection: Cross-domain misinformation detection refers to identifying and detecting misleading or false information across different domains or sources. Comito et al. (2023) propose a deep learning-based architecture to mitigate this problem by yielding high-level cross-domain features. Tang et al. (2023) design one framework to learn transferable features across domains by aligning the source and target news using Optimal Transport techniques. Shi et al. (2023) develop a rough-fuzzy graph learning framework that uses representations of cross-domain sample uncertainty structural information, and captures shared general features across domains. Tong et al. (2024) integrate domain embeddings and attention mechanisms for domain-specific knowledge extraction and combine techniques to obtain multi-domain and multi-modal information. Nan et al. (2021) adopt domain gates to aggregate multiple representations extracted by a mixture of experts (MoE) for fake news detection. Silva et al. (2021) jointly leverage domain-specific and cross-domain knowledge and introduces an unsupervised technique to train a multi-domain fake news detection model. Yue et al. (2022) propose a contrastive adaptation network, which leverages pseudo-labeling to
+
+generate target examples and design a label correction component to solve label shift problems. Yue et al. (2023) develop a domain-adaptive few-shot method based on meta-learning, which adopts limited target examples to provide feedback and guide knowledge transfer from the source domain to the target domain. However, these methods require complex structures and fine-tuning strategies.
+
+Retrieval augmented misinformation detection: Retrieval augmented generation (RAG) combines LLMs with retrieval systems to utilize external knowledge, enabling models to generate more accurate content. Xuan et al. (2024) leverage LVLM intuition and reasoning capabilities to enhance the accuracy of multimodal misinformation by retrieving external knowledge. Yue et al. (2024) collect supporting evidence from scientific sources and generate responses for combating misinformation online based on this evidence. Cheung and Lam (2023) adopt external, most up-to-date information available on the Internet to bridge the knowledge gap in an LLM to enhance fake news detection performance. Li et al. (2024) employ a multi-round retrieval strategy, which can extract key evidence from web sources for claim verification. These findings demonstrate the effectiveness of RAG technology in detecting misinformation.
+
+Affect-based misinformation detection: Emotion and sentiment are important features for misinformation detection (Liu et al., 2024d). Zhang et al. (2023b) combine the use of semantic and sentiment information, along with propagation information for rumour detection. Dong et al. (2022a) design a sentiment-aware hyper-graph attention network for fake news detection. Liu et al. (2024b) develop a conspiracy theory detection LLM by finetuning EmoLLaMA (Liu et al., 2024c). Choudhry et al. (2022) utilize emotional information for fake news detection based on an adversarial learning structure. Unfortunately, these works either have complex structural designs or fine-tuned models, which require significant time and computational resources. RAEmoLLM in this paper applies the ICL method based on retrieving demonstration examples through affective information, which has a simple structure and does not involve fine-tuning.
+
+# A.2 In-context learning
+
+In-context learning (ICL) is a specific prompting engineering method, in which the task demonstrations are included in prompts for LLMs learning (Xu et al., 2024). Wang et al. (2023) develop a
+
+framework to provide high-quality context examples for LLMs, which firstly evaluate the quality of candidate examples through a reward model, and then conduct knowledge distillation to train a dense retriever. Wang et al. (2024) introduce an algorithm that utilizes a small LM to select the best demonstrations from a set of annotated data, and subsequently expand these demonstrations to larger LMs. Liu et al. (2024a) develop in-context curriculum learning, a simple but helpful demonstration ordering method for ICL that gradually increases the complexity of prompt demonstrations. Xu and Zhang (2024) propose in-context reflection to strategically select demonstrations that reduce the discrepancy between the LLM's outputs and the actual input-output mappings. Long et al. (2023) propose a retrieval-enhanced language model to address cross-domain problems, in which they train language models by learning both target domain distribution and the discriminative task signal simultaneously with the augmented cross-domain in-context examples. Inspired by these works, we propose the RAEmoLLM.
+
+# B Task Prompt and Instruction Example
+
+For AMTCele, we utilize "Determine whether the target text is 0. Fake or 1. Legit." For PHEME, we employ "Determine if the target text is 0. non-rumours or 1. rumours." For COCO, we apply "Classify the text regarding COVID-19 conspiracy theories or misinformation into one of the following three classes: 0. Unrelated. 1. Related (but not supporting). 2. Conspiracy (related and supporting)." Here we keep the 0. Unrelated category to test the robustness of the LLM by increasing the complexity of the task.
+
+Table 6 presents a specific instruction example.
+
+# C The results from different domains in the AMTCele and PHEME datasets. (Table 7 and 8)
+
+# D Special cases analysis
+
+Misinformation and true information often convey different affective information (as shown in Table 2 and Table 13). For example, fake news and conspiracy theories tend to evoke more negative sentiments and emotions (e.g. anger or fear) and less joy. However, these results are based on statistics derived from the entire dataset. The special cases need to be analyzed. We investigate some special
+
+Task: Determine if the target text is 0. non-rumours or 1. rumours.  
+Target text: UPDATE: Reports of 1 more shooter being SHOT. This is in addition to one shot and killed earlier in Parliament Hill #OttawaShooting. Sentiment intensity: 0.234.  
+Here are a few examples retrieved through sentiment intensity:  
+Text: UPDATE: Reports gunman says four devices are located around Sydney. Security response underway. Police calling for calm. #9News. Sentiment intensity: 0.429. The label of this text: 1. rumours.  
+Text: JUST IN: Police confirm to ABC there is a second hostage situation underway in eastern Paris. Sentiment intensity: 0.328. The label of this text: 1. rumours.  
+Text: UPDATE: There are reports police have discovered the identity of the lone gunman, with the #SydneySiege in its sixth hour. #9News Sentiment intensity: 0.435. The label of this text: 1. rumours.  
+Text: JUST IN: A separate shooting and hostage situation at a supermarket in eastern Paris has been reported ... developing. Sentiment intensity: 0.236. The label of this text: 1. rumours.  
+According to the above information, the label of target text:
+
+Table 6: An example in the PHEME instruction dataset.
+
+cases retrieved based on ELoc. The results are listed in Table 9.
+
+For AMTCele, we investigate cases where fake news lacks anger or exhibits higher levels of joy, as well as cases where legit news displays higher levels of anger or lacks joy. We can see that the examples retrieved are mostly of the same category as the target, and their results have not been greatly influenced. For PHEME and COCO, we calculate statistics on cases of rumour and conspiracy without fear or exhibiting higher levels of joy (we do not report conspiracy with higher joy due to its low occurrence), as well as cases where non-rumour and related display higher levels of fear or without joy. We can see that the results for rumours in PHEME and related in COCO are poor. The most likely reason is due to the imbalance of categories in the original data, and these special cases are in the minority. This has resulted in the retrieval of more data from the larger category in original datasets, causing the model to learn less useful information and ultimately affecting the final results.
+
+# E Data leakage example in AMTCele (Table 10)
+
+# F Comparison of time consumption between RAEmoLLM and fine-tuning methods (Table 11)
+
+We take the PHEME dataset (6425 items) as an example to compare the time consumption between RAEmoLLM (applying ChatGPT as the base
+
+<table><tr><td rowspan="2">Model</td><td colspan="2">biz</td><td colspan="2">edu</td><td colspan="2">entmt</td><td colspan="2">polit</td><td colspan="2">sports</td><td colspan="2">tech</td><td colspan="2">celebrity</td></tr><tr><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td></tr><tr><td>BERT</td><td>0.5975</td><td>0.5930</td><td>0.5725</td><td>0.5436</td><td>0.5800</td><td>0.5610</td><td>0.5450</td><td>0.5180</td><td>0.5525</td><td>0.5293</td><td>0.5650</td><td>0.5409</td><td>0.5152</td><td>0.5039</td></tr><tr><td>Mistral-7b-zs</td><td>0.7250</td><td>0.7135</td><td>0.8000</td><td>0.7954</td><td>0.7625</td><td>0.7595</td><td>0.5750</td><td>0.5157</td><td>0.7750</td><td>0.7714</td><td>0.6000</td><td>0.5442</td><td>0.6980</td><td>0.6925</td></tr><tr><td>Mistral-7b-random</td><td>0.7375</td><td>0.7218</td><td>0.6625</td><td>0.6191</td><td>0.7375</td><td>0.7251</td><td>0.5500</td><td>0.4357</td><td>0.6875</td><td>0.6761</td><td>0.5625</td><td>0.4589</td><td>0.7580</td><td>0.7489</td></tr><tr><td>Mistral-7b-Vreg</td><td>0.7750</td><td>0.7656</td><td>0.8250</td><td>0.8222</td><td>0.8250</td><td>0.8222</td><td>0.6125</td><td>0.5706</td><td>0.8125</td><td>0.8089</td><td>0.7250</td><td>0.7067</td><td>0.7320</td><td>0.7275</td></tr><tr><td>Mistral-7b-Vreg-addexpl</td><td>0.8000</td><td>0.7968</td><td>0.8625</td><td>0.8620</td><td>0.8500</td><td>0.8496</td><td>0.6625</td><td>0.6423</td><td>0.8375</td><td>0.8373</td><td>0.8625</td><td>0.8607</td><td>0.7360</td><td>0.7346</td></tr></table>
+
+Table 7: The results from different domains in the AMTCele dataset  
+
+<table><tr><td rowspan="2">Model</td><td colspan="2">sydneyseige</td><td colspan="2">ottawashooting</td><td colspan="2">charliehebdo</td><td colspan="2">ferguson</td><td colspan="2">germanwings</td><td colspan="2">prince</td><td colspan="2">putinmissing</td><td colspan="2">gurlitt</td><td colspan="2">ebola</td></tr><tr><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td><td>Acc</td><td>F1</td></tr><tr><td>BERT</td><td>0.7463</td><td>0.7418</td><td>0.7497</td><td>0.7490</td><td>0.7971</td><td>0.8113</td><td>0.7053</td><td>0.7147</td><td>0.7275</td><td>0.7260</td><td>0.1296</td><td>0.1985</td><td>0.5866</td><td>0.5297</td><td>0.5391</td><td>0.4949</td><td>0.5714</td><td>0.7220</td></tr><tr><td>Mistral-7b-zs</td><td>0.6536</td><td>0.6552</td><td>0.6506</td><td>0.6504</td><td>0.6075</td><td>0.6407</td><td>0.4051</td><td>0.4146</td><td>0.6716</td><td>0.6638</td><td>0.7382</td><td>0.8344</td><td>0.5546</td><td>0.4807</td><td>0.4420</td><td>0.4389</td><td>0.4286</td><td>0.6000</td></tr><tr><td>Mistral-7b-random</td><td>0.6822</td><td>0.6838</td><td>0.5719</td><td>0.5232</td><td>0.6946</td><td>0.7153</td><td>0.4506</td><td>0.4653</td><td>0.6652</td><td>0.6646</td><td>0.6395</td><td>0.7636</td><td>0.5378</td><td>0.4569</td><td>0.5362</td><td>0.4225</td><td>0.3571</td><td>0.5263</td></tr><tr><td>Mistral-7b-Vreg</td><td>0.7215</td><td>0.7195</td><td>0.6652</td><td>0.6596</td><td>0.7335</td><td>0.7521</td><td>0.5818</td><td>0.6102</td><td>0.7143</td><td>0.7139</td><td>0.5451</td><td>0.6881</td><td>0.6008</td><td>0.5716</td><td>0.4928</td><td>0.4514</td><td>0.5000</td><td>0.6667</td></tr><tr><td>Mistral-7b-Vreg-addexpl</td><td>0.7437</td><td>0.7403</td><td>0.6753</td><td>0.6683</td><td>0.7431</td><td>0.7613</td><td>0.6527</td><td>0.6655</td><td>0.7036</td><td>0.7033</td><td>0.4592</td><td>0.6128</td><td>0.6050</td><td>0.6023</td><td>0.4348</td><td>0.4308</td><td>0.4286</td><td>0.6000</td></tr></table>
+
+Table 8: The results from different domains in the PHEME dataset  
+
+<table><tr><td rowspan="2">Datasets</td><td rowspan="2">Eloc</td><td rowspan="2">num</td><td rowspan="2">F1</td><td colspan="2">mean num of retrieval</td></tr><tr><td>legit/non-rum/related</td><td>fake/rumour/consp</td></tr><tr><td rowspan="4">AMT</td><td>fake anger=0</td><td>218</td><td>0.8152</td><td>1.0780</td><td>2.9220</td></tr><tr><td>legit anger=2/3</td><td>29</td><td>0.9643</td><td>2.2414</td><td>1.7586</td></tr><tr><td>fake joy=2/3</td><td>14</td><td>0.6667</td><td>1.5000</td><td>2.5000</td></tr><tr><td>legit joy=0</td><td>304</td><td>0.8571</td><td>2.1217</td><td>1.8783</td></tr><tr><td rowspan="4">PHEME</td><td>non-rum fear=2/3</td><td>446</td><td>0.6978</td><td>2.4776</td><td>1.5224</td></tr><tr><td>rumour fear=0</td><td>1039</td><td>0.3804</td><td>2.4658</td><td>1.5342</td></tr><tr><td>non-rum joy=0</td><td>3795</td><td>0.8949</td><td>2.9057</td><td>1.0943</td></tr><tr><td>rumour joy=2/3</td><td>25</td><td>0.2759</td><td>3.7600</td><td>0.2400</td></tr><tr><td rowspan="3">COCO</td><td>related fear=2/3</td><td>47</td><td>0.5538</td><td>2.0426</td><td>1.9574</td></tr><tr><td>consp fear=0</td><td>171</td><td>0.9073</td><td>0.9708</td><td>3.0292</td></tr><tr><td>realted joy==0</td><td>246</td><td>0.7607</td><td>2.2927</td><td>1.7073</td></tr></table>
+
+Table 9: Special cases retrieval based on EIoc. "num" denotes number. "non-rum" denotes non-rumours. "const" denotes conspiracy. The "0", "2", and "3" in the EIoc column represent "no", "moderate", and "high" emotional intensity.  
+
+<table><tr><td>Prompt</td><td>News content: Alex Jones Apologizes for Promoting &#x27;Pizzagate&#x27; Hoax Alex Jones a prominent conspiracy theorist and the host of a popular right-wing radio show ... Where does this news come from?</td></tr><tr><td>GPT-3.5-trubo</td><td>This news comes from a New York Times article</td></tr><tr><td>GPT-4o</td><td>The news about Alex Jones apologizing for promoting the Pizzagate conspiracy theory likely comes from a reputable news source such as The New York Times, The Washington Post, or another major media outlet that covers significant events and developments involving public figures and misinformation. The details provided in your question, such as the mention of &quot;The Alex Jones Show&quot; and the Pizzagate conspiracy, align with coverage typically found in mainstream news articles addressing misinformation and its impact.</td></tr></table>
+
+Table 10: Data leakage example in AMTCele
+
+model) and the fine-tuning method (BERT). From Table 11, it can be observed that RAEmoLLM will consume about 122s to construct the retrieval database (Obtain embeddings: 72s, Retrieval examples: 50s) and 208s to obtain the affective labels. For fine-tuning methods, we take BERT as an example. The current time consumed (Training each epoch: 3906s) by BERT was measured based on a single set of hyperparameters (e.g., batch size and learning rate). In practice, fine-tuning methods may require more time and effort to optimize hyperparameters. Overall, the RAEmoLLM process is simpler and more efficient.
+
+# G Affective analysis
+
+# G.1 Five types of affective information
+
+(1) Emotion intensity (EIreg): For each of four different emotions (anger, fear, joy and sadness), assign a score between 0 and 1 to represent the intensity of emotion of the text;  
+(2) Emotion intensity classification (Eloc): The text can be classified into one of four classes of the intensity of emotion (anger, fear, joy, sadness), i.e. no/low/moderate/high emotional intensity;  
+(3) Sentiment (valence) strength (Vreg): Assign a real-valued score between 0 (most negative) and 1 (most positive) to represent the sentiment intensity of the text.  
+(4) Sentiment (valence) classification (Voc): The text can be categorized into one of seven ordinal classes (i.e. {very, moderately, slightly} negative, neutral, {slightly, moderately, very} positive);  
+(5) Emotion detection (Ec): The text can be classified as 'neutral or no emotion' or as one, or more, of eleven given emotions (anger, anticipation, disgust, fear, joy, love, optimism, pessimism, sadness, surprise, trust).
+
+<table><tr><td>RAEmoLLM</td><td>Obtain embeddings 71.68s</td><td>Obtain labels 208s</td><td>Retrieval Examples 50s</td><td>Inference (time/item) 0.48s</td></tr><tr><td>Bert</td><td>Train (time/epoch) 3906.31s</td><td>Inference (time/item) 0.093s</td><td></td><td></td></tr></table>
+
+Table 11: Time consumption of RAEmoLLM (take ChatGPT as base model) and fine-tuning methods (task BERT as the example) based on the PHEME dataset.
+
+# G.2 Further Affective Analysis
+
+We show the statistics values and distribution of labels and embeddings in this Section. In Figures 3 to Figure 8, the y-axis represents the distribution of labels within the intention class indicated on the x-axis. The affective analysis on COCO has been done by ConspEmoLLM (Liu et al., 2024b). The figures show that most fake information convey more negative sentiments/emotions and less positive emotions compared to real categories<sup>8</sup>. Figure 9 and Figure 10 present the 3D visualization of affective embeddings on AMTCele and PHEME respectively. Table 13 shows the statistics values of EReg and Vreg on PHEME and COCO.
+
+To explore the relationship between affective classification information and misinformation, we conduct a chi-squared significance test and create two categorical variables. One is the misinformation label (real and fake), and the other variable is affective information. For Eloc, we count the values for 0 (absence) and others (presence) of a certain emotion. For Voc, we count the values of 7 classes. For Ec, we count the number of instances that contain each of the 11 emotions individually. Assuming the null hypothesis that affective signals are independent of text truthfulness, the chi-squared test results in Table 12 show p-values close to 0, allowing us to reject the null hypothesis. Overall, affective classification signals are also statistically linked to the veracity of the news.
+
+Table 14 shows statistics of different affective embeddings (i.e. last hidden layer of EmoLLaMA-chat-7B). We perform t-tests on the top- $K$  cosine similarity within categories and the cosine similarity between categories. For example, "fake-legit" denotes computing the cosine similarity between each data point in the "fake" category and each data point in the "legit" category. We then selected the top- $K$  similarity values and performed t-test on them. The t-value and p-value of the top-4 similarity values between "fake-legit" and "fake-fake"
+
+are -22.516 and 0, which demonstrates that the top 4 similar data retrieved based on cosine similarity within the "fake" category are highly likely to belong to the same "fake" category. We can see from the results in Table 14 that all affective information leads to the same conclusion in the top-4 scenarios<sup>9</sup>. We also visualize the data distribution reduced to 3 dimensions using PCA in Figures 9 and 10 in Appendix. It can be observed that different categories are clearly separated in the latent space. All the above demonstrate the close relationship between affective information and misinformation.
+
+![](images/ad9a089cda50cbf67b26a15639560ce2c256da48df02e0f19920db79612242c7.jpg)
+
+![](images/9b56b89a3d1f25ab8118273814b00bb3764369d22a9a1f9a4654397600f20e85.jpg)  
+Figure 3: Emotion intensity classification on AMTCele
+
+![](images/6e522ade5cf6f23fbc5668745f6e244c287c9f0d4d46c1f0038f8e2615aa5ef8.jpg)
+
+![](images/1486d1fbc5b33b7dd6ba26ad636176fb347beb4fe09edb63ec2c1c8576c489f1.jpg)
+
+![](images/a16aaac233ffac43812d7a0474b1f6dee63b8d2aa126356e2a5cb78eea4189ee.jpg)  
+Figure 4: Sentiment classification on AMTCele
+
+<table><tr><td rowspan="2"></td><td colspan="3">AMT</td><td colspan="3">PHEME</td><td colspan="3">COCO</td></tr><tr><td>Eloc</td><td>Voc</td><td>Ec</td><td>Eloc</td><td>Voc</td><td>Ec</td><td>Eloc</td><td>Voc</td><td>Ec</td></tr><tr><td>chi-squared statistic</td><td>131.16</td><td>46.07</td><td>69.40</td><td>197.98</td><td>146.14</td><td>499.48</td><td>76.31</td><td>25.09</td><td>61.50</td></tr><tr><td>p-value</td><td>3.60E-25</td><td>2.86E-08</td><td>5.78E-11</td><td>3.08E-39</td><td>5.07E-29</td><td>5.69E-101</td><td>7.76E-14</td><td>3.28E-04</td><td>1.88E-09</td></tr></table>
+
+Table 12: Chi-squared statistics values of EIoc, Voc, Ec on different datasets.  
+
+<table><tr><td rowspan="2">Datasets</td><td rowspan="2">Affective</td><td rowspan="2">sub-emotion</td><td colspan="2">non-rumours/related</td><td colspan="2">rumours/conspiracy</td><td colspan="2">t-test</td></tr><tr><td>mean</td><td>var</td><td>mean</td><td>var</td><td>t</td><td>p</td></tr><tr><td rowspan="5">PHEME</td><td rowspan="4">EReg</td><td>Anger</td><td>0.4547</td><td>0.0102</td><td>0.4233</td><td>0.0075</td><td>12.7093</td><td>1.44E-36</td></tr><tr><td>Fear</td><td>0.5337</td><td>0.0170</td><td>0.5632</td><td>0.0198</td><td>-8.5027</td><td>2.28E-17</td></tr><tr><td>Joy</td><td>0.2134</td><td>0.0121</td><td>0.1817</td><td>0.0133</td><td>11.0177</td><td>5.58E-28</td></tr><tr><td>Sadness</td><td>0.5215</td><td>0.0152</td><td>0.5177</td><td>0.0182</td><td>1.1442</td><td>0.2526</td></tr><tr><td>Vreg</td><td></td><td>0.4331</td><td>0.0143</td><td>0.3842</td><td>0.0139</td><td>15.9786</td><td>2.18E-56</td></tr><tr><td rowspan="5">COCO</td><td rowspan="4">EReg</td><td>Anger</td><td>0.5475</td><td>0.0088</td><td>0.5641</td><td>0.0068</td><td>-4.5211</td><td>6.43E-06</td></tr><tr><td>Fear</td><td>0.5623</td><td>0.0097</td><td>0.6034</td><td>0.0077</td><td>-10.5568</td><td>1.56E-25</td></tr><tr><td>Joy</td><td>0.1800</td><td>0.0111</td><td>0.1514</td><td>0.0075</td><td>7.2230</td><td>6.66E-13</td></tr><tr><td>Sadness</td><td>0.4701</td><td>0.0098</td><td>0.4773</td><td>0.0073</td><td>-1.8808</td><td>0.0601</td></tr><tr><td>Vreg</td><td></td><td>0.3961</td><td>0.0095</td><td>0.3973</td><td>0.0066</td><td>-0.3325</td><td>0.7395</td></tr></table>
+
+Table 13: T-test statistics values of EReg and Vreg on different datasets. The t-test is conducted between non-rumours/related and rumours/conspiracy.
+
+![](images/93fe685e3708766fb3ad8c64b6286c98af913f7f25c734858d8f56e15a31094a.jpg)  
+Figure 5: Emotion classification on AMTCele
+
+![](images/98746818db3126e5a6551d17b224c08feff7813f03c4f8aa4f0b06662b1a6006.jpg)  
+Figure 7: Sentiment classification on PHEME
+
+![](images/b9e74d91bc46b4af39463e07360dbea2b07d14582356307d72794c8175bd4be6.jpg)  
+Figure 6: Emotion intensity classification on PHEME
+
+![](images/20c672fe1e684b1a98a2178b2efcf86384c8a9ba71f1833f62a905496aeaa21a.jpg)  
+Figure 8: Emotion classification on PHEME
+
+<table><tr><td rowspan="2">Datasets</td><td rowspan="2">Values</td><td rowspan="2">top 4</td><td rowspan="2">top 8</td><td rowspan="2">Vreg top 16</td><td rowspan="2">top 32</td><td rowspan="2">top 64</td><td rowspan="2">Voc</td><td rowspan="2">Ec</td><td colspan="4">EIreg</td><td colspan="4">EIoc</td></tr><tr><td>anger</td><td>fear</td><td>joy</td><td>sadness</td><td>anger</td><td>fear</td><td>joy</td><td>sadness</td></tr><tr><td rowspan="8">AMTCele</td><td>fake-legit</td><td>0.791</td><td>0.771</td><td>0.753</td><td>0.736</td><td>0.718</td><td>0.852</td><td>0.812</td><td>0.801</td><td>0.801</td><td>0.801</td><td>0.801</td><td>0.840</td><td>0.840</td><td>0.840</td><td>0.840</td></tr><tr><td>fake-fake</td><td>0.848</td><td>0.810</td><td>0.783</td><td>0.761</td><td>0.741</td><td>0.894</td><td>0.862</td><td>0.855</td><td>0.855</td><td>0.855</td><td>0.855</td><td>0.885</td><td>0.885</td><td>0.885</td><td>0.885</td></tr><tr><td>t</td><td>-22.516</td><td>-14.875</td><td>-10.951</td><td>-8.976</td><td>-8.037</td><td>-20.550</td><td>-22.617</td><td>-22.434</td><td>-22.433</td><td>-22.462</td><td>-22.461</td><td>-22.260</td><td>-22.246</td><td>-22.267</td><td>-22.244</td></tr><tr><td>p</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td></tr><tr><td>legit-fake</td><td>0.787</td><td>0.765</td><td>0.747</td><td>0.729</td><td>0.711</td><td>0.848</td><td>0.807</td><td>0.797</td><td>0.797</td><td>0.797</td><td>0.797</td><td>0.836</td><td>0.836</td><td>0.836</td><td>0.836</td></tr><tr><td>legit-legit</td><td>0.841</td><td>0.798</td><td>0.768</td><td>0.743</td><td>0.721</td><td>0.886</td><td>0.856</td><td>0.848</td><td>0.848</td><td>0.848</td><td>0.848</td><td>0.877</td><td>0.877</td><td>0.877</td><td>0.877</td></tr><tr><td>t</td><td>-21.568</td><td>-12.845</td><td>-8.052</td><td>-5.263</td><td>-3.452</td><td>-17.138</td><td>-21.024</td><td>-21.399</td><td>-21.387</td><td>-21.407</td><td>-21.396</td><td>-19.364</td><td>-19.328</td><td>-19.335</td><td>-19.315</td></tr><tr><td>p</td><td>0.000</td><td>0.000</td><td>0.001</td><td>0.008</td><td>0.063</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td></tr><tr><td rowspan="8">PHEME</td><td>nonr-rum</td><td>0.930</td><td>0.927</td><td>0.924</td><td>0.921</td><td>0.917</td><td>0.982</td><td>0.952</td><td>0.940</td><td>0.940</td><td>0.940</td><td>0.939</td><td>0.972</td><td>0.972</td><td>0.972</td><td>0.972</td></tr><tr><td>nonr-nonr</td><td>0.957</td><td>0.946</td><td>0.938</td><td>0.932</td><td>0.927</td><td>0.989</td><td>0.971</td><td>0.963</td><td>0.963</td><td>0.963</td><td>0.963</td><td>0.983</td><td>0.983</td><td>0.983</td><td>0.983</td></tr><tr><td>t</td><td>-75.127</td><td>-49.017</td><td>-35.035</td><td>-27.844</td><td>-24.327</td><td>-69.237</td><td>-78.344</td><td>-77.082</td><td>-77.231</td><td>-76.869</td><td>-78.103</td><td>-71.392</td><td>-71.732</td><td>-71.005</td><td>-72.538</td></tr><tr><td>p</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td></tr><tr><td>rum-nonr</td><td>0.935</td><td>0.932</td><td>0.929</td><td>0.925</td><td>0.921</td><td>0.984</td><td>0.957</td><td>0.945</td><td>0.944</td><td>0.945</td><td>0.944</td><td>0.974</td><td>0.974</td><td>0.974</td><td>0.974</td></tr><tr><td>rum-rum</td><td>0.961</td><td>0.950</td><td>0.942</td><td>0.935</td><td>0.928</td><td>0.990</td><td>0.974</td><td>0.966</td><td>0.966</td><td>0.967</td><td>0.966</td><td>0.984</td><td>0.984</td><td>0.984</td><td>0.984</td></tr><tr><td>t</td><td>-58.813</td><td>-38.823</td><td>-27.206</td><td>-19.693</td><td>-14.156</td><td>-54.654</td><td>-58.600</td><td>-59.494</td><td>-59.637</td><td>-59.377</td><td>-60.266</td><td>-55.874</td><td>-56.306</td><td>-56.033</td><td>-56.759</td></tr><tr><td>p</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td></tr><tr><td rowspan="8">COCO</td><td>rela-consp</td><td>0.873</td><td>0.870</td><td>0.866</td><td>0.861</td><td>0.856</td><td>0.955</td><td>0.905</td><td>0.885</td><td>0.885</td><td>0.886</td><td>0.885</td><td>0.936</td><td>0.936</td><td>0.937</td><td>0.936</td></tr><tr><td>rela-rela</td><td>0.907</td><td>0.887</td><td>0.875</td><td>0.865</td><td>0.857</td><td>0.967</td><td>0.931</td><td>0.916</td><td>0.916</td><td>0.916</td><td>0.916</td><td>0.953</td><td>0.953</td><td>0.954</td><td>0.954</td></tr><tr><td>t</td><td>-44.603</td><td>-23.007</td><td>-11.581</td><td>-5.437</td><td>-2.012</td><td>-37.288</td><td>-43.522</td><td>-44.744</td><td>-44.772</td><td>-44.253</td><td>-44.800</td><td>-38.201</td><td>-38.337</td><td>-37.684</td><td>-38.281</td></tr><tr><td>p</td><td>0.000</td><td>0.093</td><td>0.428</td><td>0.457</td><td>0.312</td><td>0.004</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.001</td><td>0.000</td><td>0.001</td><td>0.001</td><td>0.002</td><td>0.002</td></tr><tr><td>consp-rela</td><td>0.863</td><td>0.858</td><td>0.852</td><td>0.846</td><td>0.838</td><td>0.950</td><td>0.897</td><td>0.876</td><td>0.876</td><td>0.877</td><td>0.876</td><td>0.929</td><td>0.929</td><td>0.930</td><td>0.929</td></tr><tr><td>consp-consp</td><td>0.911</td><td>0.891</td><td>0.878</td><td>0.868</td><td>0.859</td><td>0.968</td><td>0.933</td><td>0.919</td><td>0.919</td><td>0.920</td><td>0.920</td><td>0.954</td><td>0.954</td><td>0.955</td><td>0.954</td></tr><tr><td>t</td><td>-74.176</td><td>-47.239</td><td>-33.132</td><td>-25.606</td><td>-21.079</td><td>-54.114</td><td>-69.563</td><td>-73.828</td><td>-73.876</td><td>-73.190</td><td>-73.709</td><td>-60.255</td><td>-60.393</td><td>-59.577</td><td>-60.204</td></tr><tr><td>p</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td></tr></table>
+
+Table 14: Statistics values of cosine similarity between embeddings of different affective information on three datasets. Top K denotes retrieval top K examples. In addition to Vreg, the results of other affective information are all based on top 4. "A-B" represents the calculation of cosine similarity between each data point in A and each data point in B. Each element (i, j) in the resulting calculation represents the cosine similarity between the i-th vector in the A group embeddings and the j-th vector in the B group embeddings. The top 4 refers to selecting the four highest values from each row. The t-value and p-value represent the t-test results for the "A-B" results of the two lines above.
+
+![](images/89ab6742f7939391b316d5629c49984b83bc0ef0ab1787c18bac8a3217f4f0ce.jpg)  
+Figure 9: 3D visualization of affective embeddings on AMTCele. 0: Fake. 1: Legit
+
+![](images/e47f1b3ed2bb289540490aef1fbb8f3d160e47a91ca334053906b914e4d3d414.jpg)
+
+![](images/9bd4eacbda05fdff6f77acb6d6ecc198287b96002cdfc8a4db1d3a009ff4818b.jpg)
+
+![](images/23e78646e30aae0a3c6712dbf9c5549eb3aed57986d58b741a6790ddc7cd22dc.jpg)  
+Figure 10: 3D visualization of affective embeddings on PHEME. 0: Non-rumours. 1: Rumours
+
+![](images/d097138bb2c2679a3fe3acbcb8e258e01dc58cc5bb30508b0710717f6737a201.jpg)
+
+![](images/5e1b8cf2397ab41a0178778734731498117fc5370a0db380a0b6bf984de059ef.jpg)

raemollmretrievalaugmentedllmsforcrossdomainmisinformationdetectionusingincontextlearningbasedonemotionalinformation/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c994ada71ec43c6a4697b8fddea02506a1b682fb4d90e61c2f904efdf2bfbc3a
+size 1352208

raemollmretrievalaugmentedllmsforcrossdomainmisinformationdetectionusingincontextlearningbasedonemotionalinformation/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2868980b16902f0b7839da6bd5d9735c7c525953f591fb8002fd4b70187b252
+size 476020

ragcriticleveragingautomatedcriticguidedagenticworkflowforretrievalaugmentedgeneration/107e3e3f-a4f8-46d2-9879-d6c5196285ad_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82d51695e7d9bf9c2f30502a8ab613eae924efa6f4f79521e079a3e5843dfa5a
+size 179134

ragcriticleveragingautomatedcriticguidedagenticworkflowforretrievalaugmentedgeneration/107e3e3f-a4f8-46d2-9879-d6c5196285ad_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12df3572863417832c524e90245fdb0b584e1749489c1d3b0664d82e1b4c5f7a
+size 220239

ragcriticleveragingautomatedcriticguidedagenticworkflowforretrievalaugmentedgeneration/107e3e3f-a4f8-46d2-9879-d6c5196285ad_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1d2194ba481a8cc010826be31bd082c5fe3b05ee905df0c99cb4c3802ac76f8
+size 2751218

ragcriticleveragingautomatedcriticguidedagenticworkflowforretrievalaugmentedgeneration/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ec1bdc06cee9bea6f032c8b7d5538798fd0b1ecd56dc8352dfb9d9992cbdb34
+size 1454903

ragcriticleveragingautomatedcriticguidedagenticworkflowforretrievalaugmentedgeneration/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ace6dfde8e7d3ff7c277a83f24ad42816b46004f807ea7961c7d44b9476a5c9e
+size 801941

ragevalscenariospecificragevaluationdatasetgenerationframework/2befc6df-c50d-4008-957b-58aa4a8d715f_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:481c36c47565ce55eca0afa5acfd5670f0c13d42a3f0c32913c9aaf64606f252
+size 155816

ragevalscenariospecificragevaluationdatasetgenerationframework/2befc6df-c50d-4008-957b-58aa4a8d715f_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92d444a24c1598ac4081e59847ab1ce898cc20817c4b625b641fac9571a67dcc
+size 185509

ragevalscenariospecificragevaluationdatasetgenerationframework/2befc6df-c50d-4008-957b-58aa4a8d715f_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56a989d105b2f640e30b69a504a64e6837f8ca6664172e7c52bbce96f04f7680
+size 602602

ragevalscenariospecificragevaluationdatasetgenerationframework/full.md

@@ -0,0 +1,668 @@
+# RAGEval: Scenario Specific RAG Evaluation Dataset Generation Framework
+
+Kunlun Zhu $^{17*}$ , Yifan Luo $^{2*}$ , Dingling Xu $^{3*}$ , Yukun Yan $^{1\dagger}$ , Zhenghao Liu $^{4}$ , Shi Yu $^{1}$ , Ruobing Wang $^{5}$ , Shuo Wang $^{1}$ , Yishan Li $^{6}$ , Nan Zhang $^{6}$ , Xu Han $^{1}$ , Zhiyuan Liu $^{1\dagger}$ , Maosong Sun $^{1}$
+
+$^{1}$ Department of Computer Science and Technology, Institute for AI, Tsinghua University,  $^{2}$ Institute for Interdisciplinary Information Sciences, Tsinghua University,
+
+<sup>3</sup>Beijing Normal University, <sup>4</sup>Northeastern University, University of Chinese Academy of Sciences, <sup>6</sup>ModelBest, <sup>7</sup>University of Illinois Urbana-Champaign
+
+# Abstract
+
+Retrieval-Augmented Generation (RAG) is a powerful approach that enables large language models (LLMs) to incorporate external knowledge. However, evaluating the effectiveness of RAG systems in specialized scenarios remains challenging due to the high costs of data construction and the lack of suitable evaluation metrics. This paper introduces RAGEval, a framework designed to assess RAG systems across diverse scenarios by generating high-quality documents, questions, answers, and references through a schema-based pipeline. With a focus on factual accuracy, we propose three novel metrics—Completeness, Hallucination, and Irrelevance—to evaluate LLM-generated responses rigorously. Experimental results show that RAGEval outperforms zero-shot and one-shot methods in terms of clarity, safety, conformity, and richness of generated samples. Furthermore, the use of LLMs for scoring the proposed metrics demonstrates a high level of consistency with human evaluations. RAGEval establishes a new paradigm for evaluating RAG systems in real-world applications. The code and dataset are released at https://github.com/OpenBMB/RAGEval.
+
+# 1 Introduction
+
+Retrieval-Augmented Generation (RAG) systems are increasingly gaining attention (Gao et al., 2023; Asai et al., 2024) due to their ability to integrate external knowledge into large language models (LLMs). This ability is crucial in fields such as medicine, finance, and law, where factual accuracy is crucial in decision-making. However, RAG systems are still prone to hallucination, mainly due to noise introduced during retrieval and LLMs' limited capacity to exploit retrieved information fully.
+
+![](images/dde5b2560dd22ac3e4fdf3f55c15d95fbd0a3d5e3e1d869297683f8d67f35f48.jpg)  
+Figure 1: The challenges of building scenario-specific RAG evaluation datasets: scenario coverage and annotation costs.
+
+Although various benchmarks for measuring the capabilities of existing RAG systems have been proposed (Joshi et al., 2017; Nguyen et al., 2017; Kwiatkowski et al., 2019; Chen et al., 2024b; Lyu et al., 2024), they often lack sufficient coverage of diverse, domain-specific scenarios and fail to incorporate comprehensive metrics for assessing factual accuracy, which limits their applicability in real-world contexts that require precise and reliable information (Bruckhaus, 2024). Furthermore, the challenges of building scenario-specific evaluation datasets—such as dynamic real-world conditions, privacy concerns, and the need for expert annotation—further exacerbate the issue.
+
+To address these challenges, we propose RAGEval, a novel framework designed to automatically generate scenario-specific RAG evaluation datasets. By summarizing essential knowledge from seed documents, RAGEval creates a schema that forms the basis for generating questions, answers, and references for evaluation. Additionally, factual key points are extracted from each answer, enabling a more accurate assessment of the RAG system predictions.
+
+In RAG system assessments, evaluation met
+
+rics, like data, also play a pivotal role. Traditional metrics such as F1, ROUGE-L (Lin, 2004), and BLEU (Papineni et al., 2002) are often inadequate for complex or long-form responses. They mainly focus on the lexical overlap of the responses with the gold reference and overlook the semantic similarity. Some novel approaches, such as those relying on LLMs to evaluate responses directly (Es et al., 2024; Saad-Falcon et al., 2023), suffer from issues of stability and comparability. To address these limitations, we introduce three novel metrics—Completeness, Hallucination, and Irrelevance—that are grounded in factual key points and provide a more stable and comparable scoring method.
+
+Our main contributions are: (1) We propose RAGEval, a novel framework for automatically generating scenario-specific RAG evaluation datasets. (2) We introduce three novel evaluation metrics to assess the factual accuracy of generated answers more effectively than existing metrics like ROUGE-L and BLEU. (3) We develop a new RAG benchmark, DragonBall, and conduct comprehensive experiments to analyze the impact of RAG systems' retrieval and generation components on the performance results.
+
+# 2 Related Work
+
+The evaluation of question-answering (QA) and RAG systems has seen significant advancements in recent years. Traditional open-domain QA benchmarks, such as HotpotQA (Yang et al., 2018), TriviaQA (Joshi et al., 2017), MS Marco (Nguyen et al., 2017), and Natural Questions (Kwiatkowski et al., 2019), have long served as foundational datasets for general QA tasks. However, these benchmarks face limitations in evaluating modern RAG systems, particularly their ability to assess domain-specific knowledge, nuanced outputs, and retrieval accuracy. For instance, potential data leakage in these datasets and a lack of fine-grained metrics hinder their effectiveness in evaluating the nuanced behaviour of RAG systems.
+
+In response, several RAG-specific benchmarks have emerged. RGB (Chen et al., 2024b) focuses on assessing LLMs' ability to integrate retrieved information, emphasizing noise robustness. CRUDRAG (Lyu et al., 2024) categorizes RAG tasks into Create, Read, Update, and Delete operations to evaluate different aspects of information retrieval. CRAG (Yang et al., 2024) extends domain cov
+
+erage by introducing mock APIs to simulate real-world retrieval tasks, while MultiHop-RAG (Tang and Yang, 2024) challenges systems with multi-hop reasoning across multiple documents. These benchmarks, while valuable, remain constrained by predefined domains and fixed task structures, limiting their adaptability to dynamic, real-world applications.
+
+Traditional evaluation metrics, such as F1, ROUGE-L, and BLEU, have been widely used in various benchmarks to assess the quality of generated answers in RAG systems. However, these metrics focusing on lexical often fail to capture the full complexity of generative tasks, especially in the case of long-form responses where factual accuracy and contextual relevance are critical. Moreover, metrics like Hit Rate, MRR, and NDCG are commonly used for retrieval evaluation but cannot assess generative capabilities (Liu, 2023; Nguyen, 2023).
+
+In recent years, newer approaches have integrated LLMs into the evaluation process, trying to solve the problems of traditional metrics. RAGAS (Es et al., 2024) and ARES (Saad-Falcon et al., 2023) use LLM-generated data to evaluate contextual relevance and informativeness without relying on ground truth references. While these methods provide valuable insights, they fail to address the complexities of scenario-specific evaluations. RGB (Chen et al., 2024b) introduces task-oriented metrics that assess noise robustness and information integration, but it does not offer the flexibility required for dynamic, application-specific tasks. RAGTruth (Niu et al., 2024) proposes a corpus for evaluating hallucinations, a critical issue in RAG systems.
+
+While the aforementioned benchmarks and evaluation methods have made significant strides, they still face challenges in addressing the diversity of real-world application scenarios, which often require domain-specific data generation and context-sensitive evaluation. To solve the problem of scenario diversity in RAG evaluation, our method builds upon these advancements by introducing a novel framework for automatically generating evaluation datasets. Unlike existing frameworks that rely on predefined datasets and fixed benchmarks, our method offers higher contextual agility, enabling the design of scenario-specific factual queries tailored to different applications.
+
+Furthermore, we introduce three novel keypoint-based evaluation metrics—Completeness, Halluci-
+
+nation, and Irrelevance—designed to assess factual accuracy and relevance in these dynamically generated, scenario-specific contexts. These metrics stand in contrast to traditional benchmarks that assess RAG systems using a single, static set of evaluation criteria. Our framework enables the automatic generation of diverse datasets and provides more adaptable evaluation metrics, making it better suited for evolving application domains.
+
+# 3 Method
+
+In this section, we introduce the proposed RAGEval method. To provide an overview, we summarize the overall generation process as follows:
+
+$$
+\mathcal {S} \rightarrow \mathcal {C} \rightarrow \mathcal {D} \rightarrow (\mathcal {Q}, \mathcal {A}) \rightarrow \mathcal {R} \rightarrow \text {K e y p o i n t s}
+$$
+
+This sequence outlines how the schema summary  $(\mathcal{S})$  leads to configuration generation  $(\mathcal{C})$ , followed by document generation  $(\mathcal{D})$ . From there, question-answer pairs  $(\mathcal{Q},\mathcal{A})$  are derived, and supporting references  $(\mathcal{R})$  are identified. Finally, keypoints are extracted, serving as concise representations of the critical information in the answers.
+
+# 3.1 Schema Summary
+
+In scenario-specific text generation, a schema  $(S)$  is an abstract representation of key elements, encapsulating the aspects of essential factual knowledge from input documents. This schema serves as the backbone that ensures content diversity and reliability while standardizing outputs across various scenarios to maintain alignment with professional standards.
+
+The schema defines a structural framework of key elements for domain-specific documents without containing actual data. In medicine, it may outline categories for symptoms and treatments; in finance, it could establish classifications for sectors, organizations, and metrics. Specific data is later populated into this predefined framework during configuration generation. For example, in legal contexts, the schema might encompass fundamental legal concepts—such as case law, statutes, and court rulings—ensuring broad applicability without relying on specific legal instances. This approach allows the schema to remain versatile and scalable across various legal scenarios. A concrete example of a legal domain schema illustrating these principles is provided in figure 12 in Appendix E.
+
+The schema is initially generated using GPTs<sup>1</sup> based on a curated set of seed documents, which
+
+establish the foundational domain-specific knowledge. Following this, the schema undergoes a series of iterative refinements guided by human intuition and contextual understanding. This process ensures that the schema maintains a balance between comprehensiveness, accuracy, and generalizability, effectively supporting content generation across diverse sub-scenarios. The refinement process $^2$  is designed to prevent over-specialization, thereby enhancing the schema's scalability and adaptability.
+
+# 3.2 Configuration and Document Generation
+
+Generating scenario-specific documents with rich factual information and internal consistency is crucial for creating high-quality datasets, ensuring the generated content can be evaluated accurately and applied effectively in downstream tasks. To achieve this, we first generate configurations  $\mathcal{C}$ , derived from the previously established schema  $S$ . These configurations act as references and constraints for text generation, ensuring consistency across the document.
+
+We adopt a hybrid approach to generate configurations  $\mathcal{C}$ , combining rule-based methods with LLMs to assign values to schema elements. Rule-based methods (e.g., selecting values randomly from predefined scenario-specific options) ensure high accuracy and factual consistency for structured data. Meanwhile, LLMs generate more complex or diverse content, balancing consistency and creativity. For instance, in financial reports, configurations may include various sectors such as agriculture, aviation, and construction, each covering multiple aspects of its respective domain. An illustrative configuration for the legal scenario is provided in figure 12 in Appendix E, demonstrating how different elements can be combined within this domain.
+
+We then use GPT-4o to convert the factual information from the configuration  $\mathcal{C}$  into a structured narrative format tailored to a specific scenario. For example, in medical records, the generated document may include categories such as patient information, medical history, and treatment plan to ensure accuracy and relevance. Similarly, we include a company summary in financial reports to maintain continuity and distinct sections such as Financial Report, Corporate Governance, and Environmental and Social Responsibility.
+
+![](images/74309c57ec956b68b1755852942c0be078b49cb58194c7de5752d87fc28d85e3.jpg)
+
+![](images/1ac2aa84b4ac5cb4ee0556f951396357dffd78d9f6cb5894bd8d9db4163d112f.jpg)
+
+![](images/c5e8c693fec5a9af8477fb9a8c733d625c22e3ef7dec13c322a43bc22b9fa1fe.jpg)
+
+![](images/cb4e1f2ee7cad0689d104734a436da649cca4172552fbf8212d0c1b97cae905c.jpg)  
+Figure 2: RAGEval Progress:  $①$  summarizing a schema containing specific knowledge from seed documents.  $②$  filling in factual information based on this schema to generate diverse configurations.  $③$  generating documents according to the configurations.  $④$  creating evaluation data composed of questions, answers, and references derived from the configurations and documents.
+
+# 3.3 QRA Generation
+
+In this subsection, we describe the process of generating Question-Reference-Answer (QRA) triples using the documents  $\mathcal{D}$  and configurations  $\mathcal{C}$  to establish a robust evaluation framework for information retrieval and reasoning. The goal is to ensure that generated content can be evaluated comprehensively across multiple aspects of information understanding.
+
+We utilize configurations  $\mathcal{C}$  to guide the generation of questions and initial answers, ensuring the generated content is aligned with the schema elements. These configurations are embedded within prompts to ensure that the generated questions are specific and that the answers are precise and grounded in the schema elements. We address different types of questions, such as factual, multi-hop reasoning, summarization, and multi-document questions, each designed to evaluate specific facets of language understanding. To ensure the diversity and controllability of the questions generated by the model, we have designed 7 question types, as detailed in Table 16 in Appendix E. The GPT-4o model is provided with detailed instructions and examples for each question type, generating targeted questions  $\mathcal{Q}$  and initial answers  $\mathcal{A}$ .
+
+Specific prompts and examples are detailed in the Appendix E. Using the generated questions  $\mathcal{Q}$  and initial answers  $\mathcal{A}$ , we extract relevant information fragments (references)  $\mathcal{R}$  from the documents  $\mathcal{D}$ . This is accomplished using an extraction
+
+prompt, ensuring that the generated answers are grounded in the source material for reliability and traceability. Extracting these references enhances the comprehensiveness and consistency of the generated content.
+
+To ensure alignment between answers  $\mathcal{A}$  and references  $\mathcal{R}$ , we iteratively refine the answers to improve coherence and accuracy. If references contain content missing from the answers, we supplement them accordingly. Conversely, if the answers contain unsupported content, we either locate the relevant references or remove the unsupported sections. This step reduces hallucinations and ensures that the final answers are accurate and well-supported by  $\mathcal{R}$ .
+
+Keypoints are generated from answers  $\mathcal{A}$  for each question  $\mathcal{Q}$  to highlight the critical information in the responses. We employ a predefined prompt with in-context learning, including examples across different scenarios and question types. Typically, each response is distilled into 3-5 keypoints, encompassing essential factual details, relevant inferences, and conclusions. This keypoint extraction supports a precise and reliable evaluation of generated content.
+
+# 3.4 DragonBall Dataset
+
+We construct the DragonBall dataset, which stands for Diverse RAG Omni-Benchmark for All scenarios, by leveraging the generation method described above. This dataset encompasses a range
+
+of texts and RAG questions across three critical domains—finance, law, and medical—chosen for their real-world importance. In addition, the dataset features both Chinese and English texts, serving as a comprehensive resource for multilingual, scenario-specific research. Overall, the dataset contains 6,711 questions, reflecting its extensive scale and diversity. Additional details on the generated DragonBall dataset, including human evaluations of data quality, are provided in Appendix C and D.
+
+# 3.5 Evaluation Metrics for RAG Systems
+
+In this work, we propose a comprehensive evaluation framework for RAG systems, considering both retrieval and generation components.
+
+We define multiple metrics to evaluate the model's effectiveness and efficiency in the retrieval phase. These metrics are designed explicitly for RAG systems, considering the situations when generating answers with incomplete and noisy information.
+
+# 3.5.1 Retrieval Metrics
+
+Recall. We introduce the RAG Retrieval Recall metric to evaluate the effectiveness of the retrieval process in matching ground truth references. The Recall is formally defined as
+
+$$
+\operatorname {R e c a l l} = \frac {1}{n} \sum_ {i = 1} ^ {n} \mathbb {1} (M (G _ {i}, \mathcal {R})), \tag {1}
+$$
+
+where  $n$  is the total number of ground truth references,  $G_{i}$  denotes the  $i$ -th ground truth reference,  $\mathcal{R} = \{R_1, R_2, \ldots, R_k\}$  represents the set of retrieved references,  $M(G_{i}, \mathcal{R})$  is a boolean function that returns true if all sentences in  $G_{i}$  are found in at least one reference in  $\mathcal{R}$ , and false otherwise, and  $\mathbb{1}(\cdot)$  is the indicator function, returning 1 if the condition is true and 0 otherwise.
+
+This metric assesses the alignment between retrieved and ground truth references at the sentence level. A ground truth reference is considered successfully recalled if all its constituent sentences are present in at least one retrieved reference.
+
+Effective Information Rate (EIR). This metric quantifies the proportion of relevant information within the retrieved passages, ensuring that the retrieval process is accurate and efficient regarding information content. It is calculated as
+
+$$
+\operatorname {E I R} = \frac {\sum_ {i = 1} ^ {m} \left| G _ {i} \cap R _ {t} \right|}{\sum_ {j = 1} ^ {k} \left| R _ {j} \right|}, \tag {2}
+$$
+
+where  $G_{i}$  is the  $i$ -th ground truth reference,  $R_{t}$  is the set of total retrieved passages,  $m$  is the number of ground truth references successfully matched,  $|G_{i} \cap R_{t}|$  represents the number of words in the intersection of the  $i$ -th ground truth reference and the concatenated retrieved passages  $R_{t}$ , calculated only if  $G_{i}$  is matched in  $R_{t}$ ,  $|R_{j}|$  represents the total number of words in the  $j$ -th retrieved passage, and  $k$  is the total number of retrieved passages.
+
+To calculate  $|G_i \cap R_t|$  at the sentence level, follow these steps: 1) divide  $G_i$  into individual sentences, 2) for each sentence in  $G_i$ , check if it matches any sentence in  $R_t$ , 3) calculate the number of words in the matched sentences, and 4) sum the number of words from all matched sentences to get  $|G_i \cap R_t|$ . These steps ensure the overlap is calculated based on sentence-level matches, providing a more granular and accurate measure of relevant information within the retrieved passages.
+
+# 3.5.2 Generation Metrics
+
+For the generation component, we introduce novel metrics tailored for RAG evaluation. These metrics comprehensively evaluate the quality and reliability of generated answers.
+
+Completeness. Completeness measures how well the generated answer captures the key information from the ground truth. We employ LLM to generate a set of key points  $K = \{k_{1}, k_{2}, \ldots, k_{n}\}$  from the ground truth. The Completeness score is then calculated as the proportion of key points semantically covered by the generated answer  $A$ :
+
+$$
+\operatorname {C o m p} (A, K) = \frac {1}{| K |} \sum_ {i = 1} ^ {n} \mathbb {1} [ A \text {c o v e r s} k _ {i} ], \tag {3}
+$$
+
+where  $\mathbb{1}[\cdot ]$  is an indicator function that evaluates to 1 if the generated answer  $A$  semantically covers the key point  $k_{i}$ , and 0 otherwise. Here, "covers" means that the generated answer contains information consistent with and correctly representing the key point. Specifically, for a key point to be considered covered, the generated answer must include the relevant information and present it accurately without contradictions or factual errors.
+
+Hallucination. Hallucination identifies instances where the content contradicts key points, highlighting potential inaccuracies. The Hallucination score is calculated as
+
+$$
+\operatorname {H a l l u} (A, K) = \frac {1}{| K |} \sum_ {i = 1} ^ {n} \mathbb {1} [ A \text {c o n t r a d i c t s} k _ {i} ], \tag {4}
+$$
+
+where  $\mathbb{1}[\cdot ]$  is an indicator function that evaluates to 1 if the generated answer  $A$  contradicts the key point  $k_{i}$ , and 0 otherwise.
+
+Irrelevancy. Irrelevancy assesses the proportion of key points from the ground truth that are neither covered nor contradicted by the generated answer. Irrelevancy quantifies the proportion of key points neither covered nor contradicted, indicating areas where the answer fails to engage with relevant information. The Irrelevancy score is calculated as
+
+$$
+\operatorname {I r r} (A, K) = 1 - \operatorname {C o m p} (A, K) - \operatorname {H a l l u} (A, K). \tag {5}
+$$
+
+Completeness, Hallucination, and Relevance pinpoint specific RAG models' strengths and weaknesses. They ensure that generated answers are informative, accurate, and relevant, enhancing their quality and trustworthiness. More details about the prompt for evaluation and keypoints generation, and the comparison with human evaluation can refer to the Appendix C.
+
+# 4 Experiments
+
+# 4.1 Setup
+
+In our experiments (Table 1), the BGE-M3 (Chen et al., 2024a) model is used both for Chinese and English, with the following hyperparameters: the TopK retrieved documents are set to 5, the retrieval batch size is 256. The maximum length for the retrieval query is capped at 128 tokens. The default chunk size is set to 512, and meta-information (e.g., company name, patient details) is added to enhance retrieval.
+
+For generation, the maximum input length for the query generator is set to 4096 tokens, and batches of 5 are processed. The generation parameters include a maximum of 512 new tokens per output.
+
+We use the model's default generation configurations (e.g., temperature, Top-P). If not available, the default settings from Hugging Face will be applied. For ChatGPT models, temperature is set to 0.2 and TopP to 1.0, generating one response per query.
+
+We use FlashRAG (Jin et al., 2024) as the RAG inference pipeline with vLLM (Kwon et al., 2023) as the backend.
+
+# 4.2 Generation Performance Comparison
+
+In this experiment, we compare the performance of 9 popular open/close-sourced generation models
+
+with different parameter sizes, including MiniCPM2B-sft and MiniCPM3-4B (Hu et al., 2024), Baichuan-2-7B-chat (Yang et al., 2023), Llama3-8B-Instruct (AI@Meta, 2024), Qwen1.5-7B/14B-chat (Bai et al., 2023), Qwen2-7B-Instruct (Bai et al., 2023), GPT-3.5-Turbo-0125, and GPT-4o-2024-0806  $^{3}$ . We use the same input prompt to compare the outputs of the different generation models.
+
+The overall experimental results of the different generation models are shown in Table 1.
+
+GPT-4o and MiniCPM3-4B Show Superior Generation Performance. According to our proposed keypoint-based evaluation shown in Table 1, GPT-4o achieves the highest Completeness scores of  $79.13\%$  (CN) and  $69.36\%$  (EN) and the lowest Hallucination scores in Chinese at  $12.10\%$ . What's more, the best-performing small-to-medium opensource model is MiniCPM3-4B, which highlights significant room for improvement among opensource alternatives.
+
+Findings on Model Size. Our experimental results in Table 1 further validate the scaling law (Kaplan et al., 2020) within the same model family. For instance, Qwen1.5-14B-chat outperforms Qwen1.5-7B-chat and other open-source models except for MiniCPM3-4B, achieving better scores in both Completeness and Hallucination.
+
+The Effectiveness of Keypoint-Based Metrics. Our analyses reveal notable discrepancies between traditional evaluation metrics, such as Rouge-L and BLEU—and keypoint-based metrics that assess deep semantic alignment. For instance, in the Chinese setting, Baichuan-2-7B-chat achieves the highest Rouge-L (38.30%) and BLEU (21.55%) scores, yet its Completeness score is relatively low at only 60.25%. Conversely, GPT-4o performs the best on Completeness, scoring 79.13% in Chinese, while it exhibits both the low Rouge-L (21.30%) and BLEU (8.70%) in Chinese. These results suggest that while Rouge-L and BLEU primarily measure surface-level language similarity, keypoint-based metrics capture deeper semantic correspondence, thereby offering a more nuanced reflection of a model's true performance in RAG tasks.
+
+# 4.3 Hyperparameter Comparison
+
+In the RAG system, various hyperparameters—such as the chunk size, the number of re
+
+<table><tr><td rowspan="2">Model</td><td colspan="2">Completeness (↑)</td><td colspan="2">Hallucination (↓)</td><td colspan="2">Irrelevance (↓)</td><td colspan="2">Rouge-L (↑)</td><td colspan="2">BLEU (↑)</td></tr><tr><td>CN</td><td>EN</td><td>CN</td><td>EN</td><td>CN</td><td>EN</td><td>CN</td><td>EN</td><td>CN</td><td>EN</td></tr><tr><td>MiniCPM-2B-sft</td><td>54.59</td><td>57.88</td><td>28.82</td><td>19.49</td><td>16.58</td><td>22.63</td><td>31.11</td><td>26.94</td><td>15.19</td><td>6.38</td></tr><tr><td>MiniCPM3-4B</td><td>75.74</td><td>64.09</td><td>13.78</td><td>16.42</td><td>10.48</td><td>19.49</td><td>32.06</td><td>27.99</td><td>16.34</td><td>6.82</td></tr><tr><td>Baichuan-2-7B-chat</td><td>60.25</td><td>57.40</td><td>23.97</td><td>19.60</td><td>15.77</td><td>22.99</td><td>38.30</td><td>30.68</td><td>21.55</td><td>8.84</td></tr><tr><td>Qwen1.5-7B-chat</td><td>69.50</td><td>62.76</td><td>19.25</td><td>17.65</td><td>11.25</td><td>19.60</td><td>32.49</td><td>21.62</td><td>17.11</td><td>4.06</td></tr><tr><td>Qwen2-7B-Instruct</td><td>70.83</td><td>65.38</td><td>16.93</td><td>16.41</td><td>12.24</td><td>18.21</td><td>24.55</td><td>22.99</td><td>10.26</td><td>4.69</td></tr><tr><td>Llama3-8B-Instruct</td><td>69.26</td><td>63.61</td><td>18.29</td><td>15.12</td><td>12.45</td><td>21.27</td><td>21.54</td><td>25.22</td><td>9.15</td><td>5.12</td></tr><tr><td>Qwen1.5-14B-chat</td><td>73.17</td><td>64.41</td><td>14.40</td><td>15.50</td><td>12.43</td><td>20.09</td><td>31.93</td><td>23.99</td><td>15.25</td><td>4.84</td></tr><tr><td>GPT-3.5-Turbo</td><td>75.40</td><td>68.37</td><td>13.10</td><td>15.72</td><td>11.50</td><td>15.91</td><td>18.92</td><td>19.84</td><td>6.45</td><td>3.35</td></tr><tr><td>GPT-4o</td><td>79.13</td><td>69.36</td><td>12.10</td><td>13.79</td><td>8.77</td><td>16.85</td><td>21.30</td><td>23.25</td><td>8.70</td><td>4.80</td></tr></table>
+
+Table 1: Overall model performance results (\%) of nine language models in generation across Chinese (CN) and English (EN) datasets. The evaluation covers both open-source and proprietary models, with open-source models ranging from 2B to 14B parameters.  
+
+<table><tr><td rowspan="3">Model</td><td colspan="4">Retrieval</td><td colspan="6">Generation</td></tr><tr><td colspan="2">Recall (↑)</td><td colspan="2">EIR (↑)</td><td colspan="2">Completeness (↑)</td><td colspan="2">Hallucination (↓)</td><td colspan="2">Irrelevance (↓)</td></tr><tr><td>CN</td><td>EN</td><td>CN</td><td>EN</td><td>CN</td><td>EN</td><td>CN</td><td>EN</td><td>CN</td><td>EN</td></tr><tr><td>BM25</td><td>74.21</td><td>58.08</td><td>4.11</td><td>7.05</td><td>71.89</td><td>63.80</td><td>17.34</td><td>16.61</td><td>10.77</td><td>19.60</td></tr><tr><td>GTE-multilingual-Base</td><td>52.55</td><td>41.61</td><td>2.94</td><td>5.72</td><td>55.17</td><td>54.30</td><td>28.35</td><td>23.01</td><td>16.48</td><td>22.69</td></tr><tr><td>MiniCPM-Embedding</td><td>71.67</td><td>55.29</td><td>4.02</td><td>7.56</td><td>69.89</td><td>63.08</td><td>20.02</td><td>18.79</td><td>10.09</td><td>18.13</td></tr><tr><td>BGE-M3</td><td>72.94</td><td>55.10</td><td>4.03</td><td>6.84</td><td>70.24</td><td>64.08</td><td>18.62</td><td>17.03</td><td>11.14</td><td>18.89</td></tr></table>
+
+Table 2: The performance results (%) of various retrieval models on Chinese (CN) and English (EN) datasets. Metrics include Recall, EIR, Completeness, Hallucination and Irrelevance. We sample 50 queries for each query type in each domain randomly, 2100 queries in total. List of query types can be found at Figure 16
+
+trieved items (Top-K), and the selection of retrieval models—play a pivotal role in determining overall performance. To examine their impact on our dataset, we first explore different retrieval models, including BM25, GTE-multilingual-Base (Zhang et al., 2024), MiniCPM-Embedding $^{4}$ , and BGE-M3 (Chen et al., 2024a). Next, with the chunk size fixed at 512, we investigate how varying the Top-K retrieval value affects the results. Finally, we assess the impact of 3 distinct chunk Top-K selection strategies on completeness under different scenarios. In these experiments, we randomly select 50 samples from all query types, consistent with the data proportions described in Section 4.2. All other parameters remain identical to those in the main experimental setup, and we employ the Llama3-8B-Instruct model for testing. Through these hyperparameter evaluations, we aim to develop a more comprehensive understanding of our DragonBall dataset.
+
+# 4.3.1 Retrieval Model observation
+
+Our experiments shown in Table 2 demonstrate a strong correlation between retrieval metrics (Recall, EIR) and downstream generation quality. For instance, BM25 achieves the highest Recall (74.21% CN) and simultaneously attains the best Completeness (71.89% CN), aligning with expectations. High EIR score also indicates normally low Hallucination, for instance BM25 has the lowest Hallucination and the highest EIR. However, retrieval superiority alone does not guarantee optimal generation performance—BGE-M3 exhibits marginally lower Recall (55.10% EN) yet best Completeness (64.08% EN). We hypothesize that while BM25 effectively retrieves keyword-matching chunks, it may miss contextually nuanced passages requiring deeper reasoning, which are also critical for keypoint coverage in generation tasks.
+
+Notably, the strong performance of BM25 (surpassing dense retrievers like BGE-M3 in generation metrics) can be attributed to two factors: 1) Queries in our benchmark often contain explicit keywords that align with document chunks, and 2) The limited number of relevant references per
+
+<table><tr><td rowspan="3">TopK</td><td colspan="2">Retrieval</td><td colspan="6">Generation</td></tr><tr><td colspan="2">Recall (↑)</td><td colspan="2">Completeness (↑)</td><td colspan="2">Hallucination (↓)</td><td colspan="2">Irrelevance (↓)</td></tr><tr><td>CN</td><td>EN</td><td>CN</td><td>EN</td><td>CN</td><td>EN</td><td>CN</td><td>EN</td></tr><tr><td>2</td><td>49.18</td><td>38.16</td><td>55.04</td><td>51.29</td><td>24.52</td><td>22.29</td><td>20.45</td><td>26.42</td></tr><tr><td>5</td><td>72.94</td><td>55.10</td><td>70.38</td><td>63.96</td><td>18.63</td><td>16.80</td><td>10.99</td><td>19.24</td></tr><tr><td>8</td><td>78.94</td><td>64.05</td><td>72.32</td><td>69.41</td><td>17.10</td><td>13.96</td><td>10.58</td><td>16.63</td></tr></table>
+
+Table 3: TopK Performance Results (%)
+
+query allows simple methods to dominate when retrieving top-5 passages. This contrasts with GTE-multilingual-Base, which underperforms in both retrieval (52.55% CN Recall) and generation (55.17% CN Completeness), likely due to its suboptimal cross-lingual alignment.
+
+# 4.3.2 TopK Retrieval Observations.
+
+As shown in Table 3, increasing TopK from 2 to 8 improves Recall by  $60.51\%$  (CN) and  $67.8\%$  (EN) relative to Recall at  $\mathrm{TopK} = 2$ , confirming that broader retrieval enhances coverage of critical information. Due the no-equivalent of the total tokens retrieved, we don't include EIR metric in Table 3 since it would be useless.
+
+Notably, generation quality exhibits diminishing returns: expanding TopK from 2 to 5 boosts Completeness by  $27.87\%$  (CN) and  $24.70\%$  (EN), whereas further increasing to  $\mathrm{TopK} = 8$  yields only marginal gains ( $2.76\%$  CN,  $8.52\%$  EN). This suggests that while initial retrieval expansion ( $2 \rightarrow 5$ ) addresses core information gaps, subsequent additions ( $5 \rightarrow 8$ ) primarily refine minor details.
+
+Balancing Retrieval Breadth and Noise. While increasing TopK generally improves generation robustness, excessive expansion risks introducing irrelevant passages that may overwhelm the LLM's processing capacity. Although our current results show reduced hallucination with larger TopK, this trend could reverse in scenarios with lower retrieval precision, where noisy inputs mislead the generator. Thus, selecting an optimal TopK—sufficiently large to capture key information yet within the LLM's context window constraints—is critical.
+
+# 4.3.3 Different hyper-parameter trend across three Question Types
+
+As shown in Fig 3, the 3 query types (FQ, MRQ, and NCQ) respond differently to changes in the Chunk-TopK configuration. FQ achieves its highest Completeness at 256-10, MRQ peaks at 512-5,
+
+![](images/dc4366f8d25b3529adef51e79baae053081e298ce59a1c5c80f54d6cf5dcdf49.jpg)  
+Figure 3: Results  $(\%)$  of Completeness of different query types under different Chunk-TopK settings on finance scenario in English dataset. We test three query types: Factual Question (FQ), Multi-hop Reasoning Question (MRQ), Numerical Comparison Question (NCQ).
+
+and NCQ performs best at 128-20, demonstrating that each query type requires a distinct configuration for optimal performance.
+
+These results highlight that no single configuration uniformly benefits all query types. Instead, each query type demands a tailored Chunk-TopK setting. This underscores our core insight: adapting retrieval-augmented generation to the specific characteristics of each query type leads to more robust performance across different scenarios.
+
+# 5 Conclusion
+
+This paper introduces RAGEval, a framework for rapidly generating scenario-specific datasets to evaluate RAG systems. Our approach addresses the limitations of existing benchmarks by prioritizing factual accuracy and scenario-specific knowledge, which are critical across industries. Experimental results show that our metrics offer a more comprehensive and accurate RAG assessment in specific scenarios compared to conventional ones. GPT-4o outperforms overall, but the performance gap with top open-source models is small, showing potential
+
+for improvement. Our experiments also demonstrate that scenario-specific settings are crucial for RAG assessment. Future work could explore extending the framework to diverse scenarios and further close the performance gap in RAG systems.
+
+# Limitations
+
+We highlight two primary limitations of our framework. First, the text generation component heavily relies on large language models, which may produce hallucinations despite our careful prompt design and validation steps. Second, using advanced closed-source models can be costly, although open-source alternatives can help mitigate expenses.
+
+# Acknowledgment
+
+This work is partly supported by the Natural Science Foundation of China under Grant (No. 62206042). This work is also supported by the AI9Stars community.
+
+# References
+
+AI@Meta. 2024. Llama 3 model card.
+
+Akari Asai, Zexuan Zhong, Danqi Chen, Pang Wei Koh, Luke Zettlemoyer, Hannaneh Hajishirzi, and Wen-tau Yih. 2024. Reliable, adaptable, and attributable language models with retrieval. arXiv preprint arXiv:2403.03187.  
+Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenbin Ge, Yu Han, Fei Huang, Binyuan Hui, Luo Ji, Mei Li, Junyang Lin, Runji Lin, Dayiheng Liu, Gao Liu, Chengqiang Lu, Keming Lu, Jianxin Ma, Rui Men, Xingzhang Ren, Xuancheng Ren, Chuanqi Tan, Sinan Tan, Jianhong Tu, Peng Wang, Shijie Wang, Wei Wang, Shengguang Wu, Benfeng Xu, Jin Xu, An Yang, Hao Yang, Jian Yang, Shusheng Yang, Yang Yao, Bowen Yu, Hongyi Yuan, Zheng Yuan, Jianwei Zhang, Xingxuan Zhang, Yichang Zhang, Zhenru Zhang, Chang Zhou, Jingren Zhou, Xiaohuan Zhou, and Tianhang Zhu. 2023. Qwen technical report. arXiv preprint arXiv:2309.16609.  
+Tilmann Bruckhaus. 2024. Rag does not work for enterprises. Preprint, arXiv:2406.04369.  
+Jianlyu Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, and Zheng Liu. 2024a. M3-embedding: Multi-linguality, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. In Findings of the Association for Computational Linguistics ACL 2024, pages 2318–2335, Bangkok, Thailand and virtual meeting. Association for Computational Linguistics.
+
+Jiawei Chen, Hongyu Lin, Xianpei Han, and Le Sun. 2024b. Benchmarking large language models in retrieval-augmented generation. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 17754-17762.  
+Shahul Es, Jithin James, Luis Espinosa Anke, and Steven Schockaert. 2024. RAGAs: Automated evaluation of retrieval augmented generation. In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations, pages 150-158, St. Julians, Malta. Association for Computational Linguistics.  
+Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, and Haofen Wang. 2023. Retrieval-augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997.  
+Shengding Hu, Yuge Tu, Xu Han, Chaoqun He, Ganqu Cui, Xiang Long, Zhi Zheng, Yewei Fang, Yuxiang Huang, Weilin Zhao, et al. 2024. Minicpm: Unveiling the potential of small language models with scalable training strategies. arXiv preprint arXiv:2404.06395.  
+Jiajie Jin, Yutao Zhu, Xinyu Yang, Chenghao Zhang, and Zhicheng Dou. 2024. Flashrag: A modular toolkit for efficient retrieval-augmented generation research. CoRR, abs/2405.13576.  
+Mandar Joshi, Eunsol Choi, Daniel S Weld, and Luke Zettlemoyer. 2017. Triviaqa: A large scale distantly supervised challenge dataset for reading comprehension. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1601-1611.  
+Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. 2020. Scaling laws for neural language models. Preprint, arXiv:2001.08361.  
+Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, et al. 2019. Natural questions: a benchmark for question answering research. Transactions of the Association for Computational Linguistics, 7:453-466.  
+Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. 2023. Efficient memory management for large language model serving with pagedattention. In Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles.  
+Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74-81.
+
+Jerry Liu. 2023. Building production-ready rag applications.  
+Yuanjie Lyu, Zhiyu Li, Simin Niu, Feiyu Xiong, Bo Tang, Wenjin Wang, Hao Wu, Huanyong Liu, Tong Xu, and Enhong Chen. 2024. *Crud-rag: A comprehensive Chinese benchmark for retrieval-augmented generation of large language models.* arXiv preprint arXiv:2401.17043.  
+Isabelle Nguyen. 2023. Evaluating rag part i: How to evaluate document retrieval.  
+Tri Nguyen, Mir Rosenberg, Xia Song, Jianfeng Gao, Saurabh Tiwary, Rangan Majumder, and Li Deng. 2017. MS MARCO: A human-generated MAchine reading COprehension dataset.  
+Cheng Niu, Yuanhao Wu, Juno Zhu, Siliang Xu, KaShun Shum, Randy Zhong, Juntong Song, and Tong Zhang. 2024. RAGTruth: A hallucination corpus for developing trustworthy retrieval-augmented language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 10862-10878, Bangkok, Thailand. Association for Computational Linguistics.  
+Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311-318.  
+Jon Saad-Falcon, Omar Khattab, Christopher Potts, and Matei Zaharia. 2023. Ares: An automated evaluation framework for retrieval-augmented generation systems. arXiv preprint arXiv:2311.09476.  
+Yixuan Tang and Yi Yang. 2024. Multihop-rag: Benchmarking retrieval-augmented generation for multi-hop queries.  
+Aiyuan Yang, Bin Xiao, Bingning Wang, Borong Zhang, Ce Bian, Chao Yin, Chenxu Lv, Da Pan, Dian Wang, Dong Yan, et al. 2023. Baichuan 2: Open large-scale language models. arXiv preprint arXiv:2309.10305.  
+Xiao Yang, Kai Sun, Hao Xin, Yushi Sun, Nikita Bhalla, Xiangsen Chen, Sajal Choudhary, Rongze Daniel Gui, Ziran Will Jiang, Ziyu Jiang, et al. 2024. Crag-comprehensive rag benchmark. arXiv preprint arXiv:2406.04744.  
+Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D Manning. 2018. Hotpotqa: A dataset for diverse, explainable multi-hop question answering. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 2369-2380.  
+Xin Zhang, Yanzhao Zhang, Dingkun Long, Wen Xie, Ziqi Dai, Jialong Tang, Huan Lin, Baosong Yang, Pengjun Xie, Fei Huang, et al. 2024. mgte: Generalized long-context text representation and reranking
+
+models for multilingual text retrieval. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 1393-1412.
+
+# A Schema Refinement
+
+Our refinement constitutes only a small portion of the overall schema and is primarily focused on optimizing its format to better align with the requirements for generating configurations and documents. Specifically, the schema is represented in JSON format. In the original schema, the keys for events were often direct descriptions of the events themselves (e.g., "Major Asset Acquisition"), while the corresponding values were dictionaries composed of fields such as "time", "description", and "impact". However, this structure is not conducive to generating configurations or handling the schema universally in code.
+
+To address this issue, we implemented a manual refinement process, transforming most schema keys into more generic names. For instance, "Major Asset Acquisition" was converted from a specific key into a value in a general dictionary structure comprising the following fields: "event", "time", "description", and "impact". This refinement not only standardizes and unifies the schema structure but also facilitates universal handling and extensibility in subsequent configuration generation and code processing. See figure 9 and 10 for example. Due to the rapid advancement and remarkable progress in model capabilities, current models can now meet such requirements with some constraints in the prompt, potentially eliminating the need for manual refinement in the future.
+
+# B Document Generation
+
+To ensure content consistency in complex document generation, we have developed a hierarchical configuration generation mechanism. The implementation involves three key phases: First, constructing fundamental event schemas that may contain multiple sub-events. The configuration parameters of these base events are then fed back as secondary inputs to drive sub-event generation. For financial reports (particularly multi-sectional filings), we employ a modular approach: generating structured outline configurations for each section first, then producing and integrating content based on these outlines. Additionally, pre-generated standardized company profiles are dynamically embed
+
+5: The response is completely correct and fluent.  
+4: The response is correct but includes redundant information.  
+3: Most of the response is correct.  
+2: About half of the response is correct.  
+1: A small part of the response is correct, or there are logical errors.  
+0: The response is irrelevant or completely incorrect.
+
+Figure 4: QRA quality scoring criteria.
+
+ded into documents to maintain consistent corporate descriptions throughout the report.
+
+# C Quality Assessment
+
+In this section, we introduce the human verification process used to assess the quality of the generated dataset and the evaluation. The assessment is divided into three main tasks: QRA validation, generated documents quality assessment, and automated evaluation validation.
+
+QRA Quality Assessment. We ask 8 annotators to assess the quality of the QRAs by scoring the correctness of the QRAs generated under different configurations according to the standards listed in Figure 4. Those annotators are highly educated students or researchers with enough background knowledge for certain annotated fields and are adequately paid for after the annotations. We randomly select ten samples per question type for every language and scenario, resulting in 420 samples in total for annotation. When scoring, annotators are provided with the document, question, question type, generated response, and references. The results from Table 5 indicate that the QRA quality scores are consistently high across different scenarios, with slight variations between languages. Specifically, the combined proportion of scores 4 and 5 for all scenarios is approximately  $95\%$  or higher. This suggests that our approach maintains a high standard of accuracy and fluency in QRAs.
+
+Document Quality Assessment. We evaluate the quality of the documents generated using RAGEval by comparing them with documents generated using baseline methods, which include zero-shot prompting (to ask the LLM to generate the document given only a scenario prompt) and one-shot prompting (to ask the LLM to generate the document given a scenario prompt and a sample document). We randomly select 20, 20, and 19 generated documents for finance, legal, and medical sce
+
+Safety: Avoidance of real-world sensitive information.
+
+Clarity: Clear and specific information.
+
+Conformity: Remembrance to real documents like financial reports or medical records.
+
+Richness: Depth and breadth of information.
+
+Figure 5: Document quality comparison criteria.
+
+narios for both languages, respectively, and pack each document with 2 baseline documents generated by zero- and one-shot prompting into one group for comparison. Annotators are asked to rank the documents in each group in terms of clarity, safety, richness, and conformity, as defined in Figure 5, with ties allowed. Results shown in Figure 6 demonstrate that our method consistently outperforms zero-shot and one-shot methods across all criteria, particularly in safety, clarity, conformity, and richness. Specifically, for the Chinese and English datasets across the three aspects of richness, clarity, and safety, our method ranks first in over  $85\%$  of the cases. This demonstrates the effectiveness of our approach in generating high-quality articles with diverse and rich content without compromising safety and clarity.
+
+Validation of Automated Evaluation. To validate the consistency between LLM evaluations and human assessments, we compare the LLM-reported metrics—completeness, hallucination, and irrelevance—with those provided by human evaluators. Specifically, we selected the top five results for each question type across various scenarios, covering both Chinese and English, from the Baichuan-2-7B-chat model. This process yielded a total of 210 annotated questions, with each question evaluated by three independent annotators. The annotations were then aggregated using a voting mechanism, classifying each keypoint as either "relevant to the answer," "irrelevant," or "contradictory to the answer." We then calculate the three metrics and compare them with LLM-annotated results. Results in Figure 7 show that the machine and human evaluations show a high degree of alignment in all metrics. The final absolute difference between the human evaluation and the machine evaluation is  $1.67\%$ . The Fleiss' Kappa value between the three annotators is 0.7686. This validates the reliability of our automated evaluation metrics and confirms their consistency with human judgment.
+
+To ensure robust human evaluation, we invited
+
+<table><tr><td>Metric</td><td>Correlation Type</td><td>Correlation Coefficient</td><td>p-value</td></tr><tr><td>Comple</td><td>Spearman</td><td>0.8536</td><td>0.0000</td></tr><tr><td>Comple</td><td>Pearson</td><td>0.8549</td><td>0.0000</td></tr><tr><td>Irrele</td><td>Spearman</td><td>0.7589</td><td>0.0000</td></tr><tr><td>Irrele</td><td>Pearson</td><td>0.7568</td><td>0.0000</td></tr><tr><td>Hallu</td><td>Spearman</td><td>0.7983</td><td>0.0000</td></tr><tr><td>Hallu</td><td>Pearson</td><td>0.7851</td><td>0.0000</td></tr></table>
+
+Table 4: Correlation between Automated Metrics and Human Annotations. We report Spearman and Pearson correlation coefficients for three evaluation dimensions: Complete (Completeness), Irrele (Irrelevancy), and Hallu (Hallucination). All results are statistically significant ( $p < 0.0001$ ). The table is resized to fit the single-column layout.  
+
+<table><tr><td></td><td>Finance</td><td>Law</td><td>Medical</td></tr><tr><td>CN</td><td>4.94</td><td>4.81</td><td>4.76</td></tr><tr><td>EN</td><td>4.84</td><td>4.79</td><td>4.87</td></tr></table>
+
+three NLP researchers with diverse academic and industry backgrounds to annotate the responses independently, as shown in Table 4.
+
+In summary, the human evaluation results highlight the robustness and effectiveness of our method in generating accurate, safe, and rich content across various scenarios, as well as the reliability of our automated evaluation metrics in reflecting human judgment.
+
+# D DragonBall Datasets Details
+
+For document generation, the dataset includes texts from 20 different corporate scenarios in finance, with one randomly selected text per scenario; 10 different legal scenarios, with two randomly selected texts per scenario; and 19 major medical categories, each with two subcategories and one randomly selected text per major category. This ensures a balanced number of human-evaluated documents across finance, law, and medical scenarios.
+
+Table 5: QAR quality human review scores by domain.  
+
+<table><tr><td>scenario</td><td>Language</td><td>Document Count</td></tr><tr><td>Finance</td><td>CN &amp; EN</td><td>40 &amp; 40</td></tr><tr><td>Legal</td><td>CN &amp; EN</td><td>30 &amp; 30</td></tr><tr><td>Medical</td><td>CN &amp; EN</td><td>38 &amp; 38</td></tr></table>
+
+Table 6: Distribution of Documents in the DRAGONBALL Dataset, in total, we have 6711 questions.
+
+![](images/cb96025f1c240821666174212427dc7b813b5165b519bde746e08d50630c44b4.jpg)  
+Figure 6: Document generation comparison by scenario.
+
+![](images/6c8313a4f051d58d95454dbf8bcda1810cf54c634b478c8a5c100ae7a1c68abc.jpg)  
+Figure 7: Automated metric validation results. We show the absolute differences between the two evaluations.
+
+In Table 6, we present a detailed breakdown of the DRAGONBall dataset. The first section of the table shows the distribution of documents across the three scenarios (finance, legal, and medical) in both Chinese (CN) and English (EN), with an equal number of documents for each language. The second section categorizes the types of questions included in the dataset, providing percentages for each type. The third section details the distribution of the number of reference documents used in answering the questions, reflecting the complexity and variability of the dataset. In total, the dataset comprises 6711 questions.
+
+To ensure the high quality of the QRA triples, we first consider the balance and diversity among the different question types, and then we remove homogeneous and meaningless questions. For example, if the number of unanswerable questions is insufficient, we supplement them according to the article. Second, we eliminate redundant references and answer statements and correct logical reasoning errors in the answers to ensure the dataset quality.
+
+![](images/3d24e330dd83d6248ef7af56e8d47573bbdd54fd819ef080c295d2de9dcc156c.jpg)  
+Figure 8: Questions type ratios of DragonBall.
+
+The dataset and the framework will be released under a CC-BY-NC license to ensure its safe use.
+
+To assess the robustness and generalizability of our RAGEval pipeline, we extended human evaluation to three additional domains, with three annotators from diverse backgrounds annotating a total of 15 generated articles for each domain: Environment Report, Product Introduction, and Business Plan. Evaluations were conducted in both Chinese and English using a structured 5-point scoring system:
+
+- Score 5: Surpasses typical human-generated articles  
+- Score 4: High-quality, comparable to well-written human articles  
+- Score 3: Comparable to average human-written articles  
+- Score 2: Below-average compared to human-written articles  
+- Score 1: Significantly lower quality than human-written articles
+
+The evaluation results, averaged across four dimensions (Safety, Clarity, Normativity, Richness), are summarized below:
+
+These results in Table 7,8 demonstrate that our RAGEval pipeline consistently generates content of higher quality than zero-shot and one-shot baselines across multiple domains and languages. By leveraging domain-specific seed articles and structured templates, RAGEval effectively adapts to
+
+varied content types and maintains strong performance, highlighting its generalizability and scalability. The RAGEval codebase is open-sourced to facilitate further evaluation and domain extension.
+
+# E Examples
+
+![](images/7369bb76b0eebadda088003eb75aacc4b4ef672af5e14d6d07c891ae243175e2.jpg)  
+Figure 9: A schema example of Law scenario before refinement.
+
+![](images/673343610bdafa5197a451986b3f55fe4fa8f2e4c7e2f0bdca3dfc6e1fba4330.jpg)  
+Figure 10: A schema example of Law scenario after refinement.
+
+<table><tr><td>Domain</td><td>Generation Type</td><td>Safety</td><td>Clarity</td><td>Normativity</td><td>Richness</td><td>Avg</td></tr><tr><td>Finance</td><td>Zero-shot</td><td>3.33</td><td>2.80</td><td>2.73</td><td>2.47</td><td>2.83</td></tr><tr><td>Finance</td><td>One-shot</td><td>3.60</td><td>2.07</td><td>2.47</td><td>2.00</td><td>2.54</td></tr><tr><td>Finance</td><td>RAGEval (ours)</td><td>3.67</td><td>3.00</td><td>2.93</td><td>3.20</td><td>3.20</td></tr><tr><td>Law</td><td>Zero-shot</td><td>3.33</td><td>2.27</td><td>2.53</td><td>2.00</td><td>2.53</td></tr><tr><td>Law</td><td>One-shot</td><td>1.80</td><td>2.80</td><td>2.80</td><td>2.27</td><td>2.42</td></tr><tr><td>Law</td><td>RAGEval (ours)</td><td>3.67</td><td>3.00</td><td>2.93</td><td>2.93</td><td>3.13</td></tr><tr><td>Medical</td><td>Zero-shot</td><td>2.73</td><td>2.87</td><td>3.00</td><td>2.60</td><td>2.80</td></tr><tr><td>Medical</td><td>One-shot</td><td>2.33</td><td>2.67</td><td>2.87</td><td>2.67</td><td>2.64</td></tr><tr><td>Medical</td><td>RAGEval (ours)</td><td>3.33</td><td>3.00</td><td>3.00</td><td>3.00</td><td>3.08</td></tr><tr><td>Environment</td><td>Zero-shot</td><td>3.53</td><td>1.47</td><td>1.80</td><td>1.07</td><td>1.97</td></tr><tr><td>Environment</td><td>One-shot</td><td>1.00</td><td>2.93</td><td>2.80</td><td>2.47</td><td>2.30</td></tr><tr><td>Environment</td><td>RAGEval (ours)</td><td>3.47</td><td>3.07</td><td>3.00</td><td>3.13</td><td>3.17</td></tr><tr><td>Product</td><td>Zero-shot</td><td>3.13</td><td>2.93</td><td>3.00</td><td>2.67</td><td>2.93</td></tr><tr><td>Product</td><td>One-shot</td><td>3.67</td><td>2.40</td><td>2.27</td><td>1.80</td><td>2.56</td></tr><tr><td>Product</td><td>RAGEval (ours)</td><td>3.33</td><td>3.20</td><td>3.07</td><td>3.13</td><td>3.18</td></tr><tr><td>Business Plan</td><td>Zero-shot</td><td>2.40</td><td>2.40</td><td>2.60</td><td>2.20</td><td>2.40</td></tr><tr><td>Business Plan</td><td>One-shot</td><td>3.00</td><td>2.73</td><td>2.80</td><td>2.33</td><td>2.72</td></tr><tr><td>Business Plan</td><td>RAGEval (ours)</td><td>3.27</td><td>2.93</td><td>2.93</td><td>2.93</td><td>3.02</td></tr></table>
+
+Table 7: Human Evaluation (Chinese). Average scores across four evaluation dimensions for three generation methods in six Chinese domains.  
+
+<table><tr><td>Domain</td><td>Generation Type</td><td>Safety</td><td>Clarity</td><td>Normativity</td><td>Richness</td><td>Avg</td></tr><tr><td>Finance</td><td>Zero-shot</td><td>3.53</td><td>2.60</td><td>2.87</td><td>2.47</td><td>2.62</td></tr><tr><td>Finance</td><td>One-shot</td><td>3.20</td><td>2.33</td><td>2.73</td><td>2.13</td><td>2.60</td></tr><tr><td>Finance</td><td>RAGEval (ours)</td><td>3.67</td><td>3.00</td><td>2.80</td><td>3.07</td><td>3.14</td></tr><tr><td>Law</td><td>Zero-shot</td><td>3.53</td><td>2.07</td><td>2.67</td><td>1.80</td><td>2.52</td></tr><tr><td>Law</td><td>One-shot</td><td>1.73</td><td>2.93</td><td>2.93</td><td>2.13</td><td>2.43</td></tr><tr><td>Law</td><td>RAGEval (ours)</td><td>3.67</td><td>3.00</td><td>3.00</td><td>3.00</td><td>3.17</td></tr><tr><td>Medical</td><td>Zero-shot</td><td>2.73</td><td>2.87</td><td>3.00</td><td>2.53</td><td>2.78</td></tr><tr><td>Medical</td><td>One-shot</td><td>3.07</td><td>2.53</td><td>3.00</td><td>2.47</td><td>2.77</td></tr><tr><td>Medical</td><td>RAGEval (ours)</td><td>3.60</td><td>3.00</td><td>3.00</td><td>3.07</td><td>3.17</td></tr><tr><td>Environment</td><td>Zero-shot</td><td>2.27</td><td>1.53</td><td>1.73</td><td>1.07</td><td>1.65</td></tr><tr><td>Environment</td><td>One-shot</td><td>1.33</td><td>2.93</td><td>3.20</td><td>2.60</td><td>2.52</td></tr><tr><td>Environment</td><td>RAGEval (ours)</td><td>3.07</td><td>3.00</td><td>2.93</td><td>3.40</td><td>3.10</td></tr><tr><td>Product</td><td>Zero-shot</td><td>3.67</td><td>3.00</td><td>3.00</td><td>2.93</td><td>3.15</td></tr><tr><td>Product</td><td>One-shot</td><td>3.67</td><td>2.80</td><td>2.40</td><td>2.40</td><td>2.82</td></tr><tr><td>Product</td><td>RAGEval (ours)</td><td>3.60</td><td>3.00</td><td>3.07</td><td>3.33</td><td>3.25</td></tr><tr><td>Business Plan</td><td>Zero-shot</td><td>3.67</td><td>2.20</td><td>2.40</td><td>2.13</td><td>2.60</td></tr><tr><td>Business Plan</td><td>One-shot</td><td>3.27</td><td>2.73</td><td>2.67</td><td>2.40</td><td>2.77</td></tr><tr><td>Business Plan</td><td>RAGEval (ours)</td><td>3.60</td><td>3.00</td><td>2.87</td><td>3.07</td><td>3.14</td></tr></table>
+
+Table 8: Human Evaluation (English). Average scores across four evaluation dimensions for three generation methods in six English domains.
+
+<table><tr><td rowspan="3">Model</td><td colspan="4">Retrieval</td></tr><tr><td colspan="2">Recall (↑)</td><td colspan="2">EIR (↑)</td></tr><tr><td>CN</td><td>EN</td><td>CN</td><td>EN</td></tr><tr><td>BM25</td><td>80.69</td><td>66.17</td><td>4.46</td><td>8.80</td></tr><tr><td>GTE-multilingual-Base</td><td>65.24</td><td>52.64</td><td>3.69</td><td>7.83</td></tr><tr><td>MiniCPM-Embedding</td><td>81.23</td><td>65.85</td><td>4.62</td><td>9.47</td></tr><tr><td>BGE-M3</td><td>82.54</td><td>64.98</td><td>4.64</td><td>8.60</td></tr></table>
+
+Table 9: The performance results (%) of various retrieval models on DragonBall dataset. The primary metrics evaluated include Recall and EIR. We test all queries in Dragonball dataset.
+
+You are an experienced court clerk, you have rich experience in writing legal documents, and have a strong imagination.
+
+You need to write a detailed legal document sample.  
+You need to write a legal document item based on a JSON template:  
+- The charge to be written is: {crime_name}  
+- In the "defendant" field, the defendant's name, gender, and birth date have been filled out. You only need to fill in their ethnicity, occupation, etc.  
+- In the "defenseLawyer" field, the lawyer's name has been filled out. You only need to fill in the law firm they belong to.  
+- In the "process" field, please provide specific dates and detailed descriptions for each key event.  
+- In the "criminalFacts" field, please create a scenario involving {crime_name}, including specific behavior, time period, place, and amounts involved.  
+- The "Details" field in "criminalFacts" is a list and should have 3 to 4 incidents, each with "timePeriod", "behavior", and "evidence" elements.  
+- The offender's specific actions should be very detailed. You need to carefully supplement the case history and details, possibly involving amounts, times, places, brands, etc., that should look very real.  
+- Evidence materials supporting the verdict need to be very detailed, and you can list multiple pieces of evidence, including written evidence, physical evidence, witness testimony, etc.  
+- The "legalProcedure" field includes the trial date, trial result, and specific sentencing (if probation is applicable, also inform the probation period). {crime_detail}
+
+Figure 11: A config prompt of Law scenario.
+
+![](images/72fc0b6c5dc3f440813dc443891b4e2568b0857f91fea99671ecc7c33c6afd7a.jpg)  
+Figure 12: A config example of Law scenario.
+
+<table><tr><td rowspan="3">Query Type</td><td colspan="6">Retrieval</td><td colspan="3">Generation</td></tr><tr><td colspan="3">Recall (↑)</td><td colspan="3">EIR (↑)</td><td colspan="3">Completeness (↑)</td></tr><tr><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td></tr><tr><td>FQ</td><td>98.00</td><td>100.00</td><td>94.00</td><td>1.02</td><td>2.66</td><td>4.89</td><td>86.00</td><td>92.00</td><td>86.00</td></tr><tr><td>IIQ</td><td>59.26</td><td>68.39</td><td>73.31</td><td>1.43</td><td>5.79</td><td>13.62</td><td>77.77</td><td>82.77</td><td>71.60</td></tr><tr><td>NCQ</td><td>86.33</td><td>77.33</td><td>68.33</td><td>1.50</td><td>5.45</td><td>9.01</td><td>79.33</td><td>68.67</td><td>63.33</td></tr><tr><td>TSQ</td><td>77.00</td><td>78.47</td><td>74.93</td><td>1.94</td><td>6.36</td><td>12.47</td><td>82.00</td><td>79.67</td><td>67.67</td></tr><tr><td>MRQ</td><td>79.95</td><td>84.74</td><td>84.54</td><td>4.56</td><td>7.92</td><td>15.21</td><td>76.00</td><td>78.70</td><td>79.03</td></tr><tr><td>SQ</td><td>55.94</td><td>50.21</td><td>57.89</td><td>3.73</td><td>7.74</td><td>21.17</td><td>58.92</td><td>54.26</td><td>60.12</td></tr><tr><td>UQ</td><td>13.00</td><td>13.00</td><td>16.00</td><td>0.10</td><td>0.39</td><td>1.34</td><td>48.00</td><td>48.67</td><td>54.00</td></tr><tr><td>Avg.</td><td>67.07</td><td>67.45</td><td>67.00</td><td>2.04</td><td>5.18</td><td>11.10</td><td>72.57</td><td>72.10</td><td>68.82</td></tr></table>
+
+Table 10: Chunk-TopK results (\%) of various query types on finance scenario in English. Seven query types are evaluated: Factual Question (FQ), Information Integration Question (IIQ), Numerical Comparison Question (NCQ), Temporal Sequence Question (TSQ), Multi-hop Reasoning Question (MRQ), Summarization Question (SQ), Unanswerable Question (UQ).  
+
+<table><tr><td rowspan="3">Query Type</td><td colspan="6">Retrieval</td><td colspan="3">Generation</td></tr><tr><td colspan="3">Recall (↑)</td><td colspan="3">EIR (↑)</td><td colspan="3">Completeness (↑)</td></tr><tr><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td></tr><tr><td>FQ</td><td>100.00</td><td>100.00</td><td>98.00</td><td>1.71</td><td>2.07</td><td>1.90</td><td>98.00</td><td>88.00</td><td>90.00</td></tr><tr><td>IIQ</td><td>96.00</td><td>91.00</td><td>88.00</td><td>3.26</td><td>3.69</td><td>3.28</td><td>78.40</td><td>78.80</td><td>72.60</td></tr><tr><td>NCQ</td><td>92.00</td><td>86.67</td><td>90.67</td><td>2.84</td><td>3.16</td><td>3.22</td><td>84.00</td><td>82.67</td><td>85.67</td></tr><tr><td>TSQ</td><td>94.00</td><td>89.00</td><td>75.00</td><td>3.10</td><td>3.41</td><td>2.73</td><td>90.67</td><td>79.33</td><td>71.00</td></tr><tr><td>MRQ</td><td>99.00</td><td>98.00</td><td>99.00</td><td>6.73</td><td>8.33</td><td>7.84</td><td>85.79</td><td>80.63</td><td>84.37</td></tr><tr><td>SQ</td><td>94.04</td><td>90.86</td><td>93.25</td><td>10.05</td><td>12.87</td><td>12.98</td><td>71.85</td><td>69.33</td><td>68.85</td></tr><tr><td>UQ</td><td>16.00</td><td>16.00</td><td>12.00</td><td>0.25</td><td>0.30</td><td>0.30</td><td>21.00</td><td>30.00</td><td>22.00</td></tr><tr><td>Avg.</td><td>84.43</td><td>81.65</td><td>79.42</td><td>3.99</td><td>4.83</td><td>4.61</td><td>75.67</td><td>72.68</td><td>70.64</td></tr></table>
+
+Table 11: Chunk-TopK results (\%) of various query types on finance scenario in Chinese. Seven query types are evaluated: Factual Question (FQ), Information Integration Question (IIQ), Numerical Comparison Question (NCQ), Temporal Sequence Question (TSQ), Multi-hop Reasoning Question (MRQ), Summarization Question (SQ), Unanswerable Question (UQ).  
+
+<table><tr><td>You are an experienced court clerk, you have rich experience in writing legal documents, and have a strong imagination.</td></tr><tr><td>You need to write a detailed legal document sample.</td></tr><tr><td>You need to write a criminal judgment based on a JSON example (case key information in JSON form):</td></tr><tr><td>- Based on this completed information, write a legal document that is very detailed, supplementing the case history and details yourself, and writing in the tone of an experienced court clerk. It should be at least 8000 words.</td></tr><tr><td>- You should supplement the case history and details based on the information, such as the defendant&#x27;s confession. You can freely elaborate, but it must comply with legal regulations.</td></tr><tr><td>- When writing the document, you should include the following sections: Court and Prosecutor Information, Defendant and Defense Lawyer Information, Case Procedures, Case Statement, Charge, Evidence Description, Sentencing Considerations, Judgment Result, Appeal Rights Explanation, etc. (do not include legal basis).</td></tr><tr><td>- When writing the document, the tone should use &quot;This court&quot; as the subject. Pay attention to the formal and official tone.</td></tr><tr><td>- Do not make the evidence description and case statement too brief. Describe the case process in detail, and do not use points (1. 2. 3. etc.) in this section.</td></tr><tr><td>- You do not need to write out the referenced legal provisions&#x27; content, only the article numbers. The crime involved is: {crime_name}.</td></tr></table>
+
+Figure 13: A document prompt of Law scenario.
+
+"content":"Ashton, Clarksville Court
+
+Case No. XYZ12345
+
+**IN THE COURT OF ASHTON, CLARKSVILLE**
+
+**CRIMINAL JUDGMENT**
+
+\*\*COURT AND PROSECUTOR INFORMATION\*\*
+
+**Court:**
+
+Ashton, Clarksville Court
+
+**Prosecutorate:**
+
+Ashton, Clarksville Procuratorate
+
+**Chief Judge:**
+
+M. Gray
+
+**Judge:**
+
+H. Torres
+
+\*\*Clerk:\*\*
+
+A. Brown
+
+**DEFENDANT AND DEFENSE LAWYER INFORMATION**
+
+**Defendant:**
+
+Name: J. Gonzalez
+
+Gender: Female
+
+Birthdate: 15th June 1999
+
+Residence: 53 Bayside Street, Clarksville
+
+Ethnicity: Hispanic
+
+Occupation: Senior Inspector, Clarksville Tax Department
+
+**Defense Lawyer:**
+
+Name: M. Smith
+
+Law Firm: Clarksville Legal Associates
+
+\*\*CASE PROCEDURES\*\*
+
+The case against J. Gonzalez commenced with an investigation following a suspicious tip received by the Ashton, Clarksville Procuratorate on 1st March 2023. The investigation revealed substantial evidence implicating the defendant in the Crime of Bending the Law for Personal Gain.
+
+Consequently, J. Gonzalez was taken into detention on 5th March 2023. Criminal detention was applied on 10th March 2023, and the defendant was formally arrested on 12th March 2023.
+
+\*\*CASE STATEMENT\*\*
+
+The Crime of Bending the Law for Personal Gain by the defendant, J. Gonzalez, occurred over a span of one year, from January 2022 to December 2022. During this period, J. Gonzalez exploited her position as a Senior Inspector within the Clarksville Tax Department to manipulate tax audits, reports, and reduce penalty fees for several conspiring local businesses in exchange for substantial financial bribes. This court will detail the pertinent events chronologically to provide a comprehensive understanding of the criminal activities committed.
+
+\*\*Charge:\*\*
+
+Crime of Bending the Law for Personal Gain as per Article 397 of the applicable law.
+
+**EVIDENCE DESCRIPTION**
+
+**1. January 2022 - December 2022: Manipulation of Tax Audits in Exchange for Bribes**
+
+During the year of 2022, J. Gonzalez engaged in illicit activities using her privileged position. Emails confirmed numerous correspondences between J. Gonzalez and various local business owners. These emails explicitly outlined her agreement to manipulate tax audits and financial reports for monetary compensation. Bank statements revealed a series of significant transactions amounting to $125,000 deposited into an account owned by J. Gonzalez from suspicious sources. Testimonies from several business owners corroborated these findings, revealing a consistent pattern of bribery and exploitation.
+
+.
+
+**Date of Judgment:**
+
+15th May 2023
+
+\*\*\*
+
+M. Gray, Chief Judge
+
+\*\* \*\*
+
+H. Torres, Judge
+
+\*\* \*\*
+
+A. Brown, Clerk"
+
+Figure 14: A document example of Law scenario.
+
+```json
+{ "qafact_based": [ { "Question Type": "Factual Question", "Question": "According to the court judgment of Ashton, Clarksville, Court, what was the judgment date?", "ref": [ "Date of Judgment: 15th May 2023"] , "Answer": "15th May 2023." } ], "qa-multi_hop": [ { "Question Type": "Multi-hop Reasoning Question", "Question": "According to the judgment of Ashton, Clarksville, Court, how many instances of bending the law for personal gain did J. Gonzalez commit?", "ref": [ "The Crime of Bending the Law for Personal Gain by the defendant, J. Gonzalez, occurred over a span of one year, from January 2022 to December 2022.", "During this period, J. Gonzalez exploited her position as a Senior Inspector within the Clarksville Tax Department to manipulate tax audits, reports, and reduce penalty fees for several conspiring local businesses in exchange for substantial financial bribes.", "In March 2022, J. Gonzalez revised the tax records for Sunrise Construction Inc., drastically reducing their tax liability after receiving a bribe of \(50,000.", "In exchange for \)30,000, J. Gonzalez facilitated the undue reduction of penalty fees levied on Downtown Boutique Ltd. for late tax submissions.", "The most egregious of the offenses occurred in November 2022, when J. Gonzalez disclosed sensitive and confidential information about ongoing tax investigations to executives at Riven Pharmaceuticals, securing a bribe of \(45,000." ], "Answer": "According to the judgment, J. Gonzalez committed four instances of bending the law for personal gain: manipulating tax audits and reports, altering tax records, reducing penalty fees, and providing confidential information." } ], "qa_summary": [ { "Question Type": "Summary Question", "Summary Content": "Facts of the crime", "Question": "According to the judgment of Ashton, Clarksville, Court, summarize the facts of J. Gonzalez's crimes.", "ref": [ "The Crime of Bending the Law for Personal Gain by the defendant, J. Gonzalez, occurred over a span of one year, from January 2022 to December 2022.", "During this period, J. Gonzalez exploited her position as a Senior Inspector within the Clarksville Tax Department to manipulate tax audits, reports, and reduce penalty fees for several conspiring local businesses in exchange for substantial financial bribes.", "In March 2023, J. Gonzalez revised the tax records for Sunrise Construction Inc., drastically reducing their tax liability after receiving a bribe of \)50,000.", "In exchange for $30,000, J. Gonzalez facilitated the undue reduction of penalty fees levied on Downtown Boutique Ltd. for late tax submissions.", "The most egregious of the offenses occurred in November 2022, when J. Gonzalez disclosed sensitive and confidential information about ongoing tax investigations to executives at Riven Pharmaceuticals, securing a bribe of \(45,000." ], "Answer": "J. Gonzalez, a Senior Inspector at the Clarksville Tax Department, committed the crime of bending the law for personal gain. From January 2022 to December 2022, she manipulated tax audits and reports in exchange for bribes from multiple local businesses. In March 2023, she altered tax records to reduce the tax liability for Sunrise Construction Inc. after receiving $50,000. In August 2023, she reduced penalty fees for late tax submission of Downtown Boutique Ltd. in exchange for $30,000. In November 2023, she provided confidential information about ongoing tax investigations to Riven Pharmaceuticals in exchange for $45,000." } ]
+```
+
+Figure 15: A QRA example of Law scenario.
+
+{
+
+"prompt": "In this task, you will be given a question and a standard answer. Based on the standard answer, you need to summarize the key points necessary to answer the question. List them as follows:
+
+1. .  
+2. ... and so on, as needed.
+
+Example:
+
+Question: What are the significant changes in the newly amended Company Law?
+
+Standard Answer: The 2023 amendment to the Company Law introduced several significant changes. Firstly, the amendment strengthens the regulation of corporate governance, specifically detailing the responsibilities of the board of directors and the supervisory board [1]. Secondly, it introduces mandatory disclosure requirements for Environmental, Social, and Governance (ESG) reports [2]. Additionally, the amendment adjusts the corporate capital system, lowering the minimum registered capital requirements [3]. Finally, the amendment introduces special support measures for small and medium-sized enterprises to promote their development [4].
+
+Key Points:
+
+1. The amendment strengthens the regulation of corporate governance, detailing the responsibilities of the board of directors and the supervisory board.  
+2. It introduces mandatory disclosure requirements for ESG reports.  
+3. It adjusts the corporate capital system, lowering the minimum registered capital requirements.  
+4. It introduces special support measures for small and medium-sized enterprises.
+
+Question: Comparing the major asset acquisitions of Huaxia Entertainment Co., Ltd. in 2017 and Top Shopping Mall in 2018, which company's acquisition amount was larger?
+
+Standard Answer: Huaxia Entertainment Co., Ltd.'s asset acquisition amount in 2017 was larger [1], amounting to 120 million yuan [2], whereas Top Shopping Mall's asset acquisition amount in 2018 was 50 million yuan [3].
+
+Key Points:
+
+1. Huaxia Entertainment Co., Ltd.'s asset acquisition amount in 2017 was larger.
+
+2. Huaxia Entertainment Co., Ltd.'s asset acquisition amount was 120 million yuan in 2017.  
+3. Top Shopping Mall's asset acquisition amount was 50 million yuan in 2018.
+
+Question: Comparing the timing of sustainability and social responsibility initiatives by Meihome Housekeeping Services Co., Ltd. and Cultural Media Co., Ltd., which company initiated these efforts earlier?
+
+Standard Answer: Meihome Housekeeping Services Co., Ltd. initiated its sustainability and social responsibility efforts earlier [1], in December 2018 [2], whereas Cultural Media Co., Ltd. initiated its efforts in December 2019 [3].
+
+Key Points:
+
+1. Meihome Housekeeping Services Co., Ltd. initiated its sustainability and social responsibility efforts earlier.  
+2. Meihome Housekeeping Services Co., Ltd. initiated its efforts in December 2018.  
+3. Cultural Media Co., Ltd. initiated its efforts in December 2019.
+
+Question: Based on the 2017 Environmental and Social Responsibility Report of Green Source Environmental Protection Co., Ltd., how did the company improve community relations through participation in charitable activities, community support and development projects, and public service projects?
+
+Standard Answer: Green Source Environmental Protection Co., Ltd. improved community relations through several social responsibility activities. Firstly, in March 2017, the company participated in or funded charitable activities and institutions to support education, health, and poverty alleviation, enhancing the company's social image and brand recognition [1]. Secondly, in June 2017, the company invested in the local community, supporting education, health, and social development projects, deepening its connection with the community and promoting overall community well-being and development [2]. Finally, in August 2017, the company participated in public service projects such as urban greening and public health improvement projects, enhancing the quality of life in the community and promoting sustainable development [3]. These measures enhanced public perception of the company and improved community relations [4].
+
+Key Points:
+
+1. In March 2017, the company participated in or funded charitable activities and institutions to support education, health, and poverty alleviation, enhancing the company's social image and brand recognition.  
+2. In June 2017, the company invested in the local community, supporting education, health, and social development projects, deepening its connection with the community and promoting overall community well-being and development.  
+3. In August 2017, the company participated in public service projects such as urban greening and public health improvement projects, enhancing the quality of life in the community and promoting sustainable development.  
+4. These measures enhanced public perception of the company and improved community relations.
+
+Test Case:
+
+Question: {question}
+
+Standard Answer: {groundtruth}
+
+Key Points:
+
+Figure 16: Key points generation prompt.
+
+{"prompt":"""In this task, you will receive a question, a generated answer, and multiple key points from a standard answer. Please categorize each key point by determining whether it is Relevant, Irrelevant, or Wrong based on the generated answer. For each key point, provide a brief analysis and conclude with one of the following classifications:
+
+[[Relevant]]] indicates that the generated answer contains key information that is related to and consistent with the key point described in the standard answer.
+
+[[Irrelevant]]] indicates that the generated answer does not contain or involve information related to the key point in the standard answer.
+
+[[Wrong]]] indicates that the generated answer contains information related to the key point but it is incorrect or contradicts the standard answer keypoints.
+
+Ensure that each key point is categorized into only one of the three categories. Provide your analysis and conclusion for each key point sequentially.
+
+Example 1:
+
+Question: What is the current situation and future outlook for wealth management fee income?
+
+Generated Answer: In the first three quarters, the bank achieved wealth management fee income of 5.105 billion yuan, down  $20.7\%$  year-on-year. This decline was mainly due to the active reduction of non-standard product scales and the decline in fund sales caused by fluctuations in the securities market. To address this situation, the bank plans to continue upgrading its products, teams, professionalism, and technological capabilities. Specific measures include improving asset selection and product screening capabilities, enriching and optimizing product shelves, continuously strengthening the construction of professional teams, and focusing on enhancing team professional skills and middle-office empowerment levels to enhance the endogenous development capabilities of the teams. The future outlook includes achieving more stable and higher-quality development through these measures to cope with the challenges of declining wealth management fee income.
+
+Standard Answer Key Points:
+
+Here are 2 key points
+
+1. Current situation of wealth management fee income: Income in the first three quarters decreased by  $21.6\%$  year-on-year.  
+2. Future measures: Plans to upgrade products and teams to achieve stable development.
+
+Key Point Evaluation:
+
+Key Point 1:
+
+Comparable fragment of the generated answer: "wealth management fee income of 5.105 billion yuan, down  $20.7\%$  year-on-year."
+
+Analysis: The generated answer mentions a decrease of  $20.7\%$  year-on-year, whereas the standard key point specifies a decrease of  $21.6\%$ . The percentage difference makes this information incorrect. Conclusion: [[Wrong]]
+
+Key Point 2:
+
+Comparable fragment of the generated answer: "plans to continue upgrading its products, teams, professionalism, and technological capabilities."
+
+Analysis: The generated answer aligns with the standard key point by detailing plans to upgrade products and teams to achieve more stable development.
+
+Conclusion: [[Relevant]]] .... omit three example here
+
+Before you begin the evaluation, please pay attention to the following points:
+
+1. [[Wrong]] should only be assigned when there is a specific factual or logical conflict between the key point and the generated answer. If important content is missing, it should be categorized as [[Irrelevant]]], not [[Wrong]]. More special cases should refer to point 5 below.  
+2. [[Relevant]]] does not require the generated answer to include all the details. It only needs to contain the key information necessary to answer the question. Not all details are required. We ensure that each key point in the standard answer is typically necessary, although some details might not be important for answering the question. When making judgments, focus only on whether the most important information is included and consistent. Also, identical content in different forms can be considered relevant as long as the core key information is present.  
+3. Please ensure that the number of key points evaluated matches the number of key points in the standard answer. Each key point must be evaluated; do not skip or over-evaluate any key point.  
+4. After evaluating the key points, do not repeat your conclusions. Ensure that the total number of classifications - [[Relevant]]], [[Wrong]]], and [[Irrelevant]]] - matches the number of key points in the standard answer. .... omit three more instruction
+
+Test cases:
+
+Question: {question}
+
+Generated Answer: {prediction}
+
+Standard Answer Key Points:
+
+Here are {key_points_num} key points
+
+{key_points}
+
+Key Point Evaluation:"""
+
+}
+
+Figure 17: Key points evaluation prompt.
+
+<table><tr><td rowspan="3">Query Type</td><td colspan="6">Retrieval</td><td colspan="3">Generation</td></tr><tr><td colspan="3">Recall (↑)</td><td colspan="3">EIR (↑)</td><td colspan="3">Completeness (↑)</td></tr><tr><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td></tr><tr><td>FQ</td><td>98.00</td><td>94.00</td><td>88.83</td><td>2.24</td><td>2.52</td><td>2.07</td><td>99.00</td><td>88.00</td><td>91.00</td></tr><tr><td>IIQ</td><td>84.63</td><td>81.92</td><td>79.15</td><td>2.77</td><td>2.97</td><td>2.72</td><td>78.00</td><td>79.00</td><td>79.00</td></tr><tr><td>NCQ</td><td>93.00</td><td>78.33</td><td>68.67</td><td>2.83</td><td>2.58</td><td>2.12</td><td>84.00</td><td>71.67</td><td>65.33</td></tr><tr><td>TSQ</td><td>81.33</td><td>76.33</td><td>67.67</td><td>2.58</td><td>2.89</td><td>2.38</td><td>61.33</td><td>60.67</td><td>56.00</td></tr><tr><td>MRQ</td><td>91.41</td><td>81.93</td><td>82.47</td><td>11.95</td><td>12.10</td><td>11.23</td><td>49.37</td><td>46.41</td><td>49.41</td></tr><tr><td>SQ</td><td>74.63</td><td>73.68</td><td>74.23</td><td>13.98</td><td>14.68</td><td>14.09</td><td>58.89</td><td>64.97</td><td>60.80</td></tr><tr><td>UQ</td><td>10.00</td><td>7.00</td><td>3.00</td><td>0.14</td><td>0.08</td><td>0.08</td><td>44.00</td><td>57.00</td><td>44.00</td></tr><tr><td>Avg.</td><td>76.14</td><td>70.46</td><td>66.29</td><td>5.21</td><td>5.40</td><td>4.96</td><td>67.85</td><td>66.83</td><td>63.67</td></tr></table>
+
+Table 12: Chunk-TopK results (\%) of various query types on law scenario in Chinese. Seven query types are evaluated: Factual Question (FQ), Information Integration Question (IIQ), Numerical Comparison Question (NCQ), Temporal Sequence Question (TSQ), Multi-hop Reasoning Question (MRQ), Summarization Question (SQ), Unanswerable Question (UQ).  
+
+<table><tr><td rowspan="3">Query Type</td><td colspan="6">Retrieval</td><td colspan="3">Generation</td></tr><tr><td colspan="3">Recall (↑)</td><td colspan="3">EIR (↑)</td><td colspan="3">Completeness (↑)</td></tr><tr><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td></tr><tr><td>FQ</td><td>85.00</td><td>87.00</td><td>90.00</td><td>1.58</td><td>2.22</td><td>5.15</td><td>88.00</td><td>88.00</td><td>91.00</td></tr><tr><td>IIQ</td><td>75.07</td><td>72.23</td><td>64.03</td><td>3.33</td><td>4.08</td><td>5.20</td><td>77.50</td><td>77.00</td><td>76.17</td></tr><tr><td>NCQ</td><td>52.00</td><td>64.50</td><td>45.17</td><td>2.34</td><td>4.17</td><td>5.25</td><td>54.83</td><td>54.83</td><td>39.33</td></tr><tr><td>TSQ</td><td>79.33</td><td>59.17</td><td>52.00</td><td>5.44</td><td>4.40</td><td>4.54</td><td>58.67</td><td>45.33</td><td>40.00</td></tr><tr><td>MRQ</td><td>29.42</td><td>30.30</td><td>19.71</td><td>3.19</td><td>7.50</td><td>8.93</td><td>41.74</td><td>28.28</td><td>23.85</td></tr><tr><td>SQ</td><td>18.51</td><td>25.75</td><td>25.90</td><td>4.23</td><td>9.03</td><td>12.86</td><td>33.38</td><td>36.29</td><td>34.89</td></tr><tr><td>UQ</td><td>2.00</td><td>2.00</td><td>2.00</td><td>0.12</td><td>0.12</td><td>0.11</td><td>79.37</td><td>83.50</td><td>85.83</td></tr><tr><td>Avg.</td><td>48.76</td><td>48.71</td><td>42.69</td><td>2.89</td><td>4.50</td><td>6.01</td><td>62.01</td><td>59.10</td><td>55.93</td></tr></table>
+
+Table 13: Chunk-TopK results (\%) of various query types on law scenario in English. Seven query types are evaluated: Factual Question (FQ), Information Integration Question (IIQ), Numerical Comparison Question (NCQ), Temporal Sequence Question (TSQ), Multi-hop Reasoning Question (MRQ), Summarization Question (SQ), Unanswerable Question (UQ).  
+
+<table><tr><td rowspan="3">Query Type</td><td colspan="6">Retrieval</td><td colspan="3">Generation</td></tr><tr><td colspan="3">Recall (↑)</td><td colspan="3">EIR (↑)</td><td colspan="3">Completeness (↑)</td></tr><tr><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td></tr><tr><td>FQ</td><td>85.00</td><td>99.00</td><td>95.00</td><td>1.04</td><td>1.31</td><td>0.97</td><td>80.00</td><td>94.00</td><td>96.00</td></tr><tr><td>IIQ</td><td>86.83</td><td>83.83</td><td>83.50</td><td>1.95</td><td>2.13</td><td>1.61</td><td>87.00</td><td>83.00</td><td>79.00</td></tr><tr><td>NCQ</td><td>83.39</td><td>70.20</td><td>70.61</td><td>2.90</td><td>2.87</td><td>2.39</td><td>84.33</td><td>71.33</td><td>73.00</td></tr><tr><td>TSQ</td><td>94.00</td><td>94.00</td><td>87.67</td><td>2.72</td><td>3.21</td><td>2.35</td><td>87.33</td><td>88.00</td><td>80.33</td></tr><tr><td>MRQ</td><td>83.36</td><td>80.90</td><td>89.66</td><td>4.46</td><td>4.64</td><td>4.44</td><td>67.55</td><td>62.50</td><td>67.73</td></tr><tr><td>SQ</td><td>85.10</td><td>77.96</td><td>83.42</td><td>6.27</td><td>6.32</td><td>5.98</td><td>67.63</td><td>62.52</td><td>63.15</td></tr><tr><td>UQ</td><td>2.00</td><td>0.00</td><td>2.00</td><td>0.03</td><td>0.00</td><td>0.03</td><td>64.67</td><td>60.67</td><td>63.33</td></tr><tr><td>Avg.</td><td>74.24</td><td>72.27</td><td>73.12</td><td>2.77</td><td>2.93</td><td>2.54</td><td>76.93</td><td>74.57</td><td>74.65</td></tr></table>
+
+Table 14: Chunk-TopK results (\%) of various query types on medical scenario in Chinese. Seven query types are evaluated: Factual Question (FQ), Information Integration Question (IIQ), Numerical Comparison Question (NCQ), Temporal Sequence Question (TSQ), Multi-hop Reasoning Question (MRQ), Summarization Question (SQ), Unanswerable Question (UQ).
+
+<table><tr><td rowspan="3">Query Type</td><td colspan="6">Retrieval</td><td colspan="3">Generation</td></tr><tr><td colspan="3">Recall (↑)</td><td colspan="3">EIR (↑)</td><td colspan="3">Completeness (↑)</td></tr><tr><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td><td>128-20</td><td>256-10</td><td>512-5</td></tr><tr><td>FQ</td><td>98.67</td><td>91.67</td><td>80.00</td><td>2.61</td><td>2.24</td><td>1.70</td><td>90.00</td><td>90.00</td><td>88.00</td></tr><tr><td>IIQ</td><td>93.27</td><td>80.60</td><td>81.77</td><td>4.93</td><td>3.27</td><td>4.28</td><td>83.00</td><td>76.00</td><td>88.00</td></tr><tr><td>NCQ</td><td>90.50</td><td>80.50</td><td>44.67</td><td>2.22</td><td>1.49</td><td>0.98</td><td>81.00</td><td>74.33</td><td>40.67</td></tr><tr><td>TSQ</td><td>81.00</td><td>97.00</td><td>92.00</td><td>2.82</td><td>2.53</td><td>2.81</td><td>66.00</td><td>73.33</td><td>72.67</td></tr><tr><td>MRQ</td><td>66.17</td><td>64.16</td><td>40.46</td><td>5.83</td><td>5.23</td><td>2.64</td><td>51.22</td><td>55.38</td><td>43.20</td></tr><tr><td>SQ</td><td>57.42</td><td>64.35</td><td>50.29</td><td>13.74</td><td>14.00</td><td>11.47</td><td>52.96</td><td>59.79</td><td>52.37</td></tr><tr><td>UQ</td><td>0.00</td><td>0.00</td><td>0.00</td><td>0.00</td><td>0.00</td><td>0.00</td><td>96.00</td><td>90.00</td><td>98.00</td></tr><tr><td>Avg.</td><td>69.57</td><td>68.33</td><td>55.60</td><td>4.59</td><td>4.11</td><td>3.41</td><td>74.31</td><td>74.12</td><td>68.99</td></tr></table>
+
+Table 15: Chunk-TopK results (\%) of various query types on medical scenario in English. Seven query types are evaluated: Factual Question (FQ), Information Integration Question (IIQ), Numerical Comparison Question (NCQ), Temporal Sequence Question (TSQ), Multi-hop Reasoning Question (MRQ), Summarization Question (SQ), Unanswerable Question (UQ).  
+
+<table><tr><td>Question Type</td><td>Definition</td></tr><tr><td colspan="2">Single-document QA</td></tr><tr><td>Factual</td><td>Questions targeting specific details within a reference (e.g., a company&#x27;s profit in a report, a verdict in a legal case, or symptoms in a medical record) to test RAG&#x27;s retrieval accuracy.</td></tr><tr><td>Summarization</td><td>Questions that require comprehensive answers, covering all relevant information, to mainly evaluate the recall rate of RAG retrieval.</td></tr><tr><td>Multi-hop Reasoning</td><td>Questions involve logical relationships among events and details within a document, forming a reasoning chain to assess RAG&#x27;s logical reasoning ability.</td></tr><tr><td colspan="2">Multi-document QA</td></tr><tr><td>Information Integration</td><td>Questions that need information from two documents combined, typically containing distinct information fragments, to test cross-document retrieval accuracy.</td></tr><tr><td>Numerical Comparison</td><td>Questions requiring RAG to find and compare data fragments to draw conclusions, focusing on the model&#x27;s summarizing ability.</td></tr><tr><td>Temporal Sequence</td><td>Questions requiring RAG to determine the chronological order of events from information fragments, testing the model&#x27;s temporal reasoning skills.</td></tr><tr><td colspan="2">Unanswerable Questions</td></tr><tr><td>Unanswerable</td><td>Questions arise from potential information loss during the schema-to-article generation, where no corresponding information fragment exists, or the information is insufficient for an answer.</td></tr></table>
+
+Table 16: DragonBall Dataset question types and their definitions

ragevalscenariospecificragevaluationdatasetgenerationframework/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:129b611b04ea77dc917190e29efb53ecadd1671109beeea88f963c27741e5ab5
+size 1719494

ragevalscenariospecificragevaluationdatasetgenerationframework/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11b1f1c3d26d51b78a1635463feb395fc638db0dfa6accfd561d1b4f0dcd16b8
+size 679758

rankcotrefiningknowledgeforretrievalaugmentedgenerationthroughrankingchainofthoughts/8f6125c9-2a9d-4c94-a99a-9e98e45710ac_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b66a7ef69d89d0c27e1c666c122b2dd7ace949f4df3ce5ded001b56c589db2a5
+size 123414

rankcotrefiningknowledgeforretrievalaugmentedgenerationthroughrankingchainofthoughts/8f6125c9-2a9d-4c94-a99a-9e98e45710ac_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3ced862d684ae5ba52c6a3c3d0877fc1916eb1f9efc170338c513a383890374
+size 152421

rankcotrefiningknowledgeforretrievalaugmentedgenerationthroughrankingchainofthoughts/8f6125c9-2a9d-4c94-a99a-9e98e45710ac_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13049ff890910b602dd9fb512b02b2d52fbd33342bae2f51ca3f23d401d09c40
+size 2783930

rankcotrefiningknowledgeforretrievalaugmentedgenerationthroughrankingchainofthoughts/full.md

@@ -0,0 +1,607 @@
+# RankCoT: Refining Knowledge for Retrieval-Augmented Generation through Ranking Chain-of-Thoughts
+
+Mingyan Wu $^{1}$ , Zhenghao Liu $^{1*}$ , Yukun Yan $^{2*}$ , Xinze Li $^{1}$ , Shi Yu $^{2}$ , Zheni Zeng $^{2}$ , Yu Gu $^{1}$ , Ge Yu $^{1}$
+
+$^{1}$ School of Computer Science and Engineering, Northeastern University, China  
+ $^{2}$ Department of Computer Science and Technology, Institute for AI, Tsinghua University, China  
+Beijing National Research Center for Information Science and Technology, China
+
+# Abstract
+
+Retrieval-Augmented Generation (RAG) enhances the performance of Large Language Models (LLMs) by incorporating external knowledge. However, LLMs still encounter challenges in effectively utilizing the knowledge from retrieved documents, often being misled by irrelevant or noisy information. To address this issue, we introduce RankCoT, a knowledge refinement method that incorporates reranking signals in generating CoT-based summarization for knowledge refinement based on given query and all retrieval documents. During training, RankCoT prompts the LLM to generate Chain-of-Thought (CoT) candidates based on the query and individual documents. It then fine-tunes the LLM to directly reproduce the best CoT from these candidate outputs based on all retrieved documents, which requires LLM to filter out irrelevant documents during generating CoT-style summarization. Additionally, RankCoT incorporates a self-reflection mechanism that further refines the CoT outputs, resulting in higher-quality training data. Our experiments demonstrate the effectiveness of RankCoT, showing its superior performance over other knowledge refinement models. Further analysis reveals that RankCoT can provide shorter but effective refinement results, enabling the generator to produce more accurate answers. All code and data are available at https://github.com/NEUIR/RankCoT.
+
+# 1 Introduction
+
+Retrieval-Augmented Generation (RAG) (Lewis et al., 2020b; Guu et al., 2020; Ram et al., 2023; Shi et al., 2024) empowers Large Language Models (LLMs) to access external knowledge, providing up-to-date information during the generation process. RAG models have demonstrated their effectiveness in mitigating the hallucination problem commonly encountered by LLMs (Shuster et al.,
+
+![](images/4195761aeae5ba8f0d3c5684eab8516e89ba800b28c4fbd1b84eb598bd2188fa.jpg)  
+Figure 1: Illustration of implicit ranking and summarization in the RankCoT model. We present how knowledge refinement can be achieved by incorporating reranking into CoT-based summarization.
+
+2021), enhancing the performance of LLMs, such as GPT-4 (Achiam et al., 2023) and LLaMA (Touvron et al., 2023), in different NLP tasks.
+
+RAG models typically use dense retrieval methods (Karpukhin et al., 2020; Xiong et al., 2021) to retrieve query-relevant documents from external knowledge bases. These documents, along with the query, are then fed as the input context into LLMs (Ram et al., 2023). Thriving on their incontext learning capabilities (Brown et al., 2020; Dong et al., 2023), LLMs are able to extract relevant semantics from retrieved documents and generate appropriate responses to address the given query. However, the potential knowledge conflict between external knowledge and parameterized memory still poses a challenge for LLMs in generating precise responses (Chen et al., 2024b; Asai et al., 2024b).
+
+Many RAG models focus on building modular RAG pipelines to enhance retrieval performance (Gao et al., 2024; Asai et al., 2024a). These models primarily aim to refine the retrieved knowl
+
+edge by assessing the relevance of each document to the query and subsequently filtering out irrelevant ones (Yan et al., 2024; Asai et al., 2024a). However, the reranking model still requires feeding the remaining documents into LLMs, which means that query-unrelated content within a relevant document may still mislead the generators (Xu et al., 2024a). Some models address this problem by prompting LLMs to summarize query-relevant knowledge from the retrieved documents, thereby reducing the influence of irrelevant information (Vig et al., 2022; Yu et al., 2024a; Xu et al., 2024a). This summarization approach often incorporates information from unrelated documents as part of the summaries, resulting in the introduction of noise. Both the reranking and summarization modules have advantages for knowledge refinement. However, in existing RAG systems, these modules are typically modeled separately by prompting the same LLMs.
+
+This paper presents RankCoT, a knowledge refinement method that combines the strengths of both ranking and summarization to effectively enhance the process for retrieval result refinement. As shown in Figure 1, we feed both the query and all retrieved documents into the RankCoT model, which incorporates reranking signals in generating CoT-based summarization as knowledge refinements, thereby aiding LLMs in generating more accurate responses for answering the given query. During training the RankCoT model, we independently feed the query and retrieved document to the LLM, asking it to generate several Chain-of-Thought (CoT) responses to answer the question, which can be considered as summarization results. We then design the self-refinement model to prompt LLMs to answer the question according to these sampled CoTs, helping to refine the CoT results for more effective training. If the refined CoT contains the ground truth answer, it is considered a positive refinement result, while those that do not contain the ground truth answer are considered negative refinements.
+
+Our experiments demonstrate that RankCoT outperforms all baseline models, achieving over a  $2\%$  improvement. Notably, RankCoT proves effective across LLMs of various scales. It generates shorter knowledge refinement results compared to both reranking and summarization methods, while enhancing the response accuracy of generator. Further analysis reveals that RankCoT successfully incorporates ground truth answers into the knowl
+
+edge refinement results, while also including more query-relevant content. Additionally, RankCoT effectively extracts crucial semantics from the retrieved documents and alleviates the conflict between retrieved contents and internal knowledge.
+
+# 2 Related Work
+
+Retrieval-Augmented Generation (RAG) aims to enhance Large Language Models (LLMs) by enabling them to access external knowledge bases, providing up-to-date information during the generation process (Shi et al., 2024; Ram et al., 2023). This approach has demonstrated promising results across various NLP tasks, including open-domain question answering (Izacard et al., 2023), code generation (Zhou et al., 2023), and dialogue (Shuster et al., 2022). In these RAG models, retrieved documents are typically used as context to assist LLMs in generating more accurate responses (Ram et al., 2023). However, the conflict between the external knowledge and the parametric memory of LLMs often undermines the effectiveness of current RAG systems (Asai et al., 2024b; Xie et al., 2024; Chen et al., 2024b).
+
+To mitigate the potentially negative impact of retrieved knowledge, existing models focus on refining the external knowledge through various modules designed to help LLMs generate more precise responses. Earlier works concentrate on reranking the retrieved documents (Yu et al., 2023; Shi et al., 2024; Yu et al., 2024b), while others employ query-focused summarization techniques (Vig et al., 2022; Xu et al., 2023) to reduce noise. However, reranking models often overlook noise within individual passages, and summarization models may fail to account for query-document relevance, sometimes incorporating misleading content in the summarization results. Chain-of-Note (Yu et al., 2024a) attempts to instruct LLMs to generate query-related notes when answering a given query. This model incorporates the knowledge refinement process into the reasoning stage (Wei et al., 2022) and heavily relies on the capabilities of LLMs, which may limit its applicability in RAG systems (Gao et al., 2024).
+
+Modular RAG systems (Gao et al., 2024; Xu et al., 2024c) focus on refining external knowledge through different modules implemented by LLMs, which have become a key trend in the RAG area. For instance, Self-RAG (Asai et al., 2024a) uses different tags for adaptive retrieval (Jiang et al., 2023) and self-reflection to refine knowledge. Some ap
+
+proaches also focus on reformulating queries to identify more useful documents for answering questions (Yan et al., 2024; Trivedi et al., 2023). Yan et al. (2024) introduce a retrieval evaluator that acts as a judge to trigger query reformulation, search, and knowledge refinement actions to supply more accurate evidence for generation.
+
+To further improve the performance of modular RAG systems, these models focus on fine-tuning various components of the RAG framework. Some efforts aim to align the information needs between the retriever and the generator by optimizing the retrievers based on feedback from the generation models (Yu et al., 2023; Shi et al., 2024; Izacard and Grave, 2021). Lin et al. (2024) adapt LLMs within the RAG setting by constructing instruction-tuning data for Supervised Fine-Tuning (SFT), enabling the models to better leverage the retrieved documents. Additionally, Li et al. (2024) use Direct Preference Optimization (DPO) (Rafailov et al., 2024) to jointly optimize the modules in a RAG system, aligning their data preferences.
+
+# 3 Methodology
+
+As illustrated in Figure 2, this section introduces the RankCoT method. First, we introduce the preliminary of knowledge refinement in Retrieval-Augmented Generation (RAG) systems (Sec. 3.1). And then we describe how to optimize LLMs to produce more effective chain-of-thoughts for knowledge refinement (Sec. 3.2).
+
+# 3.1 Preliminary of Knowledge Refinement in Retrieval-Augmented Generation Systems
+
+Given a query  $q$  and a set of retrieved documents  $D = \{d_{1}, d_{2}, \ldots, d_{n}\}$ , vanilla RAG system (Ram et al., 2023) uses retrieved documents as context and leverages the in-context learning method to help the generation model  $\mathcal{M}_{\mathrm{Gen}}$  produce the answer  $y_{\mathrm{Gen}}$ :
+
+$$
+(q, D) \rightsquigarrow \mathcal {M} _ {\text {G e n}} \rightsquigarrow y _ {\text {G e n}}. \tag {1}
+$$
+
+Instead of directly feeding retrieved documents to the generation model  $(\mathcal{M}_{\mathrm{Gen}})$ , some RAG models (Gao et al., 2024; Asai et al., 2024a) design various modules to refine retrieved documents  $D$ :
+
+$$
+(q, D) \rightsquigarrow \mathcal {M} _ {\mathrm {K R}} \rightsquigarrow y _ {\mathrm {K R}}. \tag {2}
+$$
+
+The refined knowledge  $y_{\mathrm{KR}}$  is then passed to the generation model to mitigate the negative impact
+
+of retrieval noise:
+
+$$
+\left(q, y _ {\mathrm {K R}}\right)\rightsquigarrow \mathcal {M} _ {\text {G e n}} \rightsquigarrow y _ {\text {G e n}}. \tag {3}
+$$
+
+In the rest of this subsection, we will introduce different methods for implementing the knowledge refinement model  $\mathcal{M}_{\mathrm{KR}}$  in Eq. 2, including reranking, summarization, and RankCoT.
+
+Reranking. Following previous work (Asai et al., 2024a), we prompt LLMs to evaluate the relevance of the  $i$ -th retrieved document  $d_{i}$  with respect to the query  $q$ , and output a binary label  $y_{\mathrm{Rerank}}^{i}$  for filtering out noisy documents:
+
+$$
+y _ {\text {R e r a n k}} ^ {i} = \operatorname {L L M} \left(\operatorname {I n s t r u c t} _ {\text {R e r a n k}}, q, d _ {i}\right), \tag {4}
+$$
+
+where  $\mathrm{Instruct}_{\mathrm{Rerank}}$  prompts the LLM to assess the relevance between  $q$  and  $d_{i}$ . The prediction label  $y_{\mathrm{Rerank}}^{i}$  can be "YES" or "NO", indicating whether the  $i$ -th document  $d_{i}$  is relevant or irrelevant to the query  $q$ . We then retain the documents predicted as "YES" and construct the filtered document set  $\{d_1,\ldots ,d_k\}$ . The knowledge refinement result  $y_{\mathrm{KR}}$  is then represented as:
+
+$$
+y _ {\mathrm {K R}} = d _ {1} \oplus \dots \oplus d _ {k}, \tag {5}
+$$
+
+where  $\oplus$  is the concatenation operation.
+
+Summarization. Another approach for knowledge refinement is summarization, which aims to extract query-related content from the retrieved documents (Vig et al., 2022). The knowledge refinement result can be obtained as:
+
+$$
+y _ {\mathrm {K R}} = \operatorname {L L M} \left(\operatorname {I n s t r u c t} _ {\text {S u m}}, q, D\right), \tag {6}
+$$
+
+where  $\mathrm{InstructSum}$  is the instruction prompting the LLM to generate a summary. Unlike reranking, summarization directly generates the refined knowledge, avoiding the need to feed raw documents to the generation model  $\mathcal{M}_{\mathrm{Gen}}$ .
+
+RankCoT. RankCoT further incorporates a Chain-of-Thought (CoT) (Wei et al., 2022) into the knowledge refinement process:
+
+$$
+y _ {\mathrm {K R}} = \operatorname {L L M} \left(\operatorname {I n s t r u c t} _ {\mathrm {C o T}}, q, D\right), \tag {7}
+$$
+
+where  $\mathrm{Instruct}_{\mathrm{CoT}}$  is the instruction that prompts the LLM to generate CoT. RankCoT incorporates the chain-of-thought reasoning as the knowledge refinement result  $y_{\mathrm{KR}}$  to extract relevant knowledge from retrieved documents  $D$ , thereby assisting RAG models in answering the query. Unlike summarization, RankCoT integrates the reranking mechanism during the chain-of-thought generation (Sec. 3.2) to mitigate the influence of noisy documents (Liu et al., 2024; Xie et al., 2024).
+
+![](images/8d59c0901d5d77b193b8b8e9d1195daddb99078e0600d7b67bebfa322320cbd4.jpg)  
+Figure 2: Illustration of RankCoT.
+
+# 3.2 Knowledge Refinement through Ranking Chain-of-Thoughts
+
+To generate tailored knowledge refinement results  $y_{\mathrm{KR}}$  for RAG modeling, RankCoT optimizes the LLM  $(\mathcal{M})$  by incorporating reranking into the CoT generation process. Furthermore, we introduce a self-reflection method to further refine the CoT results, mitigating the risk of overfitting to undesired CoT patterns during training RankCoT.
+
+# 3.2.1 Reranking Modeling in CoT Generation
+
+To learn the query-document relevance, we feed each document  $d_{i}$  into the LLM  $(\mathcal{M})$  and sample one CoT output  $y_{\mathrm{CoT}}(d_i)$ :
+
+$$
+y _ {\mathrm {C o T}} \left(d _ {i}\right) \sim \mathcal {M} \left(q, d _ {i}\right). \tag {8}
+$$
+
+Next, we gather all generated CoT results from each document in the retrieved document set  $D$  to form the candidate CoT set  $Y_{\mathrm{CoT}}$ :
+
+$$
+Y _ {\mathrm {C o T}} = \left\{y _ {\mathrm {C o T}} \left(d _ {1}\right), \dots , y _ {\mathrm {C o T}} \left(d _ {n}\right) \right\}. \tag {9}
+$$
+
+We treat the CoT result  $y_{\mathrm{CoT}} \in Y_{\mathrm{CoT}}$  that contains the ground truth answer as the positive  $y_{\mathrm{CoT}}^{+}$ , while the result that does not contain the ground truth answer is regarded as the negative  $y_{\mathrm{CoT}}^{-}$ .
+
+Finally, we can optimize the LLM  $(\mathcal{M})$  to assign higher generation probabilities to positive knowledge refinement results  $y_{\mathrm{CoT}}^{+}$  than the negative ones  $y_{\mathrm{CoT}}^{-}$ . The training process is implemented with the Direct Preference Optimization
+
+(DPO) method (Rafailov et al., 2024):
+
+$$
+\mathcal {L} = - \mathbb {E} _ {\left(q, y _ {\mathrm {C o T}} ^ {+}, y _ {\mathrm {C o T}} ^ {-}\right) \sim \mathcal {T}} \left[ \log \sigma \left(\beta \log \right. \right.
+$$
+
+$$
+\left. \frac {\mathcal {M} \left(y _ {\mathrm {C o T}} ^ {+} \mid q , D\right)}{\mathcal {M} ^ {\operatorname {R e f}} \left(y _ {\mathrm {C o T}} ^ {+} \mid q , D\right)} - \beta \log \left. \frac {\mathcal {M} \left(y _ {\mathrm {C o T}} ^ {-} \mid q , D\right)}{\mathcal {M} ^ {\operatorname {R e f}} \left(y _ {\mathrm {C o T}} ^ {-} \mid q , D\right)}\right) \right], \tag {10}
+$$
+
+where  $\beta$  is a hyperparameter and  $\sigma$  is the Sigmoid function.  $\mathcal{M}^{\mathrm{Ref}}$  is the reference model, which remains frozen during training.
+
+In DPO training, we input all documents  $D$  into the model  $\mathcal{M}$  and aim to assign a higher probability to the positive knowledge refinement result  $y_{\mathrm{CoT}}^{+}$ , which is individually generated from one of the retrieved documents. This guides the model  $\mathcal{M}$  to rerank the retrieved documents  $D$  when generating chain-of-thought as the refinement.
+
+# 3.2.2 CoT Refinement through Self-Reflection
+
+While these generated CoTs help the RAG model generate more accurate answers, the generated CoT results of LLMs may contain undesired patterns, such as "According to the document" and "the reasoning process is". These training patterns can mislead the LLM  $(\mathcal{M})$  to overfit these CoT results during training (Gudibande et al., 2023). To address this problem, RankCoT proposes a self-reflection method to refine the CoT results  $Y_{\mathrm{CoT}}$ .
+
+Specifically, we first sample the CoT outputs  $\tilde{y}_{\mathrm{CoT}}(d_i)$  by feeding the given query  $q$  and each
+
+document  $d_{i}$  to the LLM:
+
+$$
+\tilde {y} _ {\mathrm {C o T}} \left(d _ {i}\right) \sim \mathcal {M} \left(\text {I n s t r u c t} _ {\mathrm {C o T}}, q, d _ {i}\right), \tag {11}
+$$
+
+where  $\mathrm{Instruct}_{\mathrm{CoT}}$  is used to prompt the LLM to generate a chain-of-thought. Then the CoT result  $\tilde{y}_{\mathrm{CoT}}(d_i)$  is refined as  $y_{\mathrm{CoT}}(d_i)$  by using the same LLM  $(\mathcal{M})$ :
+
+$$
+y _ {\mathrm {C o T}} \left(d _ {i}\right) = \mathcal {M} \left(\text {I n s t r u c t} _ {\text {R e f}}, q, \tilde {y} _ {\mathrm {C o T}} \left(d _ {i}\right)\right), \tag {12}
+$$
+
+where the instruction InstructRef prompts the LLM  $(\mathcal{M})$  to answer the given query  $q$  based on the initial CoT result  $\tilde{y}_{\mathrm{CoT}}(d_i)$ . Such a self-reflection mechanism helps to extract more query-related contents from the initial CoT  $\tilde{y}_{\mathrm{CoT}}(d_i)$ , producing higher-quality data to optimize LLMs. Finally, we collect the refined CoT results to form  $Y_{\mathrm{CoT}}(d_i) = \{y_{\mathrm{CoT}}(d_1), \dots, y_{\mathrm{CoT}}(d_n)\}$ , which is used to train the LLM through DPO (Eq. 10).
+
+# 4 Experimental Methodology
+
+In this section, we describe the datasets, baselines, evaluation metrics, and implementation details in our experiments. More experimental details are shown in Appendix A.4.
+
+Datasets. In our experiments, we follow previous work (Lin et al., 2024) and utilize the instruction tuning datasets to train and evaluate RAG models. For all datasets and baselines, we use BGE-large (Chen et al., 2024a) to retrieve documents from the MS MARCO V2.1 document collection (Bajaj et al., 2016). We select six datasets for evaluation, including NQ (Kwiatkowski et al., 2019), HotpotQA (Yang et al., 2018), Trivia QA (Joshi et al., 2017), PopQA (Mallen et al., 2023), ASQA (Stelmakh et al., 2022), and MARCO QA (Bajaj et al., 2016), which require models to retrieve factual knowledge or conduct more complex reasoning to help answer the given query. All data statistics are shown in Table 1.
+
+Baselines. In our experiments, we compare RankCoT with the Vanilla RAG (No Refinement) model and three knowledge refinement models, including Rerank, Summary, and CoT, which are described in Sec. 3.1. For the vanilla RAG model, we follow previous work (Ram et al., 2023) and feed 5 retrieved documents as context to answer the question. For the Rerank model, we prompt the LLM to evaluate the relevance between the query and retrieved documents (Asai et al., 2024a; Li et al., 2024). If the document is relevant to the
+
+<table><tr><td>Split</td><td>#Query</td><td>Tasks</td></tr><tr><td rowspan="2">Train</td><td>15,444</td><td>Open-Domain QA</td></tr><tr><td>8,527</td><td>Reasoning</td></tr><tr><td rowspan="2">Dev</td><td>1,752</td><td>Open-Domain QA</td></tr><tr><td>912</td><td>Reasoning</td></tr><tr><td>Test</td><td>20,294</td><td>Open-Domain QA</td></tr></table>
+
+Table 1: Data Statistics.
+
+question, it outputs "YES" and retains the document, otherwise, it outputs "NO" and discards the document. The Summary model and CoT model prompt the LLM to extract query-related knowledge from retrieved documents using summarization and Chain-of-Thought (Wei et al., 2022) formats to conclude query-related knowledge from retrieved documents.
+
+Evaluation Metrics. Following Xu et al. (2024b), we utilize Rouge-L as evaluation metric for MARCO QA task. Following Gao et al. (2023), we utilize String-EM as evaluation metric for ASQA. For other tasks, we use the Accuracy metric for evaluation.
+
+Implementation Details. We implement our RankCoT model using Llama3-8B-Instruct (Touvron et al., 2023) as the backbone model. To construct a training dataset for DPO training, we ask Llama3-8B-Instruct to generate CoTs using 10 retrieved documents independently and use the same model to refine CoTs. During training, we feed 5 relevant documents as external knowledge and ask the LLM to reproduce the refined CoT results. We use LoRA (Hu et al., 2022) method to fine-tune Llama3-8B-Instruct, with  $\beta$  set to 0.1 and the learning rate set to 2e-5.
+
+For the RAG model, we concatenate the generated CoT with the query to let Llama3-8B-Instruct generate the final answer. In addition, we also use LLMs of different scales, such as MiniCPM3-4B (Hu et al., 2024) and Qwen2.5-14B-Instruct (Yang et al., 2024), to build the RAG model and evaluate the generalization ability of RankCoT.
+
+# 5 Evaluation Result
+
+In this section, we first evaluate the performance of various RAG methods, followed by ablation studies to examine the impact of self reflection module and different training strategies. We then investigate the characteristics of RankCoT by analyzing the knowledge utilization capabilities of RAG models
+
+<table><tr><td>Method</td><td>NQ (acc)</td><td>HotpotQA (acc)</td><td>TriviaQA (acc)</td><td>PopQA (acc)</td><td>ASQA (str-em)</td><td>MARCO (rouge)</td><td>Avg.</td></tr><tr><td colspan="8">Llama3-8B-Instruct</td></tr><tr><td>No Refinement</td><td>45.68</td><td>29.43</td><td>82.85</td><td>35.60</td><td>38.79</td><td>20.73</td><td>42.18</td></tr><tr><td>Rerank</td><td>46.18</td><td>30.30</td><td>83.51</td><td>36.20</td><td>39.92</td><td>20.72</td><td>42.81</td></tr><tr><td>Summary</td><td>44.27</td><td>28.05</td><td>82.09</td><td>33.67</td><td>37.81</td><td>22.67</td><td>41.32</td></tr><tr><td>CoT</td><td>45.33</td><td>26.36</td><td>81.45</td><td>34.13</td><td>40.25</td><td>19.52</td><td>41.17</td></tr><tr><td>RankCoT</td><td>47.41</td><td>32.21</td><td>85.18</td><td>41.17</td><td>41.02</td><td>20.84</td><td>44.64</td></tr><tr><td colspan="8">MiniCPM3-4B</td></tr><tr><td>No Refinement</td><td>42.51</td><td>24.93</td><td>80.91</td><td>32.53</td><td>24.31</td><td>13.55</td><td>36.46</td></tr><tr><td>RankCoT</td><td>48.78</td><td>33.13</td><td>85.20</td><td>36.87</td><td>35.85</td><td>24.59</td><td>44.07</td></tr><tr><td colspan="8">Qwen2.5-14B-Instruct</td></tr><tr><td>No Refinement</td><td>47.66</td><td>29.70</td><td>79.49</td><td>36.97</td><td>44.73</td><td>18.50</td><td>42.84</td></tr><tr><td>RankCoT</td><td>49.98</td><td>33.91</td><td>86.68</td><td>44.45</td><td>41.94</td><td>24.62</td><td>46.93</td></tr></table>
+
+using different knowledge refinement models. We also examine the effectiveness of the refined knowledge generated by RankCoT through answering consistency in Appendix A.3. The case study is conducted in Appendix A.7.
+
+# 5.1 Overall Performance
+
+This section presents the knowledge refinement performance of different models, as shown in Table 2. Additional baseline comparison results are shown in Appendix A.2.
+
+The evaluation results reveal that these three knowledge refinement models, Rerank, Summary, and CoT, show distinct performance. Specifically, Rerank exhibits a slight improvement, while both Summary and CoT lead to a decrease in RAG performance. This highlights the challenge of effectively refining knowledge for RAG modeling. In contrast, RankCoT demonstrates a  $2.5\%$  improvement over vanilla RAG model, indicating its effectiveness in providing more meaningful refinements that help LLMs better answer questions. Furthermore, RankCoT outperforms the Rerank model with a  $1.8\%$  improvement and avoids the need to feed raw passages into the LLM twice for knowledge refinement and question answering.
+
+Then we present the performance of RankCoT by applying it to different RAG systems implemented with LLMs of various scales. The results indicate that RankCoT maintains its effectiveness across different RAG configurations, yielding a  $7.6\%$  improvement over the MiniCPM3-4B-based RAG model and a  $4.1\%$  improvement over
+
+Table 2: Overall Performance of RAG System with Different Knowledge Refinement Models. We use Llama3-8B-Instruct as the backbone model for different knowledge refinement models and apply RankCoT to the RAG system, which is implemented with Llama3-8B-Instruct, MiniCPM3-4B, and Qwen2.5-14B-Instruct.  
+
+<table><tr><td>Method</td><td>NQ (acc)</td><td>HotpotQA (acc)</td><td>TriviaQA (acc)</td><td>Avg.</td></tr><tr><td>Vanilla RAG</td><td>45.68</td><td>29.43</td><td>82.85</td><td>52.65</td></tr><tr><td colspan="5">Inference w/ Finetuned QA Model</td></tr><tr><td>Rerank</td><td>45.96</td><td>29.88</td><td>83.04</td><td>52.96</td></tr><tr><td>Summary</td><td>38.46</td><td>25.29</td><td>76.13</td><td>45.63</td></tr><tr><td>CoT</td><td>41.63</td><td>33.23</td><td>83.24</td><td>52.70</td></tr><tr><td colspan="5">SFT Training</td></tr><tr><td>Rerank</td><td>47.02</td><td>31.21</td><td>84.16</td><td>54.13</td></tr><tr><td>Summary</td><td>43.36</td><td>28.39</td><td>82.26</td><td>51.34</td></tr><tr><td>CoT</td><td>43.43</td><td>27.23</td><td>82.80</td><td>51.15</td></tr><tr><td colspan="5">DPO Training</td></tr><tr><td>Rerank</td><td>46.80</td><td>30.70</td><td>84.70</td><td>54.07</td></tr><tr><td>Summary</td><td>46.63</td><td>30.61</td><td>83.95</td><td>53.73</td></tr><tr><td>RankCoT</td><td>47.41</td><td>32.21</td><td>85.18</td><td>54.97</td></tr><tr><td>w/o Reflect</td><td>46.70</td><td>30.59</td><td>83.82</td><td>53.70</td></tr></table>
+
+Table 3: Ablation Study. Both SFT and DPO methods optimize knowledge refinement models using self-reflection labels. RankCoT w/o Reflect refers to that the RankCoT model is optimized using unrefined CoT.
+
+the RAG model implemented with Qwen2.5-14B-Instruct. These findings demonstrate that RankCoT has strong generalization ability in knowledge refinement, enabling LLMs of different scales to effectively leverage external knowledge. For more experiments, please see Appendix A.5.
+
+# 5.2 Ablation Study
+
+We conduct ablation studies to evaluate the effectiveness of various training strategies.
+
+As shown in Table 3, we first conduct the approach from prior work (Lin et al., 2024), where a fine-tuned QA model is used for knowledge re
+
+<table><tr><td rowspan="2">Method</td><td colspan="3">Has-Answer</td><td colspan="3">Miss-Answer</td><td colspan="3">Internal Knowledge</td></tr><tr><td>NQ</td><td>HotpotQA</td><td>TriviaQA</td><td>NQ</td><td>HotpotQA</td><td>TriviaQA</td><td>NQ</td><td>HotpotQA</td><td>TriviaQA</td></tr><tr><td>LLM w/o RAG</td><td>46.41</td><td>42.99</td><td>80.48</td><td>6.51</td><td>14.73</td><td>46.04</td><td>100.0</td><td>100.0</td><td>100.0</td></tr><tr><td>Vanilla RAG</td><td>63.45</td><td>57.69</td><td>88.76</td><td>2.41</td><td>8.89</td><td>19.70</td><td>77.30</td><td>67.89</td><td>90.80</td></tr><tr><td>Rerank</td><td>64.09</td><td>57.09</td><td>88.74</td><td>3.14</td><td>10.87</td><td>27.41</td><td>78.20</td><td>68.28</td><td>91.24</td></tr><tr><td>Summary</td><td>61.40</td><td>53.20</td><td>87.20</td><td>3.02</td><td>10.19</td><td>25.27</td><td>76.89</td><td>63.23</td><td>89.93</td></tr><tr><td>CoT</td><td>62.90</td><td>51.76</td><td>87.53</td><td>2.53</td><td>7.69</td><td>22.91</td><td>77.19</td><td>59.77</td><td>89.38</td></tr><tr><td>RankCoT</td><td>65.44</td><td>58.92</td><td>90.49</td><td>4.10</td><td>13.03</td><td>30.84</td><td>80.42</td><td>70.81</td><td>92.61</td></tr></table>
+
+Table 4: RAG Performance by Using Different Knowledge Refinement Models. We conduct three testing scenarios to evaluate the knowledge usage of RAG systems, including Has-Answer, Miss-Answer and Internal Knowledge.
+
+finement. We then train several knowledge refinement models—Rerank, Summary, and CoT—using the self-reflection mechanism introduced by RankCoT. Specifically, the self-reflection mechanism involves feeding the knowledge refinement results into LLMs to generate self-reflection results based on these inputs. In this setup, SFT methods select self-reflection results containing ground truth answers to train knowledge refinement models, while DPO methods select both positive and negative responses—those that contain ground truth answers and those that do not—for training.
+
+After fine-tuning with QA supervision, the LLM is able to generate knowledge refinement results using different prompts. However, the evaluation results illustrate that these QA models present limited effectiveness compared to vanilla RAG model. We then train the knowledge refinement models using training signals refined by the LLM itself. Among the SFT-based models, Rerank achieves the best performance, illustrating that the reranking signals can be easily learned by LLMs through SFT. Using the DPO training method, both Summary and RankCoT show significant improvements over these knowledge refinement models using the SFT strategy. Furthermore, RankCoT outperforms all knowledge refinement models, demonstrating its effectiveness in producing effective knowledge refinement results to help LLMs generate more accurate answers. By replacing self-refined CoTs with the raw CoT outcomes during DPO training, RankCoT achieves a  $1.3\%$  decline, showing the effectiveness of our self-reflection mechanism.
+
+# 5.3 Knowledge Usage Performance of Different Refinement Models
+
+In this experiment, we evaluate the ability of RankCoT to assist RAG models in leveraging external knowledge to generate final answers. We compare RankCoT with three knowledge refinement models: Rerank, Summary, and CoT.
+
+As shown in Table 4, we conduct three testing
+
+The internal knowledge scenario is a scenario in which the retrieval of knowledge models from external sources is not possible. This scenario involves cases where the retrieved documents contain the correct (golden) answer. This scenario evaluates whether the knowledge refinement model can effectively extract key information from these documents to aid the LLM in answering the question. The Miss-Answer scenario, on the other hand, deals with cases where the retrieved documents do not include the golden answer. This scenario further tests the ability of the knowledge refinement models to minimize the impact of retrieved noise. Finally, the Internal Knowledge scenario examines the ability of different knowledge refinement models to handle conflicts between internal and external knowledge.
+
+As shown in the evaluation results, RankCoT outperforms all knowledge refinement models across all datasets in the Has-Answer scenario. This demonstrates the effectiveness of RankCoT in incorporating more query-relevant information from retrieved documents, thereby enhancing the accuracy of the RAG system in this scenario. In the Miss-Answer scenario, the performance of vanilla RAG models significantly drops compared to LLMs w/o RAG, indicating that query-irrelevant documents mislead LLMs into producing incorrect answers. However, the use of different knowledge refinement models mitigates this performance decline. Among these models, RankCoT exhibits the most substantial improvements, demonstrating its effectiveness in filtering out noisy information from retrieval and reducing the misleading information of irrelevant documents. In the internal knowledge scenario, knowledge refinement models-Rerank, Summary, and CoT-perform comparably or even worse than vanilla RAG model, illustrating that existing methods are less effective in addressing the knowledge conflict issue. In contrast, RankCoT outperforms these knowledge refinement models, demonstrating its ability to pro
+
+![](images/3aa402721b858f526fa817009d0069ad4e1b7ae2aed8e9e90a58905828895dbc.jpg)  
+(a) Similarity between Query (b) Hit Rate of Ground Truth and Refined Knowledge. Answers.
+
+![](images/ab2fa738fec11380db252f4509ce41e93938623a914d582297e17fb35744f2e2.jpg)  
+Figure 3: Quality of Refined Knowledge Generated by Different Models. In Figure 3(a), we first estimate the text similarity between the query and the knowledge refinement results using the BGE model (Chen et al., 2024a). Then, we calculate the hit rate of these knowledge refinement results in Figure 3(b), which evaluates whether the ground truth answers are included in the knowledge refinement results.
+
+![](images/f542abced1f7dc0225080c844fb87046faccec5ca008a9b20ef6a1db30d7bb7a.jpg)  
+(a) Average Length.
+
+![](images/0dbc20f9d7153033b7a89b2b94b96f34b06ca799a4d17648b98bedf914350d83.jpg)  
+(b) Length Change Ratio.  
+Figure 4: The Length of Knowledge Refinement Results Produced by Different Models. We first present the average length of the refinement results in Figure 4(a). Then, the length change ratio relative to vanilla LLMs is illustrated in Figure 4(b).
+
+vide more tailored knowledge refinement results. The RankCoT-produced knowledge refinement results effectively alleviate knowledge conflicts, aiding LLMs in better utilizing both internal and external knowledge.
+
+# 5.4 Characteristics of the Refined Knowledge Produced by RankCoT
+
+This experiment further explores the characteristics of RankCoT-produced knowledge refinement results by estimating both the quality and length of the refined knowledge. We conduct the experiment on the NQ and TriviaQA datasets, specifically in the Has-answer test scenario, where the retrieved documents contain the ground truth answers.
+
+Refinement Quality. As shown in Figure 3, we evaluate the quality of the knowledge refinement results based on query relevance and the hit rate of the golden answer. Specifically, the similarity scores between the query and the refined knowledge generated by three different knowledge refinement mod
+
+els are presented in Figure 3(a). RankCoT achieves the highest similarity score with the query, demonstrating its effectiveness in retaining more query-related information from the retrieved documents. Furthermore, we show the hit rate of the ground truth answer in Figure 3(b). As indicated by the evaluation results, Rerank achieves the highest hit rates, while Summary performs the worst. This outcome likely stems from the fact that the Rerank model only selects the most relevant document, whereas the Summary model must extract key information, inevitably discarding some of the relevant contents that contain the ground truth answers. Although RankCoT is also a summarization-style knowledge refinement model, it achieves higher hit rates, showing that RankCoT can capture more ground truth answers in its refinement results.
+
+Length of Knowledge Refinement. Subsequently, we present the results of knowledge refinement lengths for different models in Figure 4. As shown in Figure 4(a), the summarization-style knowledge refinement methods, Summary and RankCoT, significantly reduce the length of the refined knowledge compared to the Rerank model. Notably, RankCoT achieves the shortest refinement length, demonstrating its effectiveness in minimizing the consumption of prompt inputs for LLMs (Mu et al., 2023) while helping to reduce inference time (please see Appendix A.6 for related experiments). Additionally, we investigate the length change ratio across different training methods in Figure 4(b). As shown in the results, these SFT methods generally result in shorter knowledge refinement outputs, illustrating that SFT encourages the summarization-style knowledge refinement model to overfit training signals (Li et al., 2024). In contrast, the DPO training method helps these knowledge refinement models produce longer results, facilitating more flexible responses that incorporate more crucial knowledge.
+
+# 6 Conclusion
+
+This paper proposes RankCoT, a knowledge refinement method that leverages the strengths of both ranking and summarization to effectively refine the knowledge from retrieval results, thereby aiding LLMs in generating more accurate responses. Our experimental studies show that RankCoT can effectively refine external knowledge and balance the utilization of internal and external knowledge. In-depth analysis reveals that the CoT generated by
+
+our method has a high similarity to the query and a high ground truth answer hit rate. In addition, RankCoT shows its effectiveness in minimizing the consumption of prompt inputs for LLMs while helping to reduce inference time.
+
+# Limitations
+
+Although RankCoT demonstrates its effectiveness in refining retrieved knowledge for RAG systems, the quality of the refinement is still constrained by the capabilities of LLMs. Specifically, RankCoT is optimized using the DPO method, which relies on LLMs to generate meaningful chosen and rejected pairs during optimization. Therefore, the generation of meaningful preference pairs for optimization still heavily depends on the performance of the LLMs.
+
+# Acknowledgments
+
+This work is partly supported by the Natural Science Foundation of China under Grant (No. 62206042 and No. 62137001), CCF-zhipu Large Model Innovation Fund (No. 202403), the Joint Funds of Natural Science Foundation of Liaoning Province (No. 2023-MSBA-081), and the Fundamental Research Funds for the Central Universities under Grant (No. N2416012). This work is also supported by the AI9Stars community.
+
+The authors would like to thank Weiqing Yang for his help in drawing figures during the writing process of the paper, and Sijia Yao for her careful and patient review of the final draft of the paper and her suggestions for revision.
+
+# References
+
+Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. 2023. Gpt-4 technical report. ArXiv preprint.  
+Shourya Aggarwal, Divyanshu Mandowara, Vishwajeet Agrawal, Dinesh Khandelwal, Parag Singla, and Dinesh Garg. 2021. *Explanations for Common-senseQA: New Dataset and Models*. In *Proceedings of ACL*, pages 3050-3065.  
+Aida Amini, Saadia Gabriel, Shanchuan Lin, Rik Koncel-Kedziorski, Yejin Choi, and Hannaneh Hajishirzi. 2019. MathQA: Towards interpretable math word problem solving with operation-based formalisms. In Proceedings of NAACL-HLT, pages 2357-2367.
+
+Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. 2024a. Self-RAG: Learning to retrieve, generate, and critique through self-reflection. In Proceedings of ICLR.  
+Akari Asai, Zexuan Zhong, Danqi Chen, Pang Wei Koh, Luke Zettlemoyer, Hannaneh Hajishirzi, and Wentau Yih. 2024b. Reliable, adaptable, and attributable language models with retrieval. ArXiv preprint.  
+Payal Bajaj, Daniel Campos, Nick Craswell, Li Deng, Jianfeng Gao, Xiaodong Liu, Rangan Majumder, Andrew McNamara, Bhaskar Mitra, Tri Nguyen, et al. 2016. Ms marco: A human generated machine reading comprehension dataset. ArXiv preprint.  
+Jonathan Berant, Andrew Chou, Roy Frostig, and Percy Liang. 2013. Semantic parsing on Freebase from question-answer pairs. In Proceedings of EMNLP, pages 1533-1544.  
+Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, and Dario Amodei. 2020. Language models are few-shot learners. In Proceedings of NeurIPS.  
+Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, and Zheng Liu. 2024a. Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. ArXiv preprint.  
+Jiawei Chen, Hongyu Lin, Xianpei Han, and Le Sun. 2024b. Benchmarking large language models in retrieval-augmented generation. In Proceedings of AAAI, 16, pages 17754-17762.  
+Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, et al. 2021. Training verifiers to solve math word problems. ArXiv preprint.  
+Qingxiu Dong, Lei Li, Damai Dai, Ce Zheng, Jingyuan Ma, Rui Li, Heming Xia, Jingjing Xu, Zhiyong Wu, Tianyu Liu, et al. 2023. A survey on in-context learning. ArXiv preprint.  
+Tianyu Gao, Howard Yen, Jiatong Yu, and Danqi Chen. 2023. Enabling large language models to generate text with citations. In Proceedings of EMNLP, pages 6465-6488.  
+Yunfan Gao, Yun Xiong, Meng Wang, and Haofen Wang. 2024. Modular rag: Transforming rag systems into lego-like reconfigurable frameworks. ArXiv preprint.
+
+Mor Geva, Daniel Khashabi, Elad Segal, Tushar Khot, Dan Roth, and Jonathan Berant. 2021. Did aristotle use a laptop? a question answering benchmark with implicit reasoning strategies. Transactions of the Association for Computational Linguistics, pages 346-361.  
+Arnav Gudibande, Eric Wallace, Charlie Snell, Xinyang Geng, Hao Liu, Pieter Abbeel, Sergey Levine, and Dawn Song. 2023. The false promise of imitating proprietary llms. ArXiv preprint.  
+Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Ming-Wei Chang. 2020. Retrieval augmented language model pre-training. In Proceedings of ICML, pages 3929-3938.  
+Edward J Hu, Phillip Wallis, Zeyuan Allen-Zhu, Yanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. 2022. Lora: Low-rank adaptation of large language models. In Proceedings of ICLR.  
+Shengding Hu, Yuge Tu, Xu Han, Chaoqun He, Ganqu Cui, Xiang Long, Zhi Zheng, Yewei Fang, Yuxiang Huang, Weilin Zhao, et al. 2024. Minicpm: Unveiling the potential of small language models with scalable training strategies. ArXiv preprint.  
+Gautier Izacard and Edouard Grave. 2021. Distilling knowledge from reader to retriever for question answering. In Proceedings of ICLR.  
+Gautier Izacard, Patrick S. H. Lewis, Maria Lomeli, Lucas Hosseini, Fabio Petroni, Timo Schick, Jane Dwivedi-Yu, Armand Joulin, Sebastian Riedel, and Edouard Grave. 2023. Atlas: Few-shot learning with retrieval augmented language models. Journal of Machine Learning Research, pages 251:1-251:43.  
+Zhengbao Jiang, Frank F Xu, Luyu Gao, Zhiqing Sun, Qian Liu, Jane Dwivedi-Yu, Yiming Yang, Jamie Callan, and Graham Neubig. 2023. Active retrieval augmented generation. In Proceedings of EMNLP, pages 7969-7992.  
+Mandar Joshi, Eunsol Choi, Daniel Weld, and Luke Zettlemoyer. 2017. TriviaQA: A large scale distantly supervised challenge dataset for reading comprehension. In Proceedings of ACL, pages 1601-1611.  
+Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. Dense passage retrieval for open-domain question answering. In Proceedings of EMNLP, pages 6769-6781.  
+Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, Kristina Toutanova, Llion Jones, Matthew Kelcey, Ming-Wei Chang, Andrew M. Dai, Jakob Uszkoreit, Quoc Le, and Slav Petrov. 2019. Natural questions: A benchmark for question answering research. Transactions of the Association for Computational Linguistics, pages 452-466.
+
+Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020a. BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In Proceedings of ACL, pages 7871-7880.  
+Patrick S. H. Lewis, Ethan Perez, Aleksandra Pik-tus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Kuttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. 2020b. Retrieval-augmented generation for knowledge-intensive NLP tasks. In Proceedings of NeurIPS.  
+Xinze Li, Sen Mei, Zhenghao Liu, Yukun Yan, Shuo Wang, Shi Yu, Zheni Zeng, Hao Chen, Ge Yu, Zhiyuan Liu, et al. 2024. Rag-ddr: Optimizing retrieval-augmented generation using differentiable data rewards. ArXiv preprint.  
+Xi Victoria Lin, Xilun Chen, Mingda Chen, Weijia Shi, Maria Lomeli, Richard James, Pedro Rodriguez, Jacob Kahn, Gergely Szilvasy, Mike Lewis, Luke Zettlemoyer, and Wen tau Yih. 2024. RA-DIT: Retrieval-augmented dual instruction tuning. In Proceedings of ICLR.  
+Wang Ling, Dani Yogatama, Chris Dyer, and Phil Blunson. 2017. Program induction by rationale generation: Learning to solve and explain algebraic word problems. In Proceedings of ACL, pages 158-167.  
+Nelson F Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2024. Lost in the middle: How language models use long contexts. Transactions of the Association for Computational Linguistics, pages 157-173.  
+Alex Mallen, Akari Asai, Victor Zhong, Rajarshi Das, Daniel Khashabi, and Hannaneh Hajishirzi. 2023. When not to trust language models: Investigating effectiveness of parametric and non-parametric memories. In Proceedings of ACL, pages 9802-9822.  
+Jesse Mu, Xiang Lisa Li, and Noah Goodman. 2023. Learning to compress prompts with gist tokens. In Proceedings of NeurIPS.  
+Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2024. Direct preference optimization: Your language model is secretly a reward model. In Proceedings of NeurIPS.  
+Ori Ram, Yoav Levine, Itay Dalmedigos, Dor Muhlgay, Amnon Shashua, Kevin Leyton-Brown, and Yoav Shoham. 2023. In-context retrieval-augmented language models. Transactions of the Association for Computational Linguistics, pages 1316-1331.  
+Weijia Shi, Sewon Min, Michihiro Yasunaga, Minjoon Seo, Richard James, Mike Lewis, Luke Zettlemoyer, and Wen-tau Yih. 2024. REPLUG: Retrievalaugmented black-box language models. In Proceedings of NAACL-HLT, pages 8371-8384.
+
+Kurt Shuster, Spencer Poff, Moya Chen, Douwe Kiela, and Jason Weston. 2021. Retrieval augmentation reduces hallucination in conversation. In Proceedings of EMNLP Findings, pages 3784-3803.  
+Kurt Shuster, Jing Xu, Mojtaba Komeili, Da Ju, Eric Michael Smith, Stephen Roller, Megan Ung, Moya Chen, Kushal Arora, Joshua Lane, et al. 2022. Blenderbot 3: a deployed conversational agent that continually learns to responsibly engage. *ArXiv* preprint.  
+Ivan Stelmakh, Yi Luan, Bhuwan Dhingra, and Ming-Wei Chang. 2022. ASQA: Factoid questions meet long-form answers. In Proceedings of EMNLP, pages 8273-8288.  
+Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. 2019. CommonsenseQA: A question answering challenge targeting commonsense knowledge. In Proceedings of NAACL-HLT, pages 4149-4158.  
+Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models. ArXiv preprint.  
+Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2023. Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions. In Proceedings of ACL, pages 10014-10037.  
+Jesse Vig, Alexander R. Fabbri, Wojciech Kryscinski, Chien-Sheng Wu, and Wenhao Liu. 2022. Exploring neural models for query-focused summarization. In *Findings of the Association for Computational Linguistics: NAACL*, pages 1455-1468.  
+Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, brian richter, Fei Xia, Ed H. Chi, Quoc V Le, and Denny Zhou. 2022. Chain of thought prompting elicits reasoning in large language models. In Advances in Neural Information Processing Systems.  
+Jian Xie, Kai Zhang, Jiangjie Chen, Renze Lou, and Yu Su. 2024. Adaptive chameleon or stubborn sloth: Revealing the behavior of large language models in knowledge conflicts. In Proceedings of ICLR.  
+Lee Xiong, Chenyan Xiong, Ye Li, Kwok-Fung Tang, Jialin Liu, Paul N. Bennett, Junaid Ahmed, and Arnold Overwijk. 2021. Approximate nearest neighbor negative contrastive learning for dense text retrieval. In Proceedings of ICLR.  
+Fangyuan Xu, Weijia Shi, and Eunsol Choi. 2024a. RE-COMP: Improving retrieval-augmented LMs with context compression and selective augmentation. In Proceedings of ICLR.  
+Ruochen Xu, Song Wang, Yang Liu, Shuohang Wang, Yichong Xu, Dan Iter, Pengcheng He, Chenguang Zhu, and Michael Zeng. 2023. Lmgqs: A large-scale
+
+dataset for query-focused summarization. In Proceedings of EMNLP Findings, pages 14764-14776.  
+Shicheng Xu, Liang Pang, Mo Yu, Fandong Meng, Huawei Shen, Xueqi Cheng, and Jie Zhou. 2024b. Unsupervised information refinement training of large language models for retrieval-augmented generation. ArXiv preprint.  
+Zhipeng Xu, Zhenghao Liu, Yibin Liu, Chenyan Xiong, Yukun Yan, Shuo Wang, Shi Yu, Zhiyuan Liu, and Ge Yu. 2024c. Activerag: Revealing the treasures of knowledge via active learning. ArXiv preprint.  
+Shi-Qi Yan, Jia-Chen Gu, Yun Zhu, and Zhen-Hua Ling. 2024. Corrective retrieval augmented generation. ArXiv preprint.  
+An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, et al. 2024. Qwen2.5 technical report. ArXiv preprint.  
+Yi Yang, Wen-tau Yih, and Christopher Meek. 2015. WikiQA: A challenge dataset for open-domain question answering. In Proceedings of EMNLP, pages 2013-2018.  
+Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D. Manning. 2018. HotpotQA: A dataset for diverse, explainable multi-hop question answering. In Proceedings of EMNLP, pages 2369-2380.  
+Wenhao Yu, Hongming Zhang, Xiaoman Pan, Peixin Cao, Kaixin Ma, Jian Li, Hongwei Wang, and Dong Yu. 2024a. Chain-of-note: Enhancing robustness in retrieval-augmented language models. In Proceedings of EMNLP, pages 14672-14685.  
+Yue Yu, Wei Ping, Zihan Liu, Boxin Wang, Jiaxuan You, Chao Zhang, Mohammad Shoeybi, and Bryan Catanzaro. 2024b. Rankrag: Unifying context ranking with retrieval-augmented generation in llms. ArXiv preprint.  
+Zichun Yu, Chenyan Xiong, Shi Yu, and Zhiyuan Liu. 2023. Augmentation-adapted retriever improves generalization of language models as generic plug-in. In Proceedings of ACL, pages 2421-2436.  
+Shuyan Zhou, Uri Alon, Frank F. Xu, Zhengbao Jiang, and Graham Neubig. 2023. Docprompting: Generating code by retrieving the docs. In Proceedings of ICLR.
+
+<table><tr><td>Method</td><td>NQ (acc)</td><td>HotpotQA (acc)</td><td>TriviaQA (acc)</td><td>Avg.</td></tr><tr><td>Vanilla RAG</td><td>45.68</td><td>29.43</td><td>82.85</td><td>52.65</td></tr><tr><td>Self-RAG (2024a)</td><td>39.90</td><td>24.59</td><td>78.35</td><td>47.61</td></tr><tr><td>Recomp (2024a)</td><td>40.47</td><td>25.36</td><td>79.04</td><td>48.29</td></tr><tr><td>SEGENC (2022)</td><td>40.47</td><td>24.89</td><td>76.73</td><td>47.36</td></tr><tr><td>RankCoT</td><td>47.41</td><td>32.21</td><td>85.18</td><td>54.93</td></tr></table>
+
+Table 5: Overall Performance of More Baselines.
+
+# A Appendix
+
+# A.1 License
+
+This section summarizes the licenses of the datasets used in our experiments.
+
+All of these datasets under their respective licenses and agreements allow for academic use: Natural Questions (CC-BY-SA-3.0 License); PopQA, Commonsense QA, Wiki QA, MARCO QA, StrategyQA, and Grade School Math 8K (MIT License); Web Questions and HotpotQA (CC-BY-4.0 License); TriviaQA, ASQA, Algebra QA with Rationales, and Math QA (Apache 2.0 License); Explanations for CommonsenseQ (CDLA-Sharing-1.0 License); Yahoo! Answers QA shows its terms of use at website<sup>1</sup>.
+
+# A.2 Additional Baseline Comparison Results
+
+This section presents the comparison results between RankCoT and several baseline models.
+
+In this experiment, we compare RankCoT with four baselines: vanilla LLM, Self-RAG, Recomp, and SEGENC. Self-RAG (Asai et al., 2024a) optimizes Llama3-8B-Instruct to retrieve documents on demand and ranks them by reflecting the retrieved documents using reflection tokens. SEGENC (Vig et al., 2022) is a Query-Focused Summarization model, initialized from the BART model (Lewis et al., 2020a), which summarizes documents based on a given query. Recomp (Xu et al., 2024a) proposes a method to compress retrieved documents, reducing the computational overhead of language models during inference.
+
+As shown in Table 5, Self-RAG, Recomp, and SEGENC all show performance degradation compared to vanilla RAG. This indicates that ranking, compressing, or summarizing documents inevitably lead to information loss, thereby reducing response accuracy. In contrast, RankCoT not only incorporates advantages of ranking and summarization, but also generates a CoT that preserves
+
+![](images/eb284efe6628ab047d1b86527d10d3a48701817a2af13f8bc295320842b0b6f1.jpg)  
+Figure 5: QA Consistency of the RAG Model Using Different Knowledge Refinement Models.
+
+as much useful information as possible. This approach reduces the input length for the QA model while improving response accuracy.
+
+# A.3 QA Consistency Using Different Knowledge Refinements
+
+This experiment evaluates the QA consistency based on different knowledge refinement results. Since the experiment is conducted in the internal knowledge scenario, all queries can be accurately answered by the RAG model without any external knowledge.
+
+As illustrated in Figure 5, we use the TriviaQA dataset to conduct the experiment. For each query, we input both the query and the refinement results from different models, then sample responses from the RAG model 300 times. The ratio of correct answers is calculated and denoted as Accuracy, which serves to evaluate the QA consistency of the RAG model. A higher accuracy reflects that the knowledge refinement results help the RAG model consistently produce correct answers.
+
+As shown in the evaluation results, RankCoT demonstrates its effectiveness by achieving an average accuracy of approximately  $91.3\%$ , outperforming both refinement baselines, Rerank and Summary. The accuracy of the Rerank and Summary methods shows significant variation, indicating that the knowledge refinements produced by both models still contain knowledge that causes the RAG model to lack consistency in its answers. In contrast, after applying RankCoT, the accuracy becomes concentrated at either 0 or 1, demonstrating that it better supports the RAG model in maintaining answer consistency.
+
+<table><tr><td>Split</td><td>Task</td><td>Dataset</td><td>Metric</td><td>Raw</td><td>Filtered</td></tr><tr><td rowspan="10">Training</td><td rowspan="6">Open-Domain QA</td><td>Commonsense QA (2019)</td><td>Accuracy</td><td>4,000</td><td>3,037</td></tr><tr><td>Math QA (2019)</td><td>Accuracy</td><td>4,000</td><td>3,509</td></tr><tr><td>Web Questions (2013)</td><td>Accuracy</td><td>3,578</td><td>1,810</td></tr><tr><td>Wiki QA (2015)</td><td>Rouge-L</td><td>840</td><td>840</td></tr><tr><td>Yahoo! Answers QA</td><td>Rouge-L</td><td>4,000</td><td>4,000</td></tr><tr><td>MARCO QA (2016)</td><td>Rouge-L</td><td>4,000</td><td>4,000</td></tr><tr><td rowspan="4">Reasoning</td><td>Algebra QA with Rationales (2017)</td><td>Accuracy</td><td>2,527</td><td>2,222</td></tr><tr><td>Explanations for CommonsenseQ (2021)</td><td>Accuracy</td><td>4,000</td><td>3,202</td></tr><tr><td>Grade School Math 8K (2021)</td><td>Accuracy</td><td>4,000</td><td>3,090</td></tr><tr><td>StrategyQA (2021)</td><td>Accuracy</td><td>1,860</td><td>925</td></tr><tr><td rowspan="6">Evaluation</td><td rowspan="6">Open-Domain QA</td><td>Natural Questions (2019)</td><td>Accuracy</td><td>2,837</td><td>-</td></tr><tr><td>HotpotQA (2018)</td><td>Accuracy</td><td>5,600</td><td>-</td></tr><tr><td>TriviaQA (2017)</td><td>Accuracy</td><td>5,359</td><td>-</td></tr><tr><td>PopQA (2023)</td><td>Accuracy</td><td>3,000</td><td>-</td></tr><tr><td>ASQA (2022)</td><td>STR-EM</td><td>948</td><td>-</td></tr><tr><td>MARCO QA (2016)</td><td>Rouge-L</td><td>3,000</td><td>-</td></tr></table>
+
+Table 6: Data Statistics.  
+
+<table><tr><td>Method</td><td>NQ (acc)</td><td>HotpotQA (acc)</td><td>TriviaQA (acc)</td><td>Avg.</td></tr><tr><td colspan="5">MiniCPM3-4B</td></tr><tr><td>Vanilla RAG</td><td>42.51</td><td>24.93</td><td>80.91</td><td>49.45</td></tr><tr><td>RankCoT (Not Aligned)</td><td>48.78</td><td>33.13</td><td>85.20</td><td>55.70</td></tr><tr><td>RankCoT (Aligned)</td><td>45.68</td><td>27.52</td><td>79.21</td><td>50.80</td></tr><tr><td colspan="5">Qwen2.5-14B-Instruct</td></tr><tr><td>Vanilla RAG</td><td>47.66</td><td>29.70</td><td>79.49</td><td>52.28</td></tr><tr><td>RankCoT (Not Aligned)</td><td>49.98</td><td>33.91</td><td>86.68</td><td>56.86</td></tr><tr><td>RankCoT (Aligned)</td><td>52.17</td><td>37.34</td><td>86.49</td><td>58.67</td></tr></table>
+
+Table 7: Overall Performance of the RankCoT Model with and without Backbone Model Alignment in the RAG System. "Not aligned" refers to using Llama3-8B-Instruct as the backbone model while applying RankCoT to RAG systems built with MiniCPM3-4B or Qwen2.5-14B-Instruct. "Aligned" refers to using consistent backbone models, either MiniCPM3-4B or Qwen2.5-14B-Instruct, throughout the RankCoT and RAG components.
+
+# A.4 Additional Experimental Details
+
+In this subsection, we first describe the process of constructing the training data and then show the prompt templates used in our experiments.
+
+Data Preprocessing for RankCoT. The quantities of our training and evaluation data, along with the corresponding evaluation metrics, are presented in Table 6. The "Filtered" column indicates the number of training samples used for DPO training.
+
+During RankCoT training, we collect ten datasets, obtain 32,805 samples, and process them as described in Section 3.2. Since the generated CoT can either be fully correct or incorrect, we cannot form preference data pairs from these generated CoT candidates. Thus, we filter out such samples and divide the remaining ones into training and validation datasets with a 9:1 ratio.
+
+Prompt Templates. The prompts used for CoT
+
+generation (InstructCoT) are shown in Figure 6(a). The prompt used for question answering is illustrated in Figure 7. Additionally, the prompts used for CoT refinement (InstructRef) are shown in Figure 6(b). Finally, Figure 8 presents the prompts used for implementing baselines.
+
+# A.5 Evaluating RankCoT under Aligned Parameter Scales in RAG Systems
+
+Although RankCoT (Llama3-8B-Instruct) can be applied to various RAG systems implemented with LLMs of different scales and demonstrates its effectiveness, the performance gains may diminish when larger-scale LLMs are used as the generation models in RAG systems. This is because larger LLMs inherently possess stronger knowledge refinement capabilities. To assess the impact of model scale on RankCoT, we conduct experiments by aligning the parameter scales of both the RAG models and the RankCoT model.
+
+As shown in Table 7, we compare two variants of RankCoT: one without alignment and one with alignment. Both variants implement RankCoT using either Llama3-8B-Instruct or the same LLMs employed by the corresponding RAG models. The results demonstrate that when Qwen2.5-14B-Instruct is used consistently across components, the knowledge refinement process is more effective, leading to improved answer accuracy. In contrast, aligning all components with MiniCPM3-4B results in lower overall performance compared to the unaligned setting. This outcome is expected, as the unaligned setting leverages Llama3-8B-Instruct for knowledge refinement, which benefits from a larger model size and stronger refinement capabilities.
+
+<table><tr><td rowspan="2">Method</td><td colspan="2">NQ</td><td colspan="2">HotpotQA</td><td colspan="2">TriviaQA</td><td rowspan="2">Number of LLM Calls</td></tr><tr><td>Refine</td><td>Gen.</td><td>Refine</td><td>Gen.</td><td>Refine</td><td>Gen.</td></tr><tr><td>Vanilla RAG</td><td>-</td><td>54s</td><td>-</td><td>57s</td><td>-</td><td>58s</td><td>1</td></tr><tr><td>Rerank</td><td>212s</td><td>56s</td><td>638s</td><td>58s</td><td>280s</td><td>56s</td><td>2</td></tr><tr><td>Summary</td><td>101s</td><td>11s</td><td>101s</td><td>11s</td><td>89s</td><td>9s</td><td>2</td></tr><tr><td>CoT</td><td>106s</td><td>12s</td><td>111s</td><td>13s</td><td>96s</td><td>10s</td><td>2</td></tr><tr><td>RankCoT</td><td>82s</td><td>10s</td><td>84s</td><td>10s</td><td>74s</td><td>8s</td><td>2</td></tr></table>
+
+Table 8: Inference Time of RAG with Different Knowledge Refinement Models. A total of 500 samples are drawn from each of the NQ, HotpotQA, and TriviaQA datasets. All experiments are conducted using the Llama3-8B-Instruct model with the vLLM inference engine and a batch size of 60. "Refine" denotes the time spent on knowledge extraction, while "Gen." refers to the time taken for answer generation.
+
+# A.6 Inference Time of RAG with Different Knowledge Refinement Models
+
+This section reports the inference time of various knowledge refinement models across different datasets, as summarized in Table 8.
+
+During inference, both RankCoT and the baseline refinement models make the same number of calls to the LLM: once for refining the retrieved knowledge and once for answering the question based on the refined knowledge. Compared to vanilla RAG, most of the additional time introduced by these models is spent on the knowledge refinement step. Among all evaluated methods, RankCoT achieves the lowest inference latency. As discussed in Section 5.4, RankCoT generates the shortest knowledge refinement results, which largely accounts for its superior inference efficiency.
+
+# A.7 Case Study
+
+In Table 9, we present a case study to illustrate the effectiveness of the RankCoT model. For the given query, it asks about "Australia's location in the world and region". And the retrieved documents contain both related and unrelated information about the geographical location of Australia.
+
+The summarization method captures some geographical information about Australia from the retrieved documents, such as "being in the Southern Hemisphere and located between the Indian and Pacific Oceans". However, these descriptions offer only a broad geographical scope rather than directly answering the query about the regional location of Australia. Consequently, the LLM is misled by the ambiguous information in the summarized documents and generates inaccurate answers. Moreover, the summarization contains some irrelevant information, such as "7,682,300 square kilometers" and "smallest continent and the sixth-largest country".
+
+In contrast, RankCoT refines the retrieved docu
+
+ments in a question-answering based summarization manner by generating a Chain-of-Thought (CoT). These CoT results are constructed by sequentially integrating information from different retrieved documents, while ranking and prioritizing the most query-relevant knowledge. Rather than directly summarizing keypoint information from retrieved documents, RankCoT identifies crucial geographical attributes in each document and organizes them in a structured reasoning result. At the start of the CoT, broad geographical attributes, such as "Southern Hemisphere" and "between the Indian and Pacific Oceans" are strengthened, as they appear consistently across documents. More specific regional information, such as "Oceania", is ranked higher in the reasoning process, ensuring that the final CoT provides the most accurate regional classification. This demonstrates that RankCoT is not merely a direct extraction or summary of retrieved documents, but rather a refined reasoning chain that aligns closely with the query.
+
+# Query: What is australia's location in the world and region? Ground Truth: Oceania Documents:
+
+(1) ... Australia is both a continent and a country. It is located to the south of Indonesia and Papua New Guinea, in the southeast part of the Indian Ocean. New Zealand lies to its east, and most of the Indian Ocean lies to its west. ... Australia is the world's smallest continent, though it is also the sixth-largest country in the world. The total land area of Australia is 7,682,300 sq. km.  
+(2) ... Australia is the smallest continent and the largest country in Oceania located between the Indian Ocean and Pacific Ocean in the Southern hemisphere. ... Australia is separated from Papua New Guinea by the Coral Sea and Torres Strait to the northeast; from Indonesia by the Timor and Arafura seas to the northwest; from Coral Sea Islands Territory by the Great Barrier Reef; from New Zealand by Tasman Sea to the southeast;  
+(3) ... Australia is located south of the equator and is bordered by the Indian Ocean and the Pacific Ocean.
+
+# Summary:
+
+Summary: Here is a concise summary: Australia is located in the Southern Hemisphere, between the Indian Ocean and the Pacific Ocean. It is the world's smallest continent and the sixth-largest country, covering an area of 7,682,300 square kilometers. Australia is surrounded by the Indian Ocean to the west, the Timor, Arafura, and Coral Seas to the north, the Pacific Ocean to the east, and the Tasman Sea and Southern Ocean to the south. The country is made up of six states (Western Australia, South Australia, Victoria, Tasmania, New South Wales, and Queensland) and two territories (the Northern Territory and the Australian Capital Territory)
+
+Answer: Australia is located in the Southern Hemisphere, between the Indian Ocean and the Pacific Ocean. (X)
+
+# RankCoT:
+
+Chain-of-Thought: According to the passages, Australia is:
+
+1. An island continent located in the Southern Hemisphere.  
+2. Situated between the Indian Ocean and the Pacific Ocean.  
+3. Located south of the equator.  
+4. Surrounded by the Indian Ocean to the west, the Timor, Arafura, and Coral Seas to the north, the Pacific Ocean to the east, and the Tasman Sea and Southern Ocean to the south.  
+5. An island continent in Oceania, which is a region that includes Australia, New Zealand, and many Pacific island nations.
+
+So, Australia's location in the world is in the Southern Hemisphere, in the region of Oceania.
+
+Answer: Australia is an island continent located in the Southern Hemisphere, in the region of Oceania.
+
+Table 9: Case Study. We randomly sample one case from the NQ dataset to show the knowledge refinement result. Different colors are used to annotate the key information from the retrieved knowledge retained by different knowledge refinement models: Pink for Summary, Blue for RankCoT. And we also highlight the key points that can help answer the query in Purple.
+
+# Prompts for CoT Generation
+
+[TASK]: Web Questions, Wiki QA, Yahoo! Answers QA, MARCO QA, Grade School Math 8K, StrategyQA, Natural Questions, HotpotQA, TriviaQA, PopQA, ASQA [Instruction]: Passage:{passage} Based on these passages, answer the question below. Question:{question} Let's think step by step.
+
+[TASK]:CommonsenseQA,MathQA,AlgebraQA with Rationales,   
+Explanations for CommonsenseQ   
+[Instruction]:Passage:{passage}   
+Based on these passages, please answer the multiple choice question below.   
+Question:{question}   
+Let's think step by step.
+
+(a) CoT Generation.
+
+# Prompts for CoT Refinement
+
+[TASK]: Web Questions, Wiki QA, Yahoo! Answers QA, MARCO QA, Grade School Math 8K, StrategyQA  
+[Instruction]: Task Description:  
+1. Read the given question and related chain of thought to gather relevant information.  
+2. The content of the chain of thought is the thinking process that may be used to answer the question.  
+3. If the chain of thought don't work, please answer the question based on your own knowledge.  
+4. Give a short answer to a given question.  
+Question:{question}  
+Chain of Thought:{CoT}
+
+[TASK]:CommonsenseQA,MathQA,AlgebraQA with Rationales,   
+Explanations for CommonsenseQ   
+[Instruction]:Task Description:   
+1. Read the given question and related chain of thought to gather relevant information.   
+2. The content of the chain of thought is the thinking process that may be used to answer the question.   
+3.If the chain of thought don't work, please answer the question based on your own knowledge.   
+4.Output only the choice between \*\*\*\* and \*\*\*\* at first,then give a short answer to the multiple choice question.   
+Question:{question}   
+Chain of Thought:{CoT}
+
+(b) CoT Refinement.
+
+Figure 6: Prompt Templates Used in RankCoT.
+
+# Prompts for Question Answering
+
+[TASK]: Natural Questions, HotpotQA, TriviaQA, PopQA
+
+[Instruction]: Task Description:
+
+1. Read the given question and related chain of thought to gather relevant information.  
+2. The content of the chain of thought is the thinking process that may be used to answer the question.  
+3. If the chain of thought don't work, please answer the question based on your own knowledge.  
+4. Please answer the question and only output the answer. Question:{question}
+
+Chain of Thought:{COT}
+
+[TASK]: ASQA
+
+[Instruction]: Task Description:
+
+1. Read the given question and related chain of thought to gather relevant information.  
+2. The content of the chain of thought is the thinking process that may be used to answer the question.  
+3. If the chain of thought don't work, please answer the question based on your own knowledge.  
+4. Answer the following question. The question may be ambiguous and have multiple correct answers, and in that case, you have to provide a long-form answer including all correct answers.
+
+Question:{question}
+
+Chain of Thought:{COT}
+
+[TASK]: MARCO QA
+
+[Instruction]: Task Description:
+
+1. Read the given question and related chain of thought to gather relevant information.  
+2. The content of the chain of thought is the thinking process that may be used to answer the question.  
+3. If the chain of thought don't work, please answer the question based on your own knowledge.  
+4. Give a short answer to a given question.
+
+Question:{question}
+
+Chain of Thought:{COT}
+
+Figure 7: Prompt Templates Used for Question Answering.
+
+# Prompts for Baselines
+
+[MODEL]: Vanilla RAG
+
+[TASK]: All except ASQA
+
+[Instruction]: Background:{Passages}
+
+Question:{Question}Answer:
+
+[MODEL]: Vanilla RAG
+
+[TASK]: ASQA
+
+[Instruction]: Background:{Passages}
+
+Answer the following question. The question may be ambiguous and have multiple correct answers, and in that case, you have to provide a long-form answer including all correct answers.
+
+Question:{Question}Answer:
+
+[MODEL]:Rerank
+
+[TASK]: All
+
+[Instruction]: Given the following question and context, return YES if the context is relevant to the question and NO if it isn't. Question: {question} Context:{context} Relevant (YES / NO):
+
+[MODEL]: Summary Generation
+
+[TASK]: All
+
+[Instruction]: You are tasked with summarizing the context in a concise manner, focusing on the key information relevant to answering the given question. Question: {question}
+
+Context:{context}
+
+Provide a concise summary that captures the main points and relevant details from the context to address the question:
+
+[MODEL]: Summary Question Answering
+
+[TASK]: Natural Questions, HotpotQA, TriviaQA, PopQA
+
+[Instruction]: Task Description:
+
+1. Read the given question and related summary to gather relevant information.
+
+2. If the summary don't work, please answer the question based on your own knowledge.
+
+3. Please answer the question and only output the answer.
+
+Question:{question}
+
+Summary: {summary}
+
+[MODEL]: Summary Question Answering
+
+[TASK]:ASQAI
+
+[Instruction]: Task Description:
+
+1. Read the given question and related summary to gather relevant information.
+
+2. If the summary don't work, please answer the question based on your own knowledge.
+
+3. Answer the following question. The question may be ambiguous and have multiple correct answers, and in that case, you have to provide a long-form answer including all correct answers.
+
+Question:{question}
+
+Summary: {summary}
+
+[MODEL]: Summary Question Answering
+
+[TASK]: MARCO QAI
+
+[Instruction]: Task Description:
+
+1. Read the given question and related summary to gather relevant information.
+
+2. If the summary don't work, please answer the question based on your own knowledge.
+
+3. Give a short answer to a given question Question:{question}
+
+Summary: {summary}
+
+Figure 8: Prompt Templates Used for Implementing Baselines.

rankcotrefiningknowledgeforretrievalaugmentedgenerationthroughrankingchainofthoughts/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94211e2cce8ac41bbd7be5547786e0c8dcaf36d716d6271ee55b2b35141554c7
+size 795671

rankcotrefiningknowledgeforretrievalaugmentedgenerationthroughrankingchainofthoughts/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67b3c972563234914921e98dc67fc332af1d4e1213f2f8478da446a50d8c2735
+size 569108

rankingunraveledrecipesforllmrankingsinheadtoheadaicombat/b4993003-acb3-44c9-9c58-ba518e67c4d1_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3f8acd5e19b060e14669ffe596fc128f274764aa943d1dc00cc5f0897f63b7c
+size 92309

rankingunraveledrecipesforllmrankingsinheadtoheadaicombat/b4993003-acb3-44c9-9c58-ba518e67c4d1_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36eddb4dc45be4de9f83a15d540e37bd1f0333d3e9a021f2e8c72e8ebcf26f84
+size 108072

rankingunraveledrecipesforllmrankingsinheadtoheadaicombat/b4993003-acb3-44c9-9c58-ba518e67c4d1_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1f6103ec74ea7e8f566bedd56e44af043d2bb51fac7926e01a239b6e44c12a7
+size 631368

rankingunraveledrecipesforllmrankingsinheadtoheadaicombat/full.md

@@ -0,0 +1,437 @@
+# Ranking Unraveled: Recipes for LLM Rankings in Head-to-Head AI Combat
+
+Roland Daynauth*, Christopher Clarke*, Krisztian Flautner, Lingjia Tang, Jason Mars
+
+University of Michigan
+
+{daynauth, csclarke, manowar, lingjia, profmars} @umich.edu
+
+# Abstract
+
+Evaluating large language models (LLMs) is a complex task. Pairwise ranking, where humans compare LLM outputs based on predefined criteria, has become a leading approach. By aggregating these comparisons through algorithms such as Elo, rankings across multiple LLMs can be derived. However, applying ranking algorithms in LLM evaluation presents several challenges. Traditional systems like Elo, designed initially for structured competitions such as chess, often produce inconsistent and unstable rankings due to the dynamic and context-dependent nature of LLM performance. Despite the increasing reliance on these methods, a systematic study of ranking algorithms for LLM evaluation remains lacking. This paper examines the effectiveness of various ranking systems for head-to-head LLM comparisons. We define key principles for robust ranking, conduct extensive evaluations of different ranking algorithms, and analyze their stability, accuracy, and sensitivity to real-world conditions. Our findings offer insights into the limitations of existing approaches and provide guidelines for selecting the most appropriate ranking method based on evaluation objectives and resource constraints.
+
+# 1 Introduction
+
+Rapid progress in large language models (LLMs) has highlighted the critical need for LLM evaluation methodologies, especially for complex generation tasks (Guo et al., 2023). Traditionally, NLP evaluation has relied on benchmarks such as GLUE (Wang et al., 2019), SuperGLUE (Wang et al., 2020), and more recent LLM-focused benchmarks like LM-Eval (Gao et al., 2023). However, a significant limitation of these classical benchmarks is their reliance on established ground truth, making
+
+![](images/ddb51db38da027ccbf39e13a57e2e73e68d8846e76d7774ce9850bd0ae4b5095.jpg)  
+Figure 1: Different ranking algorithms can produce different rankings with the same human evaluation data, making it difficult to determine which algorithm is appropriate for various use cases.
+
+them unsuitable for evaluating more nuanced, diverse, and complex cases where such ground truth is difficult to define or entirely unattainable (Chiang et al., 2024). This problem is particularly evident in complex tasks such as open-ended text generation, conversational dialogue, and creative writing, where human judgment is paramount. Previous work has shown that these benchmark evaluation results do not correlate well with human LLM evaluations (Zheng et al., 2024; Irugalbandara et al., 2023).
+
+This shortcoming has led the NLP community to adopt pairwise human evaluation methods (Dubois et al., 2024; Zheng et al., 2024; Chiang et al., 2024). A widely recognized example is Chatbot Arena (Chiang et al., 2024), which directly compares models through "head-to-head combat," where users vote between responses from two competing models. This method aligns with the Elo rating system (Elo, 1978), initially developed to classify chess players according to their historical competition results and relative strengths. Elo and other ranking methods have gained popularity as structured meth
+
+ods incorporating human feedback and preferences to effectively rank models (Dettmers et al., 2024; Wu and Aji, 2023; Chiang et al., 2023).
+
+However, as our experiments later demonstrated, Elo and other established ranking algorithms can often produce inconsistent and sometimes unreliable results (Figure 1), generating conflicting ranks. Thus, it is critical to understand how to select the ranking algorithm appropriately. There has been a lack of comprehensive studies across multiple ranking algorithms applied to LLM evaluations.
+
+- First, a general framework for such a study has been lacking. It remains unclear which key metrics should be used to evaluate ranking algorithms and the factors that can affect the accuracy of the ranking algorithm.  
+- There has been no comprehensive study comparing multiple ranking algorithms or offering guidelines for selecting the most suitable in the context of LLM evaluations. Although prior work such as Boubdir et al. (2023) has explored some challenges associated with using Elo for LLM ranking, its focus is limited to Elo. In addition, our experiments show that recommendations from previous work are not always effective in practice.
+
+Choosing the best model can significantly impact the user experience of the application that incorporates it. This paper explores the critical issues surrounding ranking LLMs in head-to-head human evaluations. Our contributions are as follows.
+
+- We first systemically outline a set of fundamental properties that ranking systems should adhere to when evaluating LLMs.  
+- Second, we evaluate the performance of four widely used ranking algorithms (Elo, Bradley-Terry, Glicko, and Markov Chain) in two different evaluation scenarios (Arena Style vs. Controlled Style), analyzing their ability to maintain transitivity, ensure prediction accuracy, and remain stable across hyperparameter variations.  
+- Third, we conduct an in-depth analysis of these results, highlighting flaws in the current application of these popular algorithms. Despiteits wide use for LLM ranking, Elo remains unstable even after extensive tuning is applied.
+
+- Lastly, we propose systematic guidelines for selecting the most appropriate approach based on the specific characteristics of the evaluation task and the available resources.
+
+To our knowledge, this paper represents the first work that systematically addresses the complexities and intricacies of ranking LLMs across various methodologies. We release all code, data, and models to facilitate reproducibility and further research in this area.
+
+# 2 Background
+
+In this section, we formally define the task of pairwise ranking and briefly describe the suite of algorithms often used for aggregating ranking from pair-wise comparison results. The first three algorithms—Elo (Elo, 1978), Bradley-Terry (Bradley and Terry, 1952), and Glicko (Glickman, 1999)—are well-established and widely used in contexts such as sports competitions, while the Markov chain, based on a random walker model, is simpler in design and has demonstrated strong performance in previous studies (Callaghan et al., 2003, 2004).
+
+# 2.1 Ranking
+
+Let  $M$  be the set of models under evaluation where  $M = \{1,2,\dots ,m\}$ . For any pair  $i,j\in M$ , the result  $S_{ij}$  of a head-to-head match is defined as  $S_{ij} = 1$  when  $i$  beats  $j$  and  $S_{ij} = 0$  when  $i$  loses and  $S_{ij} = 0.5$  for ties.
+
+If  $\theta_{i},\theta_{j}$  represents the strength of  $i$  and  $j$  respectively, then the probability that  $S_{ij} = 1$ , written as  $p_{ij}$ , for  $\theta_{i},\theta_{j}$  is assumed to follow a probability distribution  $F$  such that (David, 1988):
+
+$$
+p _ {i j} = F \left(\theta_ {i} - \theta_ {j}\right) \tag {1}
+$$
+
+# 2.2 Elo Rating system
+
+The Elo rating system was first adopted by the United States Chess Federation (USCF) in 1960 to measure players' strengths (Elo, 1978). For two players  $i$  and  $j$  with relative strengths  $\theta_{i}$  and  $\theta_{j}$ , the probability that  $i$  beats  $j$ , is calculated as follows:
+
+$$
+p _ {i j} = \frac {1}{1 + 1 0 ^ {\left(\theta_ {j} - \theta_ {i}\right) / 4 0 0}} \tag {2}
+$$
+
+For any match between model  $i$  and  $j$ ,  $\theta_{i}$  (also called the Elo score for  $i$ ) is updated via the formula:
+
+$$
+\theta_ {i} ^ {\prime} = \theta_ {i} + k \times \left(S _ {i j} - p _ {i j}\right) \tag {3}
+$$
+
+where  $k$  is a hyperparameter know as the  $k$ -factor. The value of  $k$  determines the rate at which the Elo score of a player changes based on the outcome of the game.
+
+# 2.3 Bradley-Terry Model
+
+The Bradley-Terry model (Bradley and Terry, 1952) was introduced in 1952 and assumes that  $F$  in equation 1 follows a logistic distribution where
+
+$$
+p _ {i j} = \frac {e ^ {\theta_ {i}}}{e ^ {\theta_ {i}} + e ^ {\theta_ {j}}} \tag {4}
+$$
+
+The strengths  $\theta_{i}$  for each model can be estimated by solving the likelihood function  $\mathcal{L}(\theta)$  such that:
+
+$$
+\mathcal {L} (\theta) = \prod_ {i <   j} p _ {i j} ^ {y _ {i j}} p _ {j i} ^ {y _ {j i}} \tag {5}
+$$
+
+where  $y_{ij}$  is the number of times that model  $i$  beats  $j$  and  $p_{ij}$  is the probability of  $i$  beating  $j$ .
+
+# 2.4 Glicko Rating system
+
+The Glicko rating system (Glickman, 1999) was introduced in 1995 as an improvement to Elo. Glicko extends Elo by introducing a secondary parameter  $(\sigma)$  known as the rating deviation. This parameter measures the reliability of a player's rating score, acting as a confidence interval for the player's rank, decreasing based on the number of games played, regardless of wins or losses. This factor is important given that two players can potentially have the same score in a tournament despite a different number of previous matches. This situation is prevalent in LLM evaluations, as new models are frequently introduced and the number of games played between them vary significantly. In Appendix A.1 we explore the details of the algorithm.
+
+# 2.5 Markov Chain
+
+The Markov Chain Model is a nonparametric ranking algorithm adapted from Callaghan et al. (2003, 2004) and adapted into a Markov Chain Model by (Kvam and Sokol, 2006) (see Appendix A.2 for details). The Markov chain employs a series of random walkers traversing a graph, with each node representing a player/model in the tournament connected by an edge denoting a match between two nodes. Each walker chooses an initial node  $i$ , staying in the node or moving to an adjacent node  $j$  according to a single condition  $p$ . In a random match between  $i$  and  $j$ , the walker moves to the winning node with a probability of  $p$  for  $p > 0.5$ ,
+
+a hyperparameter chosen in advance. Every time a walker moves or stays on a node, it represents a vote for that node, and the sum of all votes from every walker represents the rank of that player/model. In practice, the optimal value for  $p$  can vary depending on the tournament. In our experiments, we use  $p = 0.8$  as the default value similar to Callaghan et al. (2003), which achieved stable results with  $p$  ranging from  $p = 0.75$  to  $p = 0.95$ .
+
+# 3 Desirable Properties of Ranking System
+
+The goal of a ranking system is to provide a structured approach to compare and order items based on their performance or quality. In the context of LLMs, an effective ranking system allows researchers, practitioners, and end-users to identify the most capable models for their specific needs. By simplifying decision-making and optimizing resource allocation, such a system highlights models that are likely to deliver superior performance in the real-world applications. However, implementing a robust ranking system requires addressing several key properties to ensure its validity and reliability. In this section, we identify and discuss three essential properties that a ranking system for LLMs should adhere to: 1) Transitivity, 2) Prediction Accuracy, and 3) Sensitivity to Hyperparameters and Battle Conditions.
+
+# 3.1 Transitivity
+
+Transitivity is a fundamental property for a ranking system to maintain consistency.
+
+Definition 3.1. For any three models  $i,j,k\in M$  their match-ups are transitive if  $p_{ij} > 0.5,p_{jk}>$  0.5 and  $p_{ik} > 0.5$  . If  $i,j,k$  are transitive, we say that a ranking system preserves transitivity if it ranks these models in order  $i,j,k$  . Transitivity ensures that the ranking is coherent and interpretable.
+
+Human evaluation data contains transitive and intransitive relationships, whereas ranking algorithms always produce perfectly transitive rankings. Transitive and intransitive relationships are mutually exclusive; however, intransitive relationships are inevitably lost, since ranking enforces transitivity. One of the key strengths of a ranking algorithm lies in its ability to preserve as much of the original transitive structure from human data as possible. Failure to do so can lead to paradoxical rankings, where the ordering of models becomes inconsistent, ultimately diminishing trust in the rating system.
+
+# 3.2 Prediction Accuracy
+
+Like transitivity, prediction accuracy is a critical metric that assesses how well the ranking system forecasts the results in head-to-head matches. It measures the probability that the ranking system correctly predicts the winner between any two LLMs. High prediction accuracy indicates that the ranking system is reliable and can be trusted to make informed decisions about the superiority of the model for a downstream task. For combat-style evaluation in LLM, where assessment is based on subjective human judgment, prediction accuracy acts as an indicator of the ranking system's ability to align with human preferences. This alignment is crucial for tasks like open-ended text generation or conversational agents, where the model's performance is best judged by human preference. Thus, the ranking system should reflect statistical superiority and resonate with qualitative human evaluations to truly enhance the utility and adoption of LLM.
+
+# 3.3 Sensitivity to Hyperparameters
+
+Another critical aspect of evaluating ranking algorithms is understanding their sensitivity to hyperparameters, which can significantly affect the consistency and stability of rankings (Boubdir et al., 2023). Ranking algorithms such as Elo have been shown to produce drastically different rankings based on the order in which matches are evaluated and the number of permutations (Boubdir et al., 2023). Since these hyperparameters can influence the behavior and outcomes of the ranking system, it is essential to choose the right settings to ensure reliable and consistent rankings. Sensitivity to hyperparameters means that the ranking system should respond to changes in these settings in a predictable and manageable way. An overly sensitive system might produce drastically different rankings with minor parameter changes, leading to instability and unreliability. In contrast, a system with too little sensitivity might fail to capture essential differences between models. In this work, we consider the impact of two different classes of hyperparameters/conditions: 1) Algorithmic Hyperparameters: parameters native to the ranking algorithm itself and 2) Matchup Distribution: Arena style vs. controlled style.
+
+Algorithmic Hyperparameters such as  $k$ -factor in Elo,  $p$ -value in Markov Chain, and  $\sigma$  in Glicko can significantly affect the rankings produced by
+
+<table><tr><td>Algorithm</td><td>Arena (%)</td><td>SLAM (%)</td></tr><tr><td>Elo</td><td>68.24</td><td>52.5</td></tr><tr><td>Markov</td><td>51.38</td><td>51.67</td></tr><tr><td>Glicko</td><td>56.54</td><td>53.33</td></tr><tr><td>Bradley-Terry</td><td>77.29</td><td>56.67</td></tr></table>
+
+Table 1: Ranking Algorithm Performance on preserving transitivity on Chatbot Arena & SLAM Dataset.
+
+these algorithms. Similarly, the distribution of matchups between models can also influence the rankings. In practice, certain models may participate in many more matchups than others, creating an uneven distribution of data. This skews the rankings because models with fewer matches do not have sufficient data to be ranked reliably, while those with more matchups dominate the rankings. The ranking system chosen should be robust to these variations and produce consistent and reliable rankings across different hyperparameters and battle conditions.
+
+# 4 Evaluating Ranking Algorithms
+
+Building on the properties discussed in section 3, this section outlines a series of experiments designed to assess the quality of traditional ranking systems and their use in LLM rankings. We investigate how each algorithm performs under different conditions and evaluate their effectiveness in producing stable and consistent rankings.
+
+# 4.1 Datasets
+
+To evaluate these ranking systems, we use two diverse human evaluation datasets representative of two evaluation styles:
+
+- Chatbot Arena (Chiang et al., 2024) is a dynamic platform where users evaluate two randomly selected model outputs head-to-head. As newer models are released, they are added to the arena to battle the older models. This data set consists of 57 models and a total of 244,978 match-ups. The distribution of matches varies wildly, with some models having up to 30416 matches or as few as 954. As such, we refer to this dataset as an Arena-Style dataset.  
+- SLAM (Irugalbandara et al., 2023) consists of pairwise match-ups between models evaluated using the SLAM tool on domain-specific task questions. Unlike Chatbot Arena, the
+
+SLAM platform is tightly controlled with a fixed number of models and a similar number of match-ups between models, thus making it a controlled style dataset. The SLAM dataset consists of 11 models with a total of 2858 match-ups, each model having at most 529 match-ups and at least 501. As such, we refer to this dataset as a Controlled Style dataset.
+
+These two diverse datasets enable us to assess ranking algorithms under varying conditions, offering a comprehensive view of their behavior. To evaluate dataset consistency, we apply Kendall's test (Kendall and Smith, 1940) (see A.6 for details), which confirms that any inconsistencies in the datasets are statistically insignificant.
+
+# 4.2 Preserving Transitivity
+
+Using the Chatbot Arena and SLAM datasets, we evaluated the ability of each ranking algorithm to preserve transitivity (see 3.1). We evaluated the transitive performance of each ranking algorithm by first calculating the number of triples within the data set. Three models  $i,j,k$  are part of a triple if  $i$  wins the majority of comparisons with  $j$ , and  $j$  wins the majority of comparisons with  $k$ , and transitivity is considered preserved if this order is maintained in the final ranking. Table 1 shows the performance of each ranking algorithm under the arena and controlled style evaluation scenarios. We observe that the Bradley-Terry model is the most performant, preserving transitivity in  $77.29\%$  and  $56.67\%$  of the triples, respectively, compared to its closest competitor, Elo, at  $68.24\%$  and  $52.5\%$ .
+
+Bradley-Terry uses Maximum Likelihood Estimation to calculate each player's strength based on the outcome of all their matches concurrently. This is in contrast to Elo, which updates each player's rating after each match sequentially and is therefore sensitive to the other in which matches are evaluated. This gives Bradley-Terry the advantage in preserving transitivity in larger datasets such as Chatbot Arena.
+
+Bradley-Terry consistently outperforms other algorithms in preserving transitivity, achieving the highest rankings across both datasets.
+
+# 4.3 Prediction Accuracy
+
+Next, we evaluate the ability of each ranking algorithm to predict the outcome of unseen LLM matches. We first partition each dataset into a 75/25 train/test split, running each ranking algorithm on
+
+<table><tr><td>Algorithm</td><td>Arena (F1)</td><td>SLAM (F1)</td></tr><tr><td>Elo</td><td>0.90</td><td>0.87</td></tr><tr><td>Markov</td><td>0.77</td><td>0.88</td></tr><tr><td>Glicko</td><td>0.88</td><td>0.88</td></tr><tr><td>Bradley-Terry</td><td>0.82</td><td>0.87</td></tr></table>
+
+Table 2: Ranking algorithm performance in prediction accuracy (F1 score) on Chatbot Arena & SLAM Dataset.
+
+the training set to determine the final ranking and strength of each model  $\pi$ . We then use each ranking algorithm to predict the outcome of the matches in the test set and calculate the F1 score as our performance metric.
+
+Performance on the Arena Dataset Due to its uneven distribution of model match-ups, the Arena data set presents an interesting challenge for ranking algorithms to predict future outcomes. When matches between models are unevenly distributed, some algorithms struggle with prediction accuracy. Our results show the Elo model to be the most accurate, achieving an F1 score of 0.90 compared to the F1 score of 0.88, 0.82, and 0.77 for the Glicko, Bradley-Terry and Markov Chain models, respectively (Table 2).
+
+The impact of the match-up distribution on the prediction accuracy is shown in the performance of the Markov chain which is affected by the sparsity of certain model match-ups, leading to lower prediction accuracy, as shown by the poor performance on models such as deepseek-llm-67b-chat and dolphin-2.2.1-mistral-7b in figure 4 of the appendix.
+
+Another example of the impact of matchup distribution is shown in the performance of the Bradley-Terry model in the appendix figure 4. We observe that the Bradley-Terry system suffers when dealing with "powerful models," where there is a strong imbalance between a model's wins and losses. For example, in the arena dataset, gpt-4-turbo had a win/loss ratio of 12288/3979. This skew in model wins/loss results in overestimation of the model's strength as shown by the low F1 score of 0.82 achieved by gpt-4-turbo using Bradley-Terry compared to an F1 score of 0.95, 0.96 and 0.92 from Elo, Glicko, and Markov Chain, respectively. This is known as the 'rare events' problem in logistic regression (King and Zeng, 2001). One method of countering this is to use a weighted logistic regression (Chiang et al., 2024). However, we observe that this has a negligible impact on the final result.
+
+<table><tr><td>Method</td><td>Elo</td><td>Markov</td><td>Glicko</td><td>Bradley-Terry</td><td>Win-Rate</td></tr><tr><td>Elo Rating</td><td>1.00</td><td>0.89</td><td>0.93</td><td>0.95</td><td>0.91</td></tr><tr><td>Markov</td><td>-</td><td>1.00</td><td>0.99</td><td>0.97</td><td>0.98</td></tr><tr><td>Glicko</td><td>-</td><td>-</td><td>1.00</td><td>0.99</td><td>0.99</td></tr><tr><td>Bradley-Terry</td><td>-</td><td>-</td><td>-</td><td>1.00</td><td>0.98</td></tr><tr><td>Win-Rate</td><td>-</td><td>-</td><td>-</td><td>-</td><td>1.00</td></tr></table>
+
+Table 3: Spearman's correlation between ranking algorithms on the SLAM dataset  
+
+<table><tr><td>Method</td><td>Elo</td><td>Markov</td><td>Glicko</td><td>Bradley-Terry</td><td>Win-Rate</td></tr><tr><td>Elo Rating</td><td>1.00</td><td>0.74</td><td>0.86</td><td>0.94</td><td>0.93</td></tr><tr><td>Markov</td><td>-</td><td>1.00</td><td>0.76</td><td>0.81</td><td>0.78</td></tr><tr><td>Glicko</td><td>-</td><td>-</td><td>1.00</td><td>0.81</td><td>0.89</td></tr><tr><td>Bradley-Terry</td><td>-</td><td>-</td><td>-</td><td>1.00</td><td>0.91</td></tr><tr><td>Win-Rate</td><td>-</td><td>-</td><td>-</td><td>-</td><td>1.00</td></tr></table>
+
+Table 4: Spearman's correlation between ranking algorithms on the Arena dataset
+
+Furthermore, the bootstrap confidence interval used by Chatbot Arena shows a narrow range of [-2.2, 2.3], suggesting a high confidence in the model ranking. However, this interval does not guarantee how well the model will perform against unseen opponents.
+
+Prediction accuracy on the SLAM dataset The distribution of matches between models in the SLAM dataset is tightly controlled so that every reviewer evaluates an equal number of models, resulting in a near uniform distribution of matches. We observe near-identical performance for all models. This result is due to the balanced distribution of match-ups in the dataset, where if a model has the most wins in the competition, its performance is directly reflected in its win rate. We further validate this occurrence by ranking all the models in SLAM by win rate and calculating the correlation between each pair of ranking algorithms, as shown in Tables 3 and 4. We observe a high correlation between all methods and the win rate.
+
+All algorithms perform similarly on the SLAM dataset, but on Arena, Elo leads in accuracy, followed by Glicko. Bradley-Terry, however, is negatively affected by the uneven distribution of the Arena dataset.
+
+# 4.4 Hyperparameter Sensitivity
+
+To measure the sensitivity of the hyperparameters, we calculate the F1 score of each ranking system as described in Section 4.3, varying each hyperparameter over a range of 100 values. Figure 2 showcases the range of results produced by Elo, Markov and Glicko across the two evaluation scenarios.
+
+Elo k-factor The  $k$  factor in Elo dictates the extent to which the score of a player is influenced by
+
+the outcome of an individual match. For example, with large  $k$  values, highly ranked players can incur a significant penalty when losing to a player with a lower ranking, while defeating such a player only results in a slight increase in score. The evaluation of the performance of the model (depicted in Figure 2) at varying levels of hyperparameter  $k$  shows marked sensitivity, which emphasizes the need for meticulous tuning. Elo shows the largest variation across both datasets, and is less stable, particularly for the smaller SLAM dataset. Through our tuning of this parameter, we determined that the ideal  $k$  factor for Elo is less than the standard 32 commonly used in previous research (Dettmers et al., 2024; Wu and Aji, 2023; Chiang et al., 2023). This reduced convergence rate can be attributed to the complexities of human preferences as opposed to the typical assessment of real-world games, which rely on the differing abilities of opposing teams.
+
+Markov Chain:  $p$  Score The  $p$  score in the Markov Chain determines the extent to which the overall ranking is influenced by each model's win rate, with values closer to 1 indicating greater influence compared to those closer to 0.5. We observe that Markov shows stable performance on the controlled SLAM datasets, performing as well as Glicko (Figure 2), however, the algorithm struggles on the large ARENA dataset due to its over reliance on win-rates for calculating ranks. When optimizing the value of this parameter, we find that the optimal  $p$  value for the Markov chain on the SLAM dataset is larger than the Arena dataset, indicating that the model's strength is greatly influenced by its overall win rate.
+
+Glicko: Rating deviation For Glicko, the ranking deviation  $(\sigma)$  is a measure of consistency or variability in a player's performance over time and determines how flexible a player's rating can be. If  $\sigma$  is initially set too high, the system will interpret the early game results as more variable or uncertain. This leads to larger rating adjustments, making the player's rating fluctuate more. In contrast, setting  $\sigma$  too low initially restricts the system from adjusting the rating freely. As a result, players, especially newcomers, may take much longer to reach a rating that accurately reflects their true skill level. This is particularly impactful in fast-changing or skill-evolving games, where a low initial variability setting could lead to underestimation of skill. We observe that Glicko's prediction performance is consistent across both datasets, showing that its
+
+![](images/daabde9e0e9b9ef580185a7fdaefef4ee740d90be549cd662227a722d2d044c7.jpg)
+
+![](images/f22673f7fbdf4c51906718e5c1a4c03bf0ec298d9372b53c41b525114dd91a54.jpg)  
+Figure 2: Distribution of F1 scores for the SLAM (top) and ARENA (bottom) datasets, showing the performance of Elo, Markov, and Glicko algorithms across a subset of models with 100 different hyperparameter settings. The results highlight the volatility of Elo, demonstrating its high sensitivity to hyperparameter changes compared to the more stable performance of Markov and Glicko.
+
+performance is less impacted by changes in its hyperparameters compared to Elo and Markov Chain. We theorize that this is due to the large number of games played for each model in both scenarios, allowing the system to adjust the rating deviation dynamically.
+
+Elo shows high sensitivity to hyperparameter whereas Glicko demonstrates greater stability and accuracy which suggests that Glicko is a more reliable alternative for LLM evaluation.
+
+# 5 Recommended Best Practices
+
+In sections 3 and 4, we outlined the desired properties of ranking systems and presented experiments assessing the performance of these systems in achieving these criteria. Based on these observations, this section puts these results into perspective and provides practical recommendations for select
+
+ing the most appropriate ranking algorithm. We outline which algorithms perform best under specific circumstances, offering guidance on their optimal use depending on the structure of the dataset, the matchup distribution, and the model evaluation goals. These recommendations, summarized in Table 5, aim to improve the reliability and applicability of classification systems in real-world LLM evaluations.
+
+# 5.1 Challenges with Elo
+
+An important result in this paper is that Elo, even with  $>1,000$  permutations, struggles to achieve stable rankings A.5. This indicates that Elo's dependency on permutations for ranking stabilization is inadequate, challenging previous studies (Boubdir et al., 2023). Combined with the high sensitivity to hyperparameters and the average performance in preserving transitivity. We do not
+
+<table><tr><td>Criteria</td><td>Elo</td><td>Markov</td><td>Glicko</td><td>Bradley-Terry</td></tr><tr><td>Transitivity</td><td>Moderate performance on smaller datasets, better than average performance on larger datasets.</td><td>Preserves transitivity but sensitive to sparse and uneven matches.</td><td>Good transitivity preservation in both types of datasets.</td><td>Generally good at maintaining transitivity, in smaller controlled and larger uncontrolled datasets.</td></tr><tr><td>Hyperparameter Sensitivity</td><td>Highly sensitive to K-factor and permutation count; performance fluctuates with changes.</td><td>Sensitive to the p-value, with rankings varying significantly based on its setting.</td><td>Incorporates rating deviation, making it less sensitive than Elo, but hyperparameter tuning is still necessary.</td><td>Does not use any hyperparameters.</td></tr><tr><td>Prediction Accuracy</td><td>Moderate prediction accuracy but inconsistent with small sample sizes or highly uneven matches.</td><td>Moderate prediction accuracy but highly dependent on matchup distribution.</td><td>High prediction accuracy due to consideration of the uncertainty in the ratings (rating deviation).</td><td>Higher prediction accuracy in controlled datasets but suffers in uneven matches.</td></tr><tr><td>Recommended Use Cases</td><td>Not recommended for small, unevenly distributed datasets; may perform better in larger datasets with consistent matches.</td><td>Applicable where dataset size is small and even, but sensitive to distribution; effective if properly tuned in moderately structured environments.</td><td>Best suited for large, unevenly distributed datasets. Able to manage uncertainty and prevent low matchup models from being ranked too high.</td><td>Ideal for small, unevenly distributed datasets. Also effective for large, evenly distributed datasets where interpretability and scalability are valued.</td></tr></table>
+
+Table 5: Consideration for Ranking Algorithm Selection based on Desired Ranking Properties
+
+# recommend ELO as an algorithm to rank LLM performance.
+
+Best Practice 1. Elo's high sensitivity to changes in its  $K$  factor and instability (even with large permutations) make it unsuitable for efficient ranking of LLM performance in pairwise evaluation (Sections 4.4 and A.5).
+
+# 5.2 Choose Bradley-Terry for Small Controlled Datasets
+
+In section 4.3, we demonstrate that ranking algorithms exhibit varied performance based on battle conditions. When matches are evenly distributed, the ranking algorithms produce similar results. However, a heavily skewed distribution can negatively impact some methods such as Elo and Bradley-Terry. In this scenario, where matchups are balanced and the number of matches is small, objective findings show that Bradley-Terry is best due to its ability to preserve transitivity while also maintaining comparable prediction accuracy.
+
+Best Practice 2. Bradley-Terry's simplicity and lack of hyperparameters make it ideal for controlled evaluation settings, although its reliance on sufficient matchup data limits its effectiveness in unevenly distributed scenarios.
+
+# 5.3 Choose Glicko for large uneven datasets
+
+For larger datasets, especially those with uneven matchup distributions, the Bradley-Terry model encounters difficulties due to the rare event problem. This problem occurs when new models with high win rates, participate in fewerATCHups than well-established models. As a result, these new models can be ranked disproportionately higher based on limited data, leading to biased rankings. In this scenario, the Glicko system provides a superior solution. Glicko utilizes a rating deviation parameter to account for uncertainty in a model's ranking due to limitedATCHups. This allows the system to adjust the rankings more conservatively for models with
+
+fewer comparisons, preventing them from being ranked too high based only on a few matches. As data sets become larger and more uneven, Glicko's ability to incorporate rating deviation ensures that the rankings accurately reflect the true performance throughout the dataset, even when new models appear with limited match histories. This flexibility makes Glicko particularly effective for large-scale LLM assessments where uneven matchup distributions are prevalent.
+
+Best Practice 3. Glicko is effective for large data sets with uneven matches due to its rating deviation parameter, which manages uncertainty and prevents models with few matches from being overrated.
+
+# 5.4 Bradley-Terry for other Dataset Types
+
+For small, uneven datasets, Bradley-Terry's simplicity and ability to model pairwise comparisons make it a robust choice, even when the data distribution is skewed. It does not require parameter estimation, which allows it to handle small datasets efficiently without the need for complex adjustments or hyperparameter tuning, making it ideal for situations where uneven distributions might otherwise require additional preprocessing.
+
+In the case of large, evenly distributed datasets, Bradley-Terry is effective because of its scalability and ease of interpretation. For large data sets where the distribution of results is balanced, Bradley-Terry's ability to produce clear, interpretable rankings without adding excessive computational overhead is a significant advantage. Although more complex models such as Glicko can offer finer granularity in large, uneven data cases, Bradley-Terry's efficiency and interpretability make it a viable alternative when computational simplicity and transparency are prioritized.
+
+Best Practice 4. Bradley-Terry's scalability and interpretability make it a viable choice for large, evenly distributed datasets, offering clear rankings with minimal computational overhead.
+
+# 6 Related Work
+
+Pairwise ranking has recently emerged as a popular method for evaluating LLMs, as it allows for direct comparisons between models and provides a more nuanced assessment of their performance than traditional metrics (Chiang et al., 2024). This evaluation involves presenting human judges with pairs of model outputs and asking them to select the better one. The outcomes of these pairwise comparisons are then aggregated into final rankings using ranking algorithms such as Elo (Elo, 1978). Despite their widespread use, there has been limited research on the applicability of these ranking algorithms to LLM evaluation. Works such as (Boubdir et al., 2023) have critically analyzed the compatibility of Elo with LLM evaluations by examining how well it adheres to reliability and transitivity. However, Elo is not the only ranking algorithm used in LLM evaluation. There exists a number of other algorithms that have been proposed to rank in different contexts, such as the Bradley-Terry model (Bradley and Terry, 1952) and Glicko (Glickman and Jones, 1999), which have been shown to outperform Elo in certain scenarios, but their performance in the context of LLM evaluation has not been systematically evaluated. However, several factors affect the performance of these algorithms, such as the distribution of the data, the number of models that are compared, and the number of pairwise comparisons. As such, there is a need for a systematic comparison of ranking algorithms in the context of LLM evaluation to identify the most suitable algorithm for different scenarios which this paper aims to address.
+
+# 7 Conclusion
+
+In this paper, we explore the application of several ranking algorithms in the context of evaluating LLMs. Through our experiments and analysis, we identify significant limitations in the ability of these algorithms to maintain stable and accurate ranking across varying dataset types and distribution patterns. To combat these challenges, we introduce a set of guidelines for selecting the most appropriate ranking system based on the specific characteristics of the evaluation task and available resources. We
+
+believe that our findings and recommendations will be valuable to researchers and practitioners working in the field of LLM evaluation, and we hope that our work will inspire further research in this area.
+
+# 8 Limitations
+
+While our study comprehensively explores the ranking of Large Language Models (LLMs) through head-to-head combat, several limitations warrant discussion.
+
+**Scalability Constraints** One significant limitation arises from the scalability constraints inherent in pairwise evaluations. As the number of LLMs in the ecosystem expands, the number of required comparisons grows quadratically. This introduces computational challenges and resource constraints that may limit the feasibility of exhaustive head-to-head evaluations, particularly for larger sets of models.
+
+Human Feedback Variability Another concern is the variability in human feedback, on which our rankings are heavily based. Human judgments can be subjective and influenced by numerous factors, including the individual's background, expertise, and contextual understanding. This variability can introduce noise into the ranking system, affecting the evaluations' stability and reliability.
+
+# References
+
+Meriem Boubdir, Edward Kim, Beyza Ermis, Sara Hooker, and Marzieh Fadaee. 2023. Elo uncovered: Robustness and best practices in language model evaluation. arXiv preprint arXiv:2311.17295.  
+Ralph Allan Bradley and Milton E Terry. 1952. Rank analysis of incomplete block designs: I. the method of paired comparisons. Biometrika, 39(3/4):324-345.  
+Thomas Callaghan, Peter J Mucha, and Mason A Porter. 2003. Random walker ranking for ncaa division i-a football. arXiv.  
+Thomas Callaghan, Peter J Mucha, and Mason A Porter. 2004. The bowl championship series: A mathematical review. arXiv.  
+Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E. Gonzalez, Ion Stoica, and Eric P. Xing. 2023. Vicuna: An open-source chatbot impressing gpt-4 with  $90\%$  * chatgpt quality.
+
+Wei-Lin Chiang, Lianmin Zheng, Ying Sheng, Anastasios Nikolas Angelopoulos, Tianle Li, Dacheng Li, Hao Zhang, Banghua Zhu, Michael Jordan, Joseph E. Gonzalez, and Ion Stoica. 2024. Chatbot arena: An open platform for evaluating llms by human preference. Preprint, arXiv:2403.04132.  
+H. A. David. 1988. The Method of Paired Comparisons, 2nd edition. Griffin's Statistical Monographs & Courses, Oxford University Press, London: New York. Revised edition, Griffin's statistical monographs & courses; no. 41.  
+Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. 2024. Qlora: Efficient finetuning of quantized lms. Advances in Neural Information Processing Systems, 36.  
+Yann Dubois, Chen Xuechen Li, Rohan Taori, Tianyi Zhang, Ishaan Gulrajani, Jimmy Ba, Carlos Guestrin, Percy S Liang, and Tatsunori B Hashimoto. 2024. Alpacafarm: A simulation framework for methods that learn from human feedback. Advances in Neural Information Processing Systems, 36.  
+A.E. Elo. 1978. The Rating of Chessplayers, Past and Present. Arco Pub.  
+Leo Gao, Jonathan Tow, Baber Abbasi, Stella Biderman, Sid Black, Anthony DiPofi, Charles Foster, Laurence Golding, Jeffrey Hsu, Alain Le Noac'h, Haonan Li, Kyle McDonell, Niklas Muennighoff, Chris Ociepa, Jason Phang, Laria Reynolds, Hailey Schoelkopf, Aviya Skowron, Lintang Sutawika, Eric Tang, Anish Thite, Ben Wang, Kevin Wang, and Andy Zou. 2023. A framework for few-shot language model evaluation.  
+Mark E Glickman. 1999. Parameter estimation in large dynamic paired comparison experiments. Journal of the Royal Statistical Society Series C: Applied Statistics, 48(3):377-394.  
+Mark E Glickman and Albyn C Jones. 1999. Rating the chess rating system. CHANCE-BERLIN THEN NEW YORK-, 12:21-28.  
+Zishan Guo, Renren Jin, Chuang Liu, Yufei Huang, Dan Shi, Linhao Yu, Yan Liu, Jiaxuan Li, Bojian Xiong, Deyi Xiong, et al. 2023. Evaluating large language models: A comprehensive survey. arXiv preprint arXiv:2310.19736.  
+Chandra Irugalbandara, Ashish Mahendra, Roland Daynauth, Tharuka Kasthuri Arachchige, Krisztian Flautner, Lingjia Tang, Yiping Kang, and Jason Mars. 2023. Scaling down to scale up: A cost-benefit analysis of replacing openai's gpt-4 with self-hosted open source slms in production. arXiv preprint arXiv:2312.14972.  
+Maurice G Kendall and B Babington Smith. 1940. On the method of paired comparisons. Biometrika, 31(3/4):324-345.  
+Gary King and Langche Zeng. 2001. Logistic regression in rare events data. Political analysis, 9(2):137-163.
+
+Paul Kvam and Joel S Sokol. 2006. A logistic regression/markov chain model for ncaa basketball. *Naval Research Logistics (NrL)*, 53(8):788-803.  
+Ricky Sanjaya, Jun Wang, and Yaodong Yang. 2022. Measuring the Non-Transitivity in Chess. Algorithms, 15(5):152.  
+Alex Wang, Yada Pruksachatkun, Nikita Nangia, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R. Bowman. 2020. Superglue: A stickier benchmark for general-purpose language understanding systems. Preprint, arXiv:1905.00537.  
+Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R. Bowman. 2019. Glue: A multi-task benchmark and analysis platform for natural language understanding. Preprint, arXiv:1804.07461.  
+Minghao Wu and Alham Fikri Aji. 2023. Style over substance: Evaluation biases for large language models. arXiv.  
+Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, et al. 2024. Judging llm-as-a-judge with mt-bench and chatbot arena. Advances in Neural Information Processing Systems, 36.
+
+# A Appendix
+
+# A.1 Glicko Algorithm Details
+
+For model  $i \in M$ , the rating and the rating deviation are updated as:
+
+$$
+\theta_ {i} ^ {\prime} = \theta_ {i} + \frac {q}{\frac {1}{\sigma_ {i} ^ {2}} + \frac {1}{d ^ {2}}} \sum_ {j = 1} ^ {J} g \left(\sigma_ {j}\right) \left(S _ {i j} - p _ {i j}\right) \tag {6}
+$$
+
+$$
+\sigma_ {i} ^ {\prime} = \left(\frac {1}{\sigma_ {i} ^ {2}} + \frac {1}{d ^ {2}}\right) ^ {- 1} \tag {7}
+$$
+
+where  $q = \log (10) / 400$ ,  $J$  is the set of opponents for  $i$  and  $g(\sigma) = 1 / \sqrt{1 + 3q^2\sigma^2 / \pi^2}$ .
+
+Given the rating  $\theta_{i}$  and  $\theta_{j}$  of model  $i$  and  $j$ , the expected outcome  $p_{ij}$  is calculated as:
+
+$$
+p _ {i j} = \frac {1}{1 + 1 0 ^ {- g \left(\sigma_ {j}\right) \left(\theta_ {i} - \theta_ {j}\right) / 4 0 0}} \tag {8}
+$$
+
+$$
+d ^ {2} = \left(q ^ {2} \sum_ {j = 1} ^ {J} g \left(\sigma_ {j}\right) ^ {2} p _ {i j} \left(1 - p _ {i j}\right)\right) ^ {- 1} \tag {9}
+$$
+
+where  $d^2$  is the scaling factor for the system's certainty.
+
+# A.2 Markov Chain Details
+
+The Markov Chain Model was originally developed as a random walker model Callaghan et al. (2003, 2004) where ranks were calculated by solving a series of differential equations. This method was adapted into a Markov Chain Model by (Kvam and Sokol, 2006). When expressed as a Markov chain, the elements  $T_{ij}$  of the transition matrix  $T$  are the probabilities that a walker moves from model  $i$  to  $j$  in a single time step, with the initial values computed as follows:
+
+$$
+t _ {i j} = \frac {1}{N _ {i}} \left[ w _ {i j} (1 - p) + l _ {i j} p \right], \quad \forall j \neq i \tag {10}
+$$
+
+$$
+t _ {i j} = \frac {1}{N _ {i}} \left[ W _ {i} p + L _ {i} (1 - p) \right] \tag {11}
+$$
+
+where  $W_{i}, L_{i}$  are the number of matches won by model  $i$  and  $w_{ij}, l_{ij}$  are the number of games won by model  $i$  against model  $j$ .
+
+The steady-state probability vector  $\pi = [\theta_i]$  represents the final ranking of each team  $i\in M$  such that  $\pi T = \pi$  and  $\sum_{i = 1}^{m}\theta_{i} = 1$ .
+
+# A.3 F1 scores per model
+
+In this paper, F1 scores are used as an evaluation metric to assess the consistency and predictive performance of the ranking algorithms. The F1 score is calculated based on the precision and recall of the correctly predicted pairwise outcomes compared to the observed outcomes in the dataset. In the following, we detail the methodology for computing the F1 scores:
+
+Let the data set consist of  $m$  pairwise matches between models. For each pair  $(i,j)$ ,  $p_{ij}$  is the probability that the model  $i$  beats  $j$  in a random match. If model  $i$  plays  $M_i$  matches then  $E_i$ , the expected number of games won by model  $i$  is  $E_i = \lfloor M_i p_{ij} \rfloor$  and the actual number of games won is given by  $A_i$ .
+
+The Precision and Recall are calculated as:
+
+$$
+\text {p r e c i s i o n} = \frac {\left| E _ {i} \cap A _ {i} \right|}{\left| E _ {i} \right|}, \quad \text {r e c a l l} = \frac {\left| E _ {i} \cap A _ {i} \right|}{\left| A _ {i} \right|}
+$$
+
+The F1 score is calculated as the harmonic mean of precision and recall:
+
+$$
+F 1 \text {S c o r e} = 2 \cdot \frac {\text {P r e c i s i o n} \cdot \text {R e c a l l}}{\text {P r e c i s i o n} + \text {R e c a l l}}.
+$$
+
+The total F1 score for a ranking algorithm is computed by averaging the F1 scores for all pairwise matches  $(i,j)$ :
+
+$$
+F 1 \operatorname {S c o r e} _ {\text {O v e r a l l}} = \frac {1}{| M |} \sum_ {(i, j) \in M} F 1 \operatorname {S c o r e} _ {(i, j)},
+$$
+
+where  $M$  is the set of all pairwise matches.
+
+Figures 3 and 4 show the F1 scores for both the SLAM and Arena datasets, and also their overall F1 score.
+
+![](images/432adeb32d185a28c5edcf76589acad0797f49e33eb6c0ec8bf925b8e80f4d4c.jpg)  
+Figure 3: F1 scores for all models in the SLAM dataset.
+
+![](images/003be54b51f16129a40501a4e09409f21553bb345688785731f177608c2e075c.jpg)  
+Figure 4: F1 scores of the top ten (10) highest and lowest ranked models using the Elo rating system on the Chatbot Arena dataset. In general, Elo provides the best prediction accuracy by achieving an F1 score of 0.90.
+
+# A.3.1 Experimental Setup
+
+To evaluate the predictive accuracy of the ranking algorithms, each data set was split into a  $75\%$  training set and a  $25\%$  test set.
+
+# A.3.2 Procedure
+
+The ranking algorithms were applied to the training set to compute model rankings based on pairwise matchups. Using these rankings, the algorithms predicted the outcomes of matchups in the test set, expressed as the number of games each model was expected to win. These predictions were compared with the actual results of the match and the precision, recall and F1 scores were calculated for each match.
+
+The F1 scores, aggregated across all matches in the test set, quantified the alignment between the predicted and observed results, providing a measure of the ability of each algorithm to generalize to unseen data.
+
+# A.4 Elo's rank at  $k = 3$  vs  $k = 5$
+
+To analyze the sensitivity of Elo rankings to hyperparameter settings, we conducted experiments using two versions of Elo, with  $K = 3$  and  $K = 5$ , on the SLAM dataset.
+
+# A.4.1 Experimental Setup
+
+For each  $K$ -factor, we evaluated Elo rankings under varying numbers of permutations: 1, 10, 100, and 1000. Permutations involve randomizing the order of matchups, a strategy commonly employed to reduce the effect of matchup order on final rankings and improve stability.
+
+# A.4.2 Procedure
+
+For each  $K$  factor ( $K = 3$  and  $K = 5$ ), Elo rankings were calculated using the raw SLAM matchup data without permutations. Matchup orders were permuted 1, 10, 100, and 1000 times for each factor  $K$ , with Elo rankings recalculated after each set of permutations. Our findings (shown in figure 5), show volatility of Elo's ranking for each of the  $k$  value.
+
+![](images/e05a83df3175e7a38870acf0fb4670613e38b0131f08afe555fb5f7e14999473.jpg)  
+(a) Ranks produced by Elo for  $k = 3$
+
+![](images/cd10fe1126debcc754506fd61353b931ecb0e70381a0cb35817cb506fc244f88.jpg)  
+(b) Ranks produced by Elo for  $k = 5$  
+Figure 5: Elo produces different ranks based on the value of the hyperparameter  $k$ . Increasing the number of permutation can lead to more stable ratings, however, model ranks may still be unstable as is the case with orca-min_3b and neural-chat_7b.
+
+# A.5 Elo's Instability at higher permutation
+
+To analyze the performance of Elo at higher permutations, we conducted two separate experiments on the SLAM dataset.
+
+# A.5.1 Experimental Setup
+
+For a single  $K$  factor of 25, we evaluated Elo rankings by performing two separate evaluations of each 10,000 permutation using the strategy of the previous section.
+
+# A.5.2 Procedure
+
+We observed the ranks produced by each experiment, showing that even with a large permutation, ELO can produce two different ranks for the same two models6.
+
+# A.6 Evaluating the consistency of crowd-workers evaluators
+
+In this section, we recommend a method to evaluate the consistency of crowd-worker evaluations based on the Analytic hierarchy process (AHP) practices. We begin by formally defining cycles in a pairwise ranking dataset.
+
+Definition A.1. For models  $i_1, i_2, \dots, i_k \in M$ , if  $p_{i_1i_2} > 0.5, p_{i_2i_3} > 0.5, \dots, p_{iki_1} > 0.5$ , then their games is said to be cyclic. A cycle of length three is called a Rock, Paper, Scissors (RPS) cycle.
+
+![](images/751550e26a3bade4b5d8a84b3699ea37ee3ca2e3f5a8381cd247bd4ae2fa9261.jpg)  
+Figure 6: Two separate ranks produced by ELO for permutations of 10,000 each.
+
+For any three models  $i, j$ , and  $k$ , it is expected that if a reviewer prefers  $i$  over  $j$  and  $j$  over  $k$ , they prefer  $i$  over  $k$  (transitivity). However, this may not always be due to personal preference, evaluation error, or evaluator negligence. If the transitive property does not hold, we say that the three models are intransitive and the evaluation is inconsistent.
+
+Measuring the consistency of pairwise ranking falls under the domain of AHP. According to AHP theory, if the number of RPS cycles remains small, it reflects the complexity of the ranking problem rather than the judges' complacency. However, a high number of cycles represents a low quality of judgments.
+
+According to (Kendall and Smith, 1940), the expected number of RPS cycles for a pairwise comparison dataset of  $m$  items follows a  $\chi^2$  distribution, and if the actual number of RPS cycles is significantly less than the expected number, we can be reasonably confident that the dataset is consistent.
+
+The number of RPS cycles can be calculated by first generating the win rate matrix  $P$  as defined in section 4.2. This is followed by the adjacency matrix  $A$ , where  $a_{ij} = 1$  if  $p_{ij} > 0.5$  and 0 otherwise. Finally, the total number of RPS cycles is calculated as  $\rho = 1 / 3\sum_{i = 1}^{M}a_{i,i}^{*}$ , where  $a_{ii}^{*}$  are diagonals of  $A^3$  (Sanjaya et al., 2022).
+
+The  $\chi^2$  statistic is calculated as:
+
+$$
+\chi^ {2} = \frac {8}{k - 4} \left\{\frac {1}{4} \binom {k} {3} - d - \frac {1}{3} + \rho \right\} \tag {12}
+$$
+
+$$
+v = \frac {k (k - 1) (k - 2)}{(k - 4) ^ {2}} \tag {13}
+$$
+
+where  $v$  is the degrees of freedom.
+
+The RPS count and  $\chi^2$  statistic for both the SLAM and Arena datasets are shown in table 6. At a p-value close to 0, we can be confident that inconsistencies are not due to the quality of the judges.
+
+Table 6: Non-Transitivity statistic for the two datasets that were used. The p-value is calculated based on the  $\chi^2$  estimated via Kendal's Chi-squared approximation to the circular triad distribution.  
+
+<table><tr><td>Dataset</td><td>RPS Count</td><td>x²</td><td>p-value</td></tr><tr><td>SLAM</td><td>19</td><td>45.06</td><td>0.001</td></tr><tr><td>Arena</td><td>393</td><td>1107.25</td><td>0.000</td></tr></table>

rankingunraveledrecipesforllmrankingsinheadtoheadaicombat/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbff0393b6b73d2ae47148e6f17e82f3c9bb590f142c433b11b842b51584840c
+size 557307

rankingunraveledrecipesforllmrankingsinheadtoheadaicombat/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:763ac1031de286c9700ad28b5cbb856e5ea4a8d3299f3802441eb98a8eda31ce
+size 486858

rareretrievalaugmentedreasoningenhancementforlargelanguagemodels/b975a564-3470-43bc-aefc-a67598ac26f4_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e517b0318d517096b06a1ae5f848b3a10d236a5678b3072941b68106290cf1d
+size 140352

rareretrievalaugmentedreasoningenhancementforlargelanguagemodels/b975a564-3470-43bc-aefc-a67598ac26f4_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b38eddab2966d2af285cb4b06d18bf5075602689d35616e2d9a571b50984dcf7
+size 172583

rareretrievalaugmentedreasoningenhancementforlargelanguagemodels/b975a564-3470-43bc-aefc-a67598ac26f4_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3355d713659ec4ba0326761763f20b332036971fcaf88ae579fec301248b11a6
+size 1558486

rareretrievalaugmentedreasoningenhancementforlargelanguagemodels/full.md

@@ -0,0 +1,568 @@
+# RARE: Retrieval-Augmented Reasoning Enhancement for Large Language Models
+
+Hieu Tran<sup>1</sup>, Zonghai Yao<sup>1</sup>, Zhichao Yang<sup>1</sup>, Junda Wang<sup>1</sup>, Yifan Zhang<sup>2</sup>, Shuo Han<sup>2</sup>, Feiyun Ouyang<sup>2</sup>, Hong Yu<sup>1,2</sup>
+
+<sup>1</sup>University of Massachusetts, Amherst  
+<sup>2</sup>University of Massachusetts, Lowell
+
+# Abstract
+
+This work introduces RARE (Retrieval-Augmented Reasoning Enhancement), a versatile extension to the mutual reasoning framework (rStar), aimed at enhancing reasoning accuracy and factual integrity across large language models (LLMs) for complex, knowledge-intensive tasks such as medical and commonsense reasoning. RARE incorporates two innovative actions within the Monte Carlo Tree Search framework: (A6), which generates search queries based on the initial problem statement, performs information retrieval using those queries, and augments reasoning with the retrieved data to formulate the final answer; and (A7), which leverages information retrieval specifically for generated sub-questions and re-answers these sub-questions with the relevant contextual information. Additionally, a Retrieval-Augmented Factuality Scorer is proposed to replace the original discriminator, prioritizing reasoning paths that meet high standards of factuality. Experimental results with LLaMA 3.1 show that RARE enables open-source LLMs to achieve competitive performance with top closed-source models like GPT-4 and GPT-4o. This research establishes RARE as a scalable solution for improving LLMs in domains where logical coherence and factual integrity are critical<sup>1</sup>.
+
+# 1 Introduction
+
+Question answering (QA) is a cornerstone task in natural language processing that involves generating answers to questions posed in natural language. QA spans a broad spectrum of domains and types, ranging from open-domain QA (Yang et al., 2018; Kwiatkowski et al., 2019) to more specialized areas like medical QA (Jin et al., 2021; Cao et al., 2011). The overwhelming volume and complexity of medical information necessitate medical QA, which benefits many downstream tasks such as medical
+
+education, clinical decision support, and patient care optimization (Cai et al., 2023; Liu et al., 2023; Jin et al., 2024b).
+
+Medical QA represents a unique and demanding subset of QA, requiring models to navigate intricate medical knowledge, interpret clinical scenarios, and select correct and contextually appropriate options (Singhal et al., 2023b; Wu et al., 2024). Similar to general domain QA, Medical QA requires structured multi-step reasoning, where answers emerge from various sequential steps. Take Figure 1 as an example, to find appropriate treatment given patient information, the QA model should first identify patient conditions (colored in red, e.g., chief complaint and past conditions), then analyze contributing factors and diagnose the disease (colored in blue), and determine appropriate evidence-based interventions in the final step (colored in yellow). Without such structured multi-step reasoning, it would be challenging to arrive at an accurate and contextually relevant answer for such a complex medical question.
+
+Moreover, Medical QA presents two non-trivial challenges that distinguish it from general-domain QA. First, Medical QA depends heavily on domain-specific knowledge that is not always available within pre-trained models, necessitating knowledge-based retrieval from external sources (Xiong et al., 2024a). Figure 1 is an example which involves specific medical terms such as allergic conjunctivitis. In addition, medical knowledge evolves rapidly, and new treatments or updated guidelines may not be included in the model's pretrained corpus. For example, newer drugs (like epinastine hydrochloride for allergic conjunctivitis) may be recommended by recent guidelines but absent in older pre-trained models. Second, Medical QA encompasses a wide variety of question types, including not only multi-step reasoning and knowledge-based retrieval as previously mentioned, but also composite questions requiring iterative evi
+
+![](images/99f0b6c37675be0f0c0757ba5a2ab3b2e038de5ba0151802414118f98e3a68b7.jpg)  
+Figure 1: Overview of our reasoning process, which combines generation and factuality scoring. (1) A retrieval-augmented generator produces multiple candidate reasoning paths using Monte Carlo Tree Search (MCTS); (2) a retrieval-augmented factuality scorer evaluates the factual accuracy of each reasoning path; (3) the trajectory with the highest factuality score is selected as the final answer.
+
+dence retrieval, where they demand retrieval of relevant knowledge at each reasoning step to ensure accuracy and relevance throughout the process.
+
+In parallel, Commonsense Question Answering shares similar complexities with Medical QA, particularly in its reliance on structured multi-step reasoning and iterative evidence retrieval. While Medical QA draws heavily on domain-specific knowledge, Commonsense QA focuses on leveraging a model's understanding of general world knowledge and logical connections to answer questions that are often indirect or abstract. For example, tasks like StrategyQA (Geva et al., 2021) require models to infer hidden relationships and execute multi-hop reasoning, akin to diagnosing a condition in Medical QA (Trivedi et al., 2023; Bauer et al., 2018; Chen et al., 2020). This similarity in reasoning processes across both domains underscores the importance of designing frameworks that can adapt to and optimize multi-step reasoning workflows, irrespective of the domain.
+
+In this paper, we propose Retrieval-Augmented Reasoning Enhancement (RARE) to handle aforementioned challenges. We built upon rStar (Qi et al., 2024) where a language model generates reasoning steps and another verifies them, improving accuracy without fine-tuning or superior models. To answer knowledge-based questions, RARE designed a new action A6, which generates multiple search queries based on the question and retrieves relevant documents. To answer composite
+
+questions, we add action A7, which refines subquestions, retrieves targeted information, and updates the next step. RARE applies the Monte Carlo Tree Search (MCTS) algorithm to select the best action path that leads to the final answer. In addition, RARE is complemented by Retrieval-Augmented Factuality Scorer (RAFS) that evaluates and ranks reasoning paths for factual accuracy.
+
+We applied RARE and other baselines on 3 medical QA tasks and 3 general domain QA tasks. Results show that RARE significantly enhances accuracy across various LLMs, enabling the open-source LLMs (LLaMA 3) to achieve competitive performance with top closed-source LLMs like GPT-4o. Our contributions are as follows:
+
+1. Formulating Medical QA as Multi-Step Reasoning: We build upon the rStar framework to model medical QA as a structured multi-step reasoning task, addressing the complexity and sequential nature of medical queries.  
+2. Novel Retrieval Actions: We introduce two retrieval-augmented actions within the MCTS framework, enabling the integration of real-time, context-specific information to enhance reasoning accuracy and relevance.  
+3. Retrieval-Augmented Factuality Scorer: We propose a Retrieval-Augmented Factuality Scorer to evaluate and rank reasoning paths,
+
+# Actions
+
+![](images/dfde9208ff2e3bb2cefd50cd1b2356bbab50bbd6874622937e395be0659738ca.jpg)
+
+Propose a one-step thought
+
+![](images/c360cb7189ffbe8e5edaa574097d6d357c1e6642f9a5a3388ed2be80a7110305.jpg)
+
+Complete remaining thought
+
+![](images/c1029b57dbce03118420e6d45fdc107e978a72fe86006d6aba4bf70491513dfb.jpg)
+
+Propose a subquestion and
+
+answer
+
+![](images/69d74e7de4e7b1d522e1a0743789b010e062aa3747e0a7270ed56fe6e4553019.jpg)
+
+Reanswer the subquestion
+
+Rephrase the question
+
+![](images/5e62f3666fd5393fc5d826fb45b9d480f7ba45ae6026595d932b0eca616f03d7.jpg)
+
+Search Query Generation
+
+![](images/3107b7332b393211d9eae89e08cd18d0cece59cb61b1eb6bbfa0acbced33a30f.jpg)
+
+2.1.1.1.1
+
+Sub-question Retrieval and Re-answering
+
+...Therefore, warm compresses would be the most appropriate treatment for this patient. The answer is C: Warm compresses.
+
+![](images/5c9996a7847dc5f977beb246407619e7c053b093c1d2656090456025ff3681bc.jpg)  
+Figure 2: An example to illustrate the process of retrieval-augmeted generator. Highlighted nodes from top to bottom constitute a complete reasoning trace. Given a question, MCTS augments the LLM to explore a rich, human-like reasoning action space and generate the next steps based on the current state.
+
+ensuring they maintain both logical coherence and factual reliability throughout the reasoning process.
+
+# 2 Preliminaries
+
+This section introduces the foundational concepts and notations used in this work, focusing on the Monte Carlo Tree Search (MCTS) algorithm and the rStar framework (Qi et al., 2024), which serve as the basis for our proposed RARE method.
+
+# 2.1 Monte Carlo Tree Search (MCTS)
+
+Monte Carlo Tree Search (MCTS) is a decision-making algorithm that incrementally builds a search tree by simulating outcomes to estimate action values, making it effective for complex reasoning tasks (Browne et al., 2012). It operates through iterative selection, expansion, simulation, and backpropagation phases, balancing exploration and exploitation via the Upper Confidence Bound applied on Trees (UCT). MCTS enables adaptive strategy refinement in large search spaces where direct computation is infeasible. A detailed explanation of MCTS and its implementation can be found in the Appendix A.5.
+
+# 2.2 Mutual Reasoning Makes Smaller LLMs Stronger Problem-Solvers
+
+Building upon MCTS, (Qi et al., 2024) proposed rStar, a framework that augments MCTS with a diverse set of reasoning actions. This enhancement is designed to improve exploration of the solution space in complex reasoning tasks by allowing more dynamic and human-like reasoning
+
+pathways. Traditional approaches, such as Chain of Thought (CoT) reasoning (Wei et al., 2022) or self-consistency (Wang et al., 2022), often rely on a single action type, which can limit the diversity and effectiveness of generated solutions. In contrast, rStar incorporates five distinct actions that enable more adaptive exploration(the prompt for each action can be found in Appendix A.7):
+
+- A1: Propose a One-Step Thought. This action generates the next reasoning step based on previous steps, allowing the LLM to build the solution incrementally.  
+- A2: Propose Remaining Thought Steps. This action enables the LLM to produce all remaining reasoning steps in one inference, similar to CoT, for simpler questions.  
+- A3: Generate Next Sub-question and Answer. This action decomposes the main problem into a sequence of sub-questions, each solved in turn.  
+- A4: Re-answer Sub-question. This action allows the LLM to re-answer a previously generated sub-question, increasing accuracy by using few-shot prompting.  
+- A5: Rephrase Question/Sub-question. This action rephrases the question to clarify conditions and reduce misunderstandings, enhancing the LLM's interpretation of the problem.
+
+# 3 Methodology
+
+# 3.1 Overview of RARE Framework
+
+Inspired by the generator-discriminator structure of rStar (Qi et al., 2024), RARE introduces a retrieval-
+
+![](images/2359545c8cc149b411bb7892b8cf0ca3ea21e6e179eb592ba5ee7844dbe1fb5d.jpg)  
+A6
+
+The patient had bladder infection. Which of the following conditions is also commonly associated with this double-stranded DNA virus infection? A: Gastroenteritis, B: Painful dermatomal vesicles, C:Lymphadenopathy, D: Vesicular rash
+
+![](images/f2ede844e483cde96a4e9e2ebcad9104c9a144c0cf15be6c42a0cb105110aaa8.jpg)  
+Question
+
+augmented generator and a factuality scorer to enhance reasoning accuracy and factual integrity in large language models. As illustrated in Figure 1, RARE operates in two main stages:
+
+Candidate Generation with Retrieval-Augmented Generator: The retrieval-augmented generator builds on the MCTS-based rStar self-generator, incorporating two new retrieval-augmented actions that dynamically fetch relevant external information. These actions improve the relevance and accuracy of candidate reasoning paths by integrating contextually enriched knowledge into intermediate reasoning steps, especially for complex questions, illustrated in Figure 2.
+
+Factuality Evaluation with Retrieval-Augmented Factuality Scorer : Replacing the discriminator in rStar, the Retrieval-Augmented Factuality Scorer evaluates each candidate trajectory's factual reliability. This scorer verifies the alignment of intermediate reasoning steps with retrieved evidence, assigning a factuality score that reflects the trajectory's consistency with external knowledge. The trajectory with the highest
+
+factuality score is selected as the final answer, prioritizing the most factually supported reasoning path. This selection ensures coherence and factual alignment, enhancing response reliability.
+
+Through these stages, RARE systematically integrates retrieval-based evidence into the reasoning process, optimizing both reasoning coherence and factual accuracy. This approach makes RARE well-suited for knowledge-intensive tasks, such as commonsense and medical reasoning.
+
+![](images/615deca73a0bc559de3cd5a515a92d094012e32305032fc405f8bc9567ef2310.jpg)  
+Figure 3: Illustration of the proposed retrieval-augmented action A6 in RARE: Given a question, LLMs generate search queries and retrieve relevant documents to construct a contextually enriched final answer. Key entities are underlined.  
+Figure 4: Illustration of the proposed retrieval-augmented action A7 in RARE: LLMs decompose the question into sub-questions, perform retrieval for each sub-question, and re-answer them based on the retrieved documents. The final sub-question is a rephrased version of the original question, so the sub-answer to this final sub-question also serves as the answer to the original question. In comparison with previous figure, we can find that A6 tends to use existing entity from the main question where A7 tends to use additional entity from previous subanswer.
+
+# 3.2 Retrieval-Augmented Generator
+
+Traditional retrieval-augmented generation methods often rely on a single retrieval step before generating responses, which limits their effectiveness in complex reasoning tasks. However, recent advances in medical RAG have demonstrated the importance of iterative retrieval in improving reasoning quality. i-MedRAG (Xiong et al., 2024b) highlights that multi-turn retrieval, where an LLM generates follow-up queries dynamically, signifi
+
+# Retrieval-Augmented Factuality Scorer (RAFS)
+
+![](images/2220703fc39887134fb89c36dfd2522506f0e6818c163ae3ce63a81b10c54b2a.jpg)  
+Figure 5: The RAFS assesses the factual accuracy of reasoning paths in four steps. (1) Split into sentences: The reasoning is divided into individual statements. (2) Generate retrieval queries: For each statement, an LLM generates multiple queries aimed at retrieving relevant information. (3) Retrieve information: The retrieval system gathers supporting information based on these queries. (4) Rate using retrieved information: Each statement is evaluated against the retrieved information and labeled as Supported or Not Supported. The final output includes a factuality score, calculated as the proportion of supported statements, which aids in selecting the most factually reliable reasoning path.
+
+cantly enhances response accuracy by refining its knowledge base. Inspired by this approach, we introduce two new retrieval-augmented actions into the original rStar self-generator (Qi et al., 2024), transforming it into a Retrieval-Augmented Generator, as illustrated in Figures 3 and 4. These retrieval-augmented actions enable the generator to dynamically incorporate external knowledge during intermediate reasoning steps, improving both the contextual relevance and factual accuracy of generated responses.
+
+- A6: Search Query Generation and Information Retrieval. In this step, the LLM generates targeted search queries from the original question and retrieves relevant information. This additional context is then used to enrich the reasoning path, enabling the model to deliver a more comprehensive and contextually grounded final answer.  
+A7: Sub-question Retrieval and Reanswering. Unlike Action A6, which centers
+
+on the main question, this action targets subquestions generated by Action A3. For each sub-question, the model fetches relevant information and re-answers accordingly. By refining these intermediary steps, A7 improves the reasoning chain's coherence and factual reliability, resulting in more accurate overall outcomes.
+
+By incorporating these retrieval-augmented actions, the generator can explore a wider range of possible solutions, leading to reasoning paths that are both logically coherent and enriched with external knowledge. This upgrade transforms the generator into a retrieval-augmented generator, enabling RARE to better handle complex, knowledge-intensive reasoning tasks. We draw on a diverse corpus—PubMed, StatPearls, medical textbooks, and Wikipedia—for authoritative, up-to-date information. For efficient retrieval, we employ the ColBERT model (Khattab and Zaharia, 2020), a dense retrieval approach optimized for passage-level retrieval, enabling fine-grained token-level matching
+
+to enhance the precision of retrieved information. Additional details on the retrieval corpus and model can be found in Appendix A.4, the prompt for each action can be found in Appendix A.7.
+
+# 3.3 Retrieval-Augmented Factuality Scorer (RAFS)
+
+Inspired by the Search Augmented Factuality Evaluator (SAFE) (Wei et al., 2024), which combines an LLM (GPT-3.5-turbo) with Google Search to evaluate the factuality of responses, RARE introduces the Retrieval-Augmented Factuality Scorer. RAFS adapts this approach by replacing GPT-3.5-turbo with LLaMA 3.1 and Google Search with a corpus index retrieval system containing both general-domain knowledge (Wikipedia) and medical-domain resources (PubMed, StatPearls, and Medical Textbooks). This adaptation enhances domain specificity and accessibility for tasks requiring specialized knowledge. To assess the factual accuracy of generated reasoning paths, RAFS evaluates each candidate trajectory in four systematic steps, as illustrated in Figure 5. More details about the retrieval corpus can be found in Appendix A.4.
+
+1. Split into Statements: Each reasoning path is divided into individual statements.  
+2. Generate Retrieval Queries: For each statement, RAFS employs an LLM to generate multiple retrieval queries designed to retrieve contextually relevant evidence.  
+3. Retrieve Information: The retrieval system gathers documents or information that corresponds to each generated query.  
+4. Rate Using Retrieved Information: Each statement is compared against the retrieved evidence and labeled as either Supported or Not Supported, based on alignment with the information. The overall factuality score for the reasoning path is calculated as the proportion of supported statements, indicating the trajectory's factual reliability.
+
+As shown in Figure 5, RAFS outputs a factuality score along with Supported or Not Supported labels for each statement. This scoring aids in selecting the most reliable reasoning path from multiple candidates, allowing RARE to prioritize responses that align closely with verified external knowledge.
+
+<table><tr><td>Model</td><td>Method</td><td>MedQA</td><td>MedMCQA</td><td>MMLU-M</td><td>Avg</td></tr><tr><td rowspan="6">LLaMA3.2 3B</td><td>CoT</td><td>52.63</td><td>49.82</td><td>57.67</td><td>53.37</td></tr><tr><td>MedRAG</td><td>52.08</td><td>51.78</td><td>65.58</td><td>56.48</td></tr><tr><td>i-MedRAG</td><td>60.88</td><td>53.60</td><td>66.76</td><td>60.41</td></tr><tr><td>SC</td><td>56.09</td><td>50.85</td><td>58.49</td><td>55.14</td></tr><tr><td>rStar</td><td>61.27</td><td>54.26</td><td>67.22</td><td>60.92</td></tr><tr><td>RARE</td><td>63.86</td><td>56.61</td><td>70.98</td><td>63.82</td></tr><tr><td rowspan="6">LLaMA3.1 8B</td><td>CoT</td><td>61.51</td><td>55.15</td><td>71.63</td><td>62.76</td></tr><tr><td>MedRAG</td><td>63.00</td><td>56.87</td><td>74.56</td><td>64.81</td></tr><tr><td>i-MedRAG</td><td>73.61</td><td>61.61</td><td>78.42</td><td>71.21</td></tr><tr><td>SC</td><td>64.73</td><td>56.35</td><td>72.73</td><td>64.60</td></tr><tr><td>rStar</td><td>70.40</td><td>62.13</td><td>79.16</td><td>70.56</td></tr><tr><td>RARE</td><td>75.57</td><td>64.32</td><td>81.63</td><td>73.84</td></tr><tr><td rowspan="6">LLaMA3.1 70B</td><td>CoT</td><td>76.67</td><td>68.75</td><td>81.72</td><td>75.71</td></tr><tr><td>MedRAG</td><td>77.61</td><td>71.19</td><td>84.76</td><td>77.85</td></tr><tr><td>i-MedRAG</td><td>82.40</td><td>72.38</td><td>86.69</td><td>80.49</td></tr><tr><td>SC</td><td>79.49</td><td>70.19</td><td>82.73</td><td>77.47</td></tr><tr><td>rStar</td><td>84.99</td><td>72.72</td><td>87.19</td><td>81.63</td></tr><tr><td>RARE</td><td>87.43</td><td>75.18</td><td>90.91</td><td>84.51</td></tr><tr><td>Meditron 70B</td><td>CoT</td><td>51.69</td><td>46.74</td><td>64.92</td><td>54.45</td></tr><tr><td>Mixtral (8x7B)</td><td>CoT</td><td>64.10</td><td>56.28</td><td>74.01</td><td>64.80</td></tr><tr><td>GPT-3.5</td><td>CoT</td><td>65.04</td><td>55.25</td><td>72.91</td><td>64.40</td></tr><tr><td>GPT-4</td><td>CoT</td><td>83.97</td><td>69.88</td><td>89.44</td><td>81.10</td></tr><tr><td>GPT-4o Mini</td><td>CoT</td><td>73.29</td><td>66.17</td><td>84.30</td><td>74.59</td></tr><tr><td>GPT-4o</td><td>CoT</td><td>85.55</td><td>74.70</td><td>90.45</td><td>83.57</td></tr></table>
+
+Table 1: Performance of RARE and baseline methods on three medical reasoning benchmarks: MedQA, MedMCQA, and MMLU-Medical. SC is self-consistency. RARE consistently outperforms rStar across all model sizes, with improvements statistically significant at  $\mathbf{p} < 0.01$  based on paired t-tests over multiple runs.
+
+# 4 Results
+
+In this section, we evaluate the performance of our proposed method, RARE, on both medical reasoning and commonsense reasoning tasks using three large language models: LLaMA 3.2 3B Instruct, LLaMA 3.1 8B Instruct and LLaMA 3.1 70B Instruct (Dubey et al., 2024). Throughout our work, we may drop "Instruct", but we are always referring to the "Instruct" versions. Detail settings of the experiments, descriptions of the evaluation tasks and baselines can be found in Appendix.
+
+# 4.1 Performance on Medical Reasoning tasks
+
+Table 1 shows the performance of RARE and various baseline methods on three challenging medical reasoning benchmarks: MedQA, MedMCQA, and MMLU-Medical. These datasets require not only complex reasoning but also a high degree of factual accuracy, making them suitable for evaluating the effectiveness of RARE's retrieval-augmented reasoning approach. The results demonstrate the effectiveness of RARE in enhancing the reasoning capabilities of LLaMA models compared to baseline methods, including Chain of Thought, MedRAG, iMedRAG, Self-Consistency(SC), and rStar. Across all model sizes—LLaMA3.2 3B, LLaMA3.1 8B, and LLaMA3.1 70B—RARE consistently outperforms baseline methods. On LLaMA3.1 8B, RARE
+
+achieves substantial gains, outperforming rStar by  $5.17\%$  on MedQA,  $2.19\%$  on MedMCQA, and  $2.47\%$  on MMLU-Medical. The performance improvement becomes more pronounced as model size increases, with RARE-enhanced LLaMA3.1 70B outperforming GPT-4o on MedQA  $(87.43\%)$  vs.  $85.55\%$  and MMLU-Medical  $(90.91\%)$  vs.  $90.45\%$  ), highlighting its competitive edge.
+
+# 4.2 Performance on Commonsense Reasoning
+
+Table 2 presents the performance of RARE compared to other methods and larger language models on commonsense reasoning benchmarks, including StrategyQA, CommonsenseQA, and Physical IQA. These datasets test a range of commonsense reasoning skills, with StrategyQA requiring more multi-step reasoning. RARE consistently outperforms baseline methods, including CoT, MedRAG, iMedRAG, SC and rStar, across both LLaMA3.1 8B and LLaMA3.1 70B models. For LLaMA3.1 8B, RARE achieves substantial improvements over rStar, with gains of  $6.45\%$  on StrategyQA,  $4.26\%$  on CommonsenseQA, and  $2.87\%$  on PIQA. On LLaMA3.1 70B, RARE further closes the gap with state-of-the-art proprietary models, achieving  $85.74\%$  on StrategyQA,  $86.98\%$  on CommonsenseQA, and  $92.66\%$  on PIQA, surpassing GPT-4o.
+
+RARE's retrieval-augmented reasoning method provides substantially larger gains over CoT on multi-step inference tasks (like StrategyQA) compared to commonsense-heavy tasks (such as CommonsenseQA), indicating its particular effectiveness in handling implicit, multi-hop reasoning. Specifically, RARE's largest improvement over CoT is observed on StrategyQA(10.19%), where multi-step reasoning is crucial, suggesting that its retrieval-augmented reasoning enhancement method is particularly effective in handling implicit, multi-hop inference. In contrast, while RARE outperforms CoT on CommonsenseQA(7.22%), the gains are relatively smaller, indicating that tasks relying more on commonsense knowledge rather than explicit step-by-step reasoning do not benefit as significantly from retrieval-augmented reasoning.
+
+# 4.3 Ablation Study
+
+We conduct an ablation study on 250 MedQA samples using the LLaMA 3.1 8B model to assess the contributions of each RARE component (Table 3). Starting with the baseline (rStar) at  $70.0\%$  accu
+
+<table><tr><td>Model</td><td>Method</td><td>SQA</td><td>CQA</td><td>PIQA</td><td>Avg</td></tr><tr><td rowspan="6">LLaMA3.1 8B</td><td>CoT</td><td>67.83</td><td>73.62</td><td>76.17</td><td>72.54</td></tr><tr><td>MedRAG</td><td>66.08</td><td>74.45</td><td>78.67</td><td>73.07</td></tr><tr><td>i-MedRAG</td><td>68.12</td><td>75.18</td><td>80.52</td><td>74.61</td></tr><tr><td>SC</td><td>68.41</td><td>74.90</td><td>77.42</td><td>73.58</td></tr><tr><td>rStar</td><td>71.57</td><td>76.58</td><td>83.04</td><td>77.06</td></tr><tr><td>RARE</td><td>78.02</td><td>80.84</td><td>85.91</td><td>81.59</td></tr><tr><td rowspan="6">LLaMA3.1 70B</td><td>CoT</td><td>76.71</td><td>78.62</td><td>81.66</td><td>79.00</td></tr><tr><td>MedRAG</td><td>75.54</td><td>82.23</td><td>86.07</td><td>81.28</td></tr><tr><td>i-MedRAG</td><td>77.29</td><td>83.13</td><td>87.76</td><td>82.73</td></tr><tr><td>SC</td><td>77.29</td><td>78.87</td><td>82.67</td><td>79.61</td></tr><tr><td>rStar</td><td>81.80</td><td>83.66</td><td>89.27</td><td>84.91</td></tr><tr><td>RARE</td><td>85.74</td><td>86.98</td><td>92.66</td><td>88.46</td></tr><tr><td>Claude-3 Haiku</td><td>CoT</td><td>69.58</td><td>67.40</td><td>82.32</td><td>73.10</td></tr><tr><td>Claude-3.5 Sonnet</td><td>CoT</td><td>76.86</td><td>74.12</td><td>89.39</td><td>80.12</td></tr><tr><td>GPT-4o Mini</td><td>CoT</td><td>78.60</td><td>82.31</td><td>88.41</td><td>83.11</td></tr><tr><td>GPT-4o</td><td>CoT</td><td>80.64</td><td>86.50</td><td>91.13</td><td>86.09</td></tr></table>
+
+Table 2: Performance comparison on common sense reasoning tasks with various LLMs and reasoning methods, evaluated on StrategyQA (SQA), Common-senseQA (CQA), and Physical IQA (PIQA). SC is self-consistency. RARE consistently outperforms rStar across all model sizes, with improvements statistically significant at  $\mathbf{p} < 0.01$  based on paired t-tests over multiple runs.  
+
+<table><tr><td>Configuration</td><td>Accuracy</td></tr><tr><td>rStar</td><td>70.0</td></tr><tr><td>rStar + RAFS</td><td>71.4</td></tr><tr><td>rStar + A6</td><td>72.4</td></tr><tr><td>rStar + A7</td><td>71.2</td></tr><tr><td>rStar + A6 + A7</td><td>73.2</td></tr><tr><td>RARE (rStar + A6 + A7 + RAFS)</td><td>74.8</td></tr></table>
+
+Table 3: Ablation study on RARE components, evaluated on 250 MedQA samples using LLaMA 3.1 8B.
+
+racy, adding the RAFS improves reliability, increasing accuracy to  $71.4\%$ . Incorporating A6 and A7 further boosts accuracy to  $72.4\%$  and  $71.2\%$ , respectively. Combining both actions yields  $73.2\%$ , demonstrating their synergy in strengthening reasoning. The full RARE configuration, integrating rStar, A6, A7, and RAFS, achieves the highest accuracy  $(74.8\%)$ , highlighting the collective impact of retrieval and factuality scoring in enhancing reasoning accuracy.
+
+# 4.4 Common Reasoning Paths Patterns
+
+Exploring common paths leading to correct answers helps us understand and refine the reasoning patterns that most reliably produce accurate solutions. Figures 8 and 9 illustrate the top 10 most common reasoning paths that lead to correct answers on MedQA and StrategyQA, respectively.
+
+In MedQA, prominent paths such as A1  $\rightarrow$  A2,A3  $\rightarrow$  A2,and A1  $\rightarrow$  A6 constitute a large
+
+share of successful reasoning. More complex sequences like  $\mathrm{A}3 \to \mathrm{A}7 \to \mathrm{A}3$  also appear, demonstrating the generator's capacity to explore multiple approaches adaptively.
+
+A similar pattern occurs on StrategyQA, where paths like A1  $\rightarrow$  A2, A3  $\rightarrow$  A2, and A1  $\rightarrow$  A6 remain dominant. Still, simpler actions such as A6 and A2 also play a noticeable role, reflecting the distinct reasoning demands of this task.
+
+Overall, both figures highlight the RARE's flexibility in navigating a variety of reasoning strategies. They further illustrate the importance of both straightforward and more intricate paths in effectively tackling different categories of tasks.
+
+# 4.5 Human Evaluation of the Retrieval-Augmented Factuality Scorer
+
+<table><tr><td>Metric</td><td>Score (%)</td></tr><tr><td>Inter-Annotator Agreement</td><td>86.49</td></tr><tr><td>RAFS Alignment with Annotator 1</td><td>87.84</td></tr><tr><td>RAFS Alignment with Annotator 2</td><td>82.43</td></tr><tr><td>Average RAFS-Annotator Alignment</td><td>85.14</td></tr></table>
+
+Table 4: Human evaluation of RAFS.
+
+To assess the effectiveness of RAFS, we conducted a human evaluation comparing its factual assessments with expert judgments. We selected 10 medical questions from MedQA and generated corresponding responses using RARE with LLaMA 3 8B, segmenting each response into approximately 100 individual statements. RAFS classified each statement as either Supported or Not Supported. Two medical experts—both of whom have passed the United States Medical Licensing Examination (USMLE)—independently labeled each statement as Correct or Incorrect, corresponding to RAFS's Supported and Not Supported labels, respectively.
+
+Table 4 shows a high inter-annotator agreement of  $86.49\%$ , confirming consistency among expert evaluations. RAFS achieves strong alignment with human judgments, with an average agreement of  $85.14\%$  (87.84% with Annotator 1 and  $82.43\%$  with Annotator 2), demonstrating its reliability in factual assessment. The slightly lower alignment with Annotator 2 suggests some subjectivity in evaluating borderline statements. Overall, RAFS effectively prioritizes factually accurate reasoning paths, closely aligning with expert validation.
+
+![](images/766d2cdff86020e7d6dde4ecad4d4c8a9242c3b1a692372678efcf5ef49bcb70.jpg)  
+Figure 6: Accuracy vs. Number of Generated Tokens for Different Methods
+
+# 4.6 Computational Analysis
+
+Figure 6 illustrates the trade-off between inference cost and reasoning performance, averaged across commonsense reasoning tasks. CoT requires approximately 400 tokens per example, while rStar averages 92.6k tokens, and RARE increases this to 119.9k tokens due to its additional retrieval-augmented steps. On average, rStar improves accuracy over CoT by  $4.52\%$ , and RARE achieves a  $9.05\%$  improvement—the highest among all methods. Although RARE uses roughly 27.3k more tokens than rStar, its superior accuracy demonstrates that the additional computation enhances reasoning robustness. This finding aligns with recent trends such as OpenAI O1 and DeepSeek R1 (El-Kishky et al., 2025; Guo et al., 2025), which prioritize deeper reasoning over minimal latency, reinforcing that improved accuracy can justify increased computational cost in complex reasoning tasks.
+
+# 5 Related Work
+
+# 5.1 Prompting LLMs to reason
+
+Prompting LLMs to reason has become a central focus of recent research, particularly through the development of prompting-based techniques such as Chain-of-Thought prompting (Wei et al., 2022). These approaches aim to improve inference capabilities by structuring prompts to encourage multi-step reasoning. Key advancements include planning-based prompting (Hao et al., 2023; Ding et al., 2023), problem decomposition (Zhou et al., 2022; Khot et al., 2022), abstraction (Zheng et al., 2023), and program-based reasoning (Chen et al., 2022; Zhou et al., 2023b). While effective for single-step inference, these methods often rely on a limited range of operations, which can restrict the diversity and generality of the solutions generated.
+
+MCTS has emerged as a powerful tool for optimizing reasoning paths across a vast solution space, enhancing both exploration and decision-making efficiency (Silver et al., 2018). MCTS has been widely adopted in domains such as game theory (Sironi et al., 2018; Ontanón, 2016) and strategic planning (Zhou et al., 2023a; Yu et al., 2023). When integrated with reinforcement learning, MCTS enables self-play training, achieving human-level or even superhuman performance in complex domains like Go (Silver et al., 2016).
+
+More recently, MCTS has been applied to LLMs to identify optimal reasoning trajectories without requiring additional supervision (Feng et al., 2023; Zhou et al., 2023a; Tian et al., 2024; Huang et al., 2024). For instance, Feng et al. (2023) restricted the search granularity to word or sentence level, while Tian et al. (2024) introduced  $\eta$ MCTS to enable hierarchical planning with tailored reward functions. MCTS has also been employed to curate high-quality reasoning paths for training reward models, leading to iterative improvements in model reasoning (Zhang et al., 2024). Zhou et al. (2023a) further generalized the use of MCTS by incorporating all possible reasoning and action steps into a unified search space, supporting joint inference, action, and planning.
+
+The most relevant work to ours is rStar (Qi et al., 2024), which extends MCTS by incorporating a diverse set of reasoning operations. rStar improves solution-space exploration through dynamic, human-like reasoning and introduces five distinct operations that enable more flexible planning. Moreover, it employs a discriminator—another LLM with reasoning capabilities similar to the main model—to verify the plausibility of each candidate reasoning path.
+
+# 5.2 Medical RAG
+
+Retrieval-augmented generation has proven highly effective in grounding LLM reasoning with up-to-date external knowledge, especially in medical tasks such as question answering and generation (Xiong et al., 2024a; Tian et al., 2019; Xia et al., 2022; Wang et al., 2024; Yao et al., 2024; Yao and Yu, 2025). RAG has also been applied to classification, information extraction, lay-language generation (Li et al., 2024; Guo et al., 2024; Yao et al., 2023), and medical dialogue systems (Shi et al., 2024). To enhance retrieval effectiveness, various improvements have been proposed, such as
+
+query rewriting (Zhang et al., 2022) and multi-step retrieval (Mrini et al., 2022), enabling LLMs to iteratively refine their knowledge intake for clinical decision-making and literature synthesis (Zakka et al., 2024; Wang et al., 2023; Jeong et al., 2024). For example, Wang et al. (2023) developed a hybrid retriever with complex filtering mechanisms to identify high-quality documents, while Jeong et al. (2024) introduced SelfBioRAG, a self-reflective retriever that integrates reasoning signals into the retrieval loop. Iterative RAG variants like i-MedRAG allow LLMs to refine their understanding by posing follow-up queries in multiple rounds (Xiong et al., 2024b). The most relevant RAG-based system to our work is SeRTS (Hu et al., 2024), which leverages MCTS to guide query generation. However, SeRTS follows a depth-first strategy, generating and executing one query at a time, whereas our framework—RARE—adopts a more flexible, agentic planning approach.
+
+RARE integrates both RAG-based and non-RAG operations, allowing for breadth-first exploration via A6, which generates multiple queries simultaneously. This enables RARE to handle a wider variety of reasoning tasks. In contrast to SeRTS, which always requires iterative retrieval, RARE can also answer simple problems directly via step-by-step (CoT-like) reasoning without invoking retrieval. This design makes RARE more versatile, allowing it to dynamically adapt between retrieval-driven and retrieval-free reasoning depending on the task complexity.
+
+# 6 Conclusion
+
+We introduced RARE (Retrieval-Augmented Reasoning Enhancement), a framework designed to improve the reasoning accuracy and factual reliability of large language models (LLMs) through retrieval-augmented actions and factuality scoring. RARE operates entirely as an autonomous language agent, requiring no additional training or fine-tuning of the underlying LLM. This makes the framework robust to overfitting and highly adaptable across tasks and datasets, as it relies solely on real-time retrieval and reasoning mechanisms. Experiments on medical and commonsense reasoning benchmarks demonstrate RARE's effectiveness, RARE bridges the gap between open-source models and state-of-the-art proprietary systems, showcasing its potential as a scalable and effective solution for knowledge-intensive reasoning tasks.
+
+# 7 Limitations
+
+While RARE demonstrates significant improvements in reasoning accuracy and factual reliability, it has certain limitations that present opportunities for future work.
+
+First, the framework has only been tested on open-source models like LLaMA 3.1 and not on larger proprietary models such as GPT-4. This is due to the high number of API calls required by RARE's iterative retrieval and reasoning process, making evaluations on closed-source models prohibitively costly. However, the framework is designed to be model-agnostic and can be directly applied to proprietary models if resources permit.
+
+Second, RARE is designed to identify a single reasoning path that leads to a correct answer but does not necessarily optimize for the best or shortest path that maximizes robustness (e.g., achieving the highest model confidence). Future work could explore designing better reward functions to prevent reward hacking and improve the selection of the most reliable reasoning paths.
+
+Finally, RARE is currently limited to using Monte Carlo Tree Search for exploring action paths. While effective, this approach does not leverage a trained reward model to dynamically guide the search process. Future extensions could incorporate reward models or alternative optimization strategies to further enhance reasoning quality and efficiency.
+
+These limitations highlight areas for improvement and potential research directions to make RARE more robust, generalizable, and applicable to a wider range of models and reasoning tasks.
+
+# 8 Ethics Statement
+
+This work aims to advance the field of Medical QA by enhancing the reasoning capabilities of language models through the RARE framework. While the results demonstrate significant improvements, several ethical considerations must be addressed to ensure responsible development and deployment:
+
+Clinical Applicability. RARE is designed to improve reasoning and factual reliability but is not intended to replace healthcare professionals or serve as a standalone diagnostic or treatment tool. Any integration into medical workflows must be supervised by qualified practitioners to ensure patient safety and ethical use.
+
+Bias and Fairness. Language models, including those tested with RARE, may reflect biases
+
+present in their training data. These biases could impact the fairness and reliability of the reasoning process, particularly in sensitive medical contexts. Future work must include rigorous audits for bias and fairness to minimize potential harm.
+
+Generalizability. As RARE has been primarily evaluated in English-language, text-based general and medical domain QA tasks, its applicability to non-English-speaking contexts and multimodal scenarios remains untested. Efforts should be made to extend the framework to diverse linguistic and cultural contexts to ensure equitable access to its benefits.
+
+Societal Impacts. While RARE demonstrates the potential for improving medical reasoning tasks, its outputs should be considered supplementary to human expertise. The ethical deployment of RARE requires clear guidelines to avoid overreliance on AI and ensure that it enhances, rather than replaces, human decision-making in healthcare.
+
+Human Annotation. The human evaluation in this study was conducted by two medical experts who have passed the United States Medical Licensing Examination (USMLE). Annotators were compensated fairly for their time at a rate of $40/hour. All participants were informed about the scope of their role and participated voluntarily.
+
+# References
+
+Lisa Bauer, Yicheng Wang, and Mohit Bansal. 2018. Commonsense for generative multi-hop question answering tasks. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 4220-4230.  
+Yonatan Bisk, Rowan Zellers, Jianfeng Gao, Yejin Choi, et al. 2020. Piqa: Reasoning about physical commonsense in natural language. In Proceedings of the AAAI conference on artificial intelligence, volume 34, pages 7432-7439.  
+Cameron B Browne, Edward Powley, Daniel Whitehouse, Simon M Lucas, Peter I Cowling, Philipp Rohlfshagen, Stephen Tavener, Diego Perez, Spyridon Samothrakis, and Simon Colton. 2012. A survey of monte carlo tree search methods. IEEE Transactions on Computational Intelligence and AI in games, 4(1):1-43.  
+Pengshan Cai, Zonghai Yao, Fei Liu, Dakuo Wang, Meghan Reilly, Huixue Zhou, Lingxi Li, Yi Cao, Alok Kapoor, Adarsha Bajracharya, et al. 2023. Paniniqa: Enhancing patient education through interactive question answering. Transactions of the Association for Computational Linguistics, 11:1518-1536.
+
+YongGang Cao, Feifan Liu, Pippa Simpson, Lamont Antieau, Andrew Bennett, James J Cimino, John Ely, and Hong Yu. 2011. Askhermes: An online question answering system for complex clinical questions. Journal of biomedical informatics, 44(2):277-288.  
+Qianglong Chen, Feng Ji, Haiqing Chen, and Yin Zhang. 2020. Improving commonsense question answering by graph-based iterative retrieval over multiple knowledge sources. In Proceedings of the 28th International Conference on Computational Linguistics, pages 2583-2594.  
+Wenhu Chen, Xueguang Ma, Xinyi Wang, and William W Cohen. 2022. Program of thoughts prompting: Disentangling computation from reasoning for numerical reasoning tasks. arXiv preprint arXiv:2211.12588.  
+Ruomeng Ding, Chaoyun Zhang, Lu Wang, Yong Xu, Minghua Ma, Wei Zhang, Si Qin, Saravan Rajmohan, Qingwei Lin, and Dongmei Zhang. 2023. Everything of thoughts: Defying the law of penrose triangle for thought generation. arXiv preprint arXiv:2311.04254.  
+Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.  
+Ahmed El-Kishky, Alexander Wei, Andre Saraiva, Borys Minaev, Daniel Selsam, David Dohan, Francis Song, Hunter Lightman, Ignasi Clavera, Jakub Pachocki, et al. 2025. Competitive programming with large reasoning models. arXiv preprint arXiv:2502.06807.  
+Xidong Feng, Ziyu Wan, Muning Wen, Stephen Marcus McAleer, Ying Wen, Weinan Zhang, and Jun Wang. 2023. Alphazero-like tree-search can guide large language model decoding and training. arXiv preprint arXiv:2309.17179.  
+Mor Geva, Daniel Khashabi, Elad Segal, Tushar Khot, Dan Roth, and Jonathan Berant. 2021. Did aristotle use a laptop? a question answering benchmark with implicit reasoning strategies. Transactions of the Association for Computational Linguistics, 9:346-361.  
+Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948.  
+Yue Guo, Wei Qiu, Gondy Leroy, Sheng Wang, and Trevor Cohen. 2024. Retrieval augmentation of large language models for lay language generation. Journal of Biomedical Informatics, 149:104580.  
+Shibo Hao, Yi Gu, Haodi Ma, Joshua Jiahua Hong, Zhen Wang, Daisy Zhe Wang, and Zhiting Hu. 2023. Reasoning with language model is planning with world model. arXiv preprint arXiv:2305.14992.
+
+Minda Hu, Licheng Zong, Hongru Wang, Jingyan Zhou, Jingjing Li, Yichen Gao, Kam-Fai Wong, Yu Li, and Irwin King. 2024. Serts: Self-rewarding tree search for biomedical retrieval-augmented generation. In *Findings of the Association for Computational Linguistics: EMNLP* 2024, pages 1321-1335.  
+Jiatan Huang, Mingchen Li, Zonghai Yao, Zhichao Yang, Yongkang Xiao, Feiyun Ouyang, Xiaohan Li, Shuo Han, and Hong Yu. 2024. Ritek: A dataset for large language models complex reasoning over textual knowledge graphs. arXiv preprint arXiv:2410.13987.  
+Minbyul Jeong, Jiwoong Sohn, Mujeen Sung, and Jaewoo Kang. 2024. Improving medical reasoning through retrieval and self-reflection with retrieval-augmented large language models. Bioinformatics, 40(Supplement_1):i119-i129.  
+Di Jin, Eileen Pan, Nassim Oufattole, Wei-Hung Weng, Hanyi Fang, and Peter Szolovits. 2021. What disease does this patient have? a large-scale open domain question answering dataset from medical exams. Applied Sciences, 11(14):6421.  
+Qiao Jin, Robert Leaman, and Zhiyong Lu. 2024a. Pubmed and beyond: biomedical literature search in the age of artificial intelligence. *EBioMedicine*, 100.  
+Qiao Jin, Zifeng Wang, Charalampos S Floudas, Fangyuan Chen, Changlin Gong, Dara Bracken-Clarke, Elisabetta Xue, Yifan Yang, Jimeng Sun, and Zhiyong Lu. 2024b. Matching patients to clinical trials with large language models. Nature Communications, 15(1):9074.  
+Omar Khattab and Matei Zaharia. 2020. Colbert: Efficient and effective passage search via contextualized late interaction over bert. In Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval, pages 39-48.  
+Tushar Khot, Harsh Trivedi, Matthew Finlayson, Yao Fu, Kyle Richardson, Peter Clark, and Ashish Sabharwal. 2022. Decomposed prompting: A modular approach for solving complex tasks. arXiv preprint arXiv:2210.02406.  
+Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, et al. 2019. Natural questions: a benchmark for question answering research. Transactions of the Association for Computational Linguistics, 7:453-466.  
+Mingchen Li, Halil Kilicoglu, Hua Xu, and Rui Zhang. 2024. Biomedrag: A retrieval augmented large language model for biomedicine. arXiv preprint arXiv:2405.00465.
+
+Siru Liu, Aileen P Wright, Barron L Patterson, Jonathan P Wanderer, Robert W Turer, Scott D Nelson, Allison B McCoy, Dean F Sittig, and Adam Wright. 2023. Using ai-generated suggestions from chatgpt to optimize clinical decision support. Journal of the American Medical Informatics Association, 30(7):1237-1245.  
+Zhiyong Lu. 2011. Pubmed and beyond: a survey of web tools for searching biomedical literature. Database, 2011:baq036.  
+Khalil Mrini, Harpreet Singh, Franck Dernoncourt, Seunghyun Yoon, Trung Bui, Walter Chang, Emilia Farcas, and Ndapa Nakashole. 2022. Medical question understanding and answering with knowledge grounding and semantic self-supervision. arXiv preprint arXiv:2209.15301.  
+Santiago Ontanón. 2016. Informed monte carlo tree search for real-time strategy games. In 2016 IEEE Conference on Computational Intelligence and Games (CIG), pages 1-8. IEEE.  
+Ankit Pal, Logesh Kumar Umapathi, and Malaikannan Sankarasubbu. 2022. Medmcqa: A large-scale multi-subject multi-choice dataset for medical domain question answering. In Conference on health, inference, and learning, pages 248-260. PMLR.  
+Zhenting Qi, Mingyuan Ma, Jiahang Xu, Li Lyna Zhang, Fan Yang, and Mao Yang. 2024. Mutual reasoning makes smaller llms stronger problem-solvers. arXiv preprint arXiv:2408.06195.  
+Xiaoming Shi, Zeming Liu, Li Du, Yuxuan Wang, Hongru Wang, Yuhang Guo, Tong Ruan, Jie Xu, and Shaoting Zhang. 2024. Medical dialogue: A survey of categories, methods, evaluation and challenges. arXiv preprint arXiv:2405.10630.  
+David Silver, Aja Huang, Chris J Maddison, Arthur Guez, Laurent Sifre, George Van Den Driessche, Julian Schrittwieser, Ioannis Antonoglou, Veda Panneershelvam, Marc Lanctot, et al. 2016. Mastering the game of go with deep neural networks and tree search. nature, 529(7587):484-489.  
+David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Arthur Guez, Marc Lanctot, Laurent Sifre, Dharshan Kumaran, Thore Graepel, et al. 2018. A general reinforcement learning algorithm that masters chess, shogi, and go through self-play. Science, 362(6419):1140-1144.  
+Karan Singhal, Shekoofeh Azizi, Tao Tu, S Sara Mahdavi, Jason Wei, Hyung Won Chung, Nathan Scales, Ajay Tanwani, Heather Cole-Lewis, Stephen Pfohl, et al. 2023a. Large language models encode clinical knowledge. Nature, 620(7972):172-180.  
+Karan Singhal, Tao Tu, Juraj Gottweis, Rory Sayres, Ellery Wulczyn, Le Hou, Kevin Clark, Stephen Pfohl, Heather Cole-Lewis, Darlene Neal, et al. 2023b. Towards expert-level medical question answering with large language models. arXiv preprint arXiv:2305.09617.
+
+Chiara F Sironi, Jialin Liu, Diego Perez-Liebana, Raluca D Gaina, Ivan Bravi, Simon M Lucas, and Mark HM Winands. 2018. Self-adaptive mcts for general video game playing. In International Conference on the Applications of Evolutionary Computation, pages 358-375. Springer.  
+Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. 2018. Commonsenseqa: A question answering challenge targeting commonsense knowledge. arXiv preprint arXiv:1811.00937.  
+Nandan Thakur, Nils Reimers, Andreas Rückle, Abhishek Srivastava, and Iryna Gurevych. 2021. Beir: A heterogenous benchmark for zero-shot evaluation of information retrieval models. arXiv preprint arXiv:2104.08663.  
+Ye Tian, Baolin Peng, Linfeng Song, Lifeng Jin, Dian Yu, Haitao Mi, and Dong Yu. 2024. Toward self-improvement of llms via imagination, searching, and criticizing. arXiv preprint arXiv:2404.12253.  
+Yuanhe Tian, Weicheng Ma, Fei Xia, and Yan Song. 2019. Chimed: A chinese medical corpus for question answering. In Proceedings of the 18th BioNLP Workshop and Shared Task, pages 250-260.  
+Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2023. Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 10014-10037.  
+Junda Wang, Zhichao Yang, Zonghai Yao, and Hong Yu. 2024. Jmlr: Joint medical llm and retrieval training for enhancing reasoning and professional question answering capability. arXiv preprint arXiv:2402.17887.  
+Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2022. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171.  
+Yubo Wang, Xueguang Ma, and Wenhu Chen. 2023. Augmenting black-box llms with medical textbooks for clinical question answering. arXiv preprint arXiv:2309.02233.  
+Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824-24837.  
+Jerry Wei, Chengrun Yang, Xinying Song, Yifeng Lu, Nathan Hu, Jie Huang, Dustin Tran, Daiyi Peng, Ruibo Liu, Da Huang, et al. 2024. Long-form factuality in large language models. arXiv preprint arXiv:2403.18802.
+
+Chaoyi Wu, Weixiong Lin, Xiaoman Zhang, Ya Zhang, Weidi Xie, and Yanfeng Wang. 2024. Pmc-llama: toward building open-source language models for medicine. Journal of the American Medical Informatics Association, page ocae045.  
+Fei Xia, Bin Li, Yixuan Weng, Shizhu He, Kang Liu, Bin Sun, Shutao Li, and Jun Zhao. 2022. Medconqa: medical conversational question answering system based on knowledge graphs. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 148-158.  
+Guangzhi Xiong, Qiao Jin, Zhiyong Lu, and Aidong Zhang. 2024a. Benchmarking retrieval-augmented generation for medicine. arXiv e-prints, pages arXiv-2402.  
+Guangzhi Xiong, Qiao Jin, Xiao Wang, Minjia Zhang, Zhiyong Lu, and Aidong Zhang. 2024b. Improving retrieval-augmented generation in medicine with iterative follow-up questions. arXiv preprint arXiv:2408.00727.  
+Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William W Cohen, Ruslan Salakhutdinov, and Christopher D Manning. 2018. Hotpotqa: A dataset for diverse, explainable multi-hop question answering. arXiv preprint arXiv:1809.09600.  
+Zonghai Yao, Nandyala Siddharth Kantu, Guanghao Wei, Hieu Tran, Zhangqi Duan, Sunjae Kwon, Zhichao Yang, Hong Yu, et al. 2023. Readme: Bridging medical jargon and lay understanding for patient education through data-centric nlp. arXiv preprint arXiv:2312.15561.  
+Zonghai Yao, Aditya Parashar, Huixue Zhou, Won Seok Jang, Feiyun Ouyang, Zhichao Yang, and Hong Yu. 2024. Mcqq-srefine: Multiple choice question generation and evaluation with iterative self-critique, correction, and comparison feedback. arXiv preprint arXiv:2410.13191.  
+Zonghai Yao and Hong Yu. 2025. A survey on llm-based multi-agent ai hospital.  
+Xiao Yu, Maximillian Chen, and Zhou Yu. 2023. Prompt-based monte-carlo tree search for goal-oriented dialogue policy planning. arXiv preprint arXiv:2305.13660.  
+Cyril Zakka, Rohan Shad, Akash Chaurasia, Alex R Dalal, Jennifer L Kim, Michael Moor, Robyn Fong, Curran Phillips, Kevin Alexander, Euan Ashley, et al. 2024. Almanac—retrieval-augmented language models for clinical medicine. NEJM AI, 1(2):AIoa2300068.  
+Dan Zhang, Sining Zhoubian, Ziniu Hu, Yisong Yue, Yuxiao Dong, and Jie Tang. 2024. Rest-mcts*: Llm self-training via process reward guided tree search. arXiv preprint arXiv:2406.03816.
+
+Ming Zhang, Shuai Dou, Ziyang Wang, and Yunfang Wu. 2022. Focus-driven contrastive learning for medical question summarization. In Proceedings of the 29th international conference on computational linguistics, pages 6176-6186.  
+Huaixiu Steven Zheng, Swaroop Mishra, Xinyun Chen, Heng-Tze Cheng, Ed H Chi, Quoc V Le, and Denny Zhou. 2023. Take a step back: Evoking reasoning via abstraction in large language models. arXiv preprint arXiv:2310.06117.  
+Andy Zhou, Kai Yan, Michal Shlapentokh-Rothman, Haohan Wang, and Yu-Xiong Wang. 2023a. Language agent tree search unifies reasoning acting and planning in language models. arXiv preprint arXiv:2310.04406.  
+Aojun Zhou, Ke Wang, Zimu Lu, Weikang Shi, Sichun Luo, Zipeng Qin, Shaoqing Lu, Anya Jia, Linqi Song, Mingjie Zhan, et al. 2023b. Solving challenging math word problems using gpt-4 code interpreter with code-based self-verification. arXiv preprint arXiv:2308.07921.  
+Denny Zhou, Nathanael Scharli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuurmans, Claire Cui, Olivier Bousquet, Quoc Le, et al. 2022. Least-to-most prompting enables complex reasoning in large language models. arXiv preprint arXiv:2205.10625.
+
+# A Appendix
+
+# A.1 Implementation Details
+
+In the MCTS framework of the Retrievalaugmented Generator, we set the number of rolls to 4 for LLaMA 3.2 3B and LLaMA 3.1 8B models, and 2 for the LLaMA 3.1 70B model. This configuration strikes a balance between effective inference and computational efficiency, particularly for larger models where inference costs are higher.
+
+In the factuality scoring stage, we perform a self-scorer setup, where the Retrieval-Augmented Factuality Scorer uses the same backbone model as the generator. For instance, when the generator utilizes LLaMA 3.1 3B, the RAFS also employs the LLaMA 3.2 3B model for factuality evaluation. This ensures consistency between the generator and scorer while maintaining efficient inference. All inference processes, including factuality scoring, are parallelized to further enhance efficiency.
+
+# A.2 Evaluation tasks
+
+To rigorously test the reasoning capabilities of RARE, we evaluate it on a range of reasoning tasks, categorized into two main domains:
+
+Medical Reasoning Tasks: We use three medical datasets that require complex, domain-specific reasoning, including:
+
+- MedQA-USMLE (Jin et al., 2021): A medical question-answering dataset based on the United States Medical Licensing Examination (USMLE) questions.  
+- MedMCQA (Pal et al., 2022): A dataset consisting of multiple-choice medical questions designed to test clinical knowledge.  
+- MMLU-Medical (Singhal et al., 2023a): The medical subset of the Massive Multitask Language Understanding (MMLU) benchmark, focusing on diverse topics in the medical field.
+
+Commonsense Reasoning Tasks: We evaluate RARE's general reasoning ability on commonsense datasets. While StrategyQA requires complex, implicit reasoning strategies, other tasks benefit from advanced reasoning but may not require it to the same extent:
+
+- StrategyQA (Geva et al., 2021): A dataset of open-domain questions that require implicit reasoning strategies.
+
+- CommonsenseQA (Talmor et al., 2018): A multiple-choice question-answering dataset designed to test commonsense knowledge.  
+PIQA (Physical Interaction QA) (Bisk et al., 2020): A dataset for physical reasoning, where models must answer questions about common physical interactions.
+
+# A.3 Baselines
+
+We compare the performance of RARE with several baseline reasoning methods, including:
+
+- Chain of Thought (CoT) (Wei et al., 2022): A reasoning approach that generates explanations step-by-step, aiming for more coherent answers.  
+- Self-Consistency (Wang et al., 2022): A method that uses majority voting among multiple reasoning paths to increase response accuracy.  
+- rStar (Qi et al., 2024): A framework that extends MCTS with a diverse set of reasoning actions, improving reasoning accuracy.  
+- MedRAG (Xiong et al., 2024a): A retrieval-augmented framework designed for medical question answering, integrating knowledge retrieval from domain-specific corpora.  
+- i-MedRAG (Xiong et al., 2024b): An iterative retrieval-based medical reasoning model that dynamically refines queries to improve response accuracy in medical QA tasks.
+
+A.4 Retrieval Model and Corpus  
+
+<table><tr><td>Corpus</td><td>#Docs</td><td>#Snippets</td><td>Avg. L</td><td>Domain</td></tr><tr><td>PubMed</td><td>23.9M</td><td>23.9M</td><td>296</td><td>Biomed.</td></tr><tr><td>StatPearls</td><td>9.3k</td><td>301.2k</td><td>119</td><td>Clinics</td></tr><tr><td>Textbooks</td><td>18</td><td>125.8k</td><td>182</td><td>Medicine</td></tr><tr><td>Wikipedia</td><td>6.5M</td><td>29.9M</td><td>162</td><td>General</td></tr><tr><td>MedCorp</td><td>30.4M</td><td>54.2M</td><td>221</td><td>Mixed</td></tr></table>
+
+Table 5: Statistics of the retrieval corpus used in our experiments. #Docs refers to the number of documents, #Snippets represents extracted text units, and Avg. L indicates the average snippet length.
+
+For information retrieval, we leverage the MedCorp (Xiong et al., 2024a) corpus, a curated collection of high-quality and domain-specific resources
+
+that serve as knowledge bases for medical and general question answering. The MedCorp corpus consists of (table 5):
+
+- PubMed<sup>2</sup>: A widely used biomedical literature database containing over 36 million articles (Lu, 2011; Jin et al., 2024a). For our retrieval tasks, we utilize a subset of 23.9 million articles with valid titles and abstracts, similar to the MedRAG setup (Xiong et al., 2024a).  
+- StatPearls<sup>3</sup>: A clinical decision support resource with publicly available medical articles hosted on NCBI Bookshelf<sup>4</sup>. The corpus includes 9,330 peer-reviewed StatPearls articles, structured into hierarchical snippets where each paragraph is treated as a retrieval unit, with corresponding hierarchical headings as metadata.  
+Medical Textbooks: A collection of 18 widely used medical textbooks (Jin et al., 2021) that are commonly referenced for foundational medical knowledge and United States Medical Licensing Examination (USMLE) preparation. The textbook corpus is segmented into passages of up to 1,000 characters using the RecursiveCharacterTextSplitter from LangChain.  
+- Wikipedia: A large-scale, general-domain encyclopedia frequently used in information retrieval tasks (Thakur et al., 2021). We incorporate a processed version of Wikipedia from Huggingface and apply text chunking techniques to facilitate passage-level retrieval.
+
+To retrieve relevant passages from MedCorp, we employ the ColBERT retrieval model (Khattab and Zaharia, 2020), an efficient neural retrieval framework optimized for passage ranking. ColBERT's late interaction mechanism enables fine-grained token-level relevance matching, allowing our retrieval-augmented generator to identify and incorporate contextually relevant medical knowledge. This approach enhances the factual accuracy and reliability of generated responses while efficiently handling large-scale biomedical corpora.
+
+# A.5 Monte Carlo Tree Search (MCTS)
+
+Monte Carlo Tree Search (MCTS) is a decision-making algorithm widely used in complex decision processes, such as games, by building a search tree and simulating outcomes to estimate the value of potential actions. MCTS operates through four main phases (Browne et al., 2012).
+
+Selection: Starting from the root node, the algorithm traverses through child nodes based on strategies like the Upper Confidence Bound applied on Trees (UCT), which balances exploration and exploitation, continuing until a leaf node is reached.
+
+Expansion: At the leaf node, if it does not represent a terminal state, one or more feasible child nodes are added to represent potential future actions.
+
+Simulation (Evaluation): From one of the newly added nodes (typically selected randomly), random simulations (or "rollouts") are performed by selecting actions randomly until reaching a terminal state, thereby estimating the node's potential.
+
+Backpropagation: After simulation, the results (win, loss, or draw) are propagated back through the traversed nodes, updating the statistical data (e.g., rewards or visit counts) to guide future decision-making.
+
+By iterating through these phases, MCTS incrementally builds a decision tree, enabling optimal strategy refinement in scenarios where direct calculation of the best strategy is infeasible due to a vast state space. A crucial component of MCTS is the Upper Confidence Bound applied on Trees (UCT) algorithm, used during the selection phase to balance exploration and exploitation. The UCT formula for choosing actions is defined as follows:
+
+$$
+\mathrm {U C T} _ {j} = \bar {X} _ {j} + C \sqrt {\frac {2 \ln N}{N _ {j}}} \tag {1}
+$$
+
+where  $\bar{X}_j$  is the average reward of action  $j$ ,  $N$  is the total visit count of the parent node,  $N_j$  is the visit count of node  $j$ , and  $C$  is a constant that controls the balance between exploration and exploitation.
+
+rStar enhanced MCTS-based exploration of candidate solutions. Specifically, rStar leverages a reward mechanism to guide tree expansion. Each node  $s$  generated under action  $a$  has a reward value  $Q(s,a)$ . Unexplored nodes are initialized with  $Q(s_{i},a_{i}) = 0$ , leading to random tree expansions initially. Upon reaching a terminal node  $s_d$ , a reward score  $Q(s_d,a_d)$  is computed based on
+
+whether the trajectory reaches the correct answer and is then back-propagated to each intermediate node in the trajectory  $t = x \oplus s_1 \oplus s_2 \oplus \ldots \oplus s_d$ . For each intermediate node  $s_i$ , the reward is updated as  $Q(s_i, a_i) = Q(s_i, a_i) + Q(s_d, a_d)$ , with  $Q(s_d, a_d)$  determined by the likelihood or confidence from self-consistency majority voting. This reward propagation directs the search toward paths with higher probabilities of correctness, enhancing decision-making within the reasoning process.
+
+# A.6 Evaluation of Reasoning Candidates with RAFS
+
+Figure 7 demonstrates how the Retrieval-Augmented Factuality Scorer (RAFS) evaluates reasoning candidates for a medical question. The example involves diagnosing and treating a patient presenting with symptoms consistent with seasonal allergic conjunctivitis. RAFS assigns factuality scores to each candidate reasoning path based on their alignment with retrieved evidence.
+
+Candidate Reasoning 1 accurately identifies the condition and treatment, correctly selecting ketotifen eye drops (Answer B) with a factuality score of 1.0. Candidate Reasoning 2 suggests fluorometholone eye drops (Answer D) but includes conflicting reasoning and inaccuracies, resulting in a lower factuality score of 0.625. Candidate Reasoning 3 incorrectly recommends warm compresses (Answer C) as the most appropriate treatment, reflecting limited factual alignment and receiving a score of 0.6.
+
+This example highlights RAFS's ability to differentiate between reasoning paths based on factual correctness, ensuring that the most reliable and evidence-supported answer is prioritized.
+
+# A.7 Prompt of Different Actions in MCTS
+
+# RAFS Examples
+
+Question: A 35-year-old man comes to the physician because of itchy, watery eyes for the past week. He has also been sneezing multiple times a day during this period. He had a similar episode 1 year ago around springtime. He has iron deficiency anemia and ankylosing spondylitis. Current medications include ferrous sulfate, artificial tear drops, and indomethacin. He works as an elementary school teacher. His vital signs are within normal limits. Visual acuity is 20/20 without correction. Physical examination shows bilateral conjunctival injection with watery discharge. The pupils are  $3\mathrm{mm}$ , equal, and reactive to light. Examination of the anterior chamber of the eye is unremarkable. Which of the following is the most appropriate treatment? A: Erythromycin ointment, B: Ketotifen eye drops, C: Warm compresses, D: Fluorometholone eye drops
+
+Candidate Reasoning 1: Let's think step by step. Based on the symptoms described - itchy, watery eyes, sneezing, bilateral conjunctival injection, and watery discharge - along with the history of a similar episode around springtime, it's clear that the patient is experiencing allergic conjunctivitis, likely triggered by seasonal allergens like pollen given his profession as an elementary school teacher. For symptomatic relief of allergic conjunctivitis, antihistamine eyedrops or mast cell stabilizers can be effective. Among the options provided, ketotifen eye drops act as a mast cell stabilizer and would be appropriate for treating allergic conjunctivitis. The answer is B: Ketotifen eye drops.  $\checkmark$  The factuality score is 1.0
+
+Candidate Reasoning 2: Based on the patient's symptoms of itchy, watery eyes, sneezing, and physical examination findings of bilateral conjunctival injection with watery discharge, the diagnosis is allergic conjunctivitis. Considering the timing of the episode and the patient's profession, the most likely causative agent is seasonal pollen. Treatment should focus on reducing symptoms rather than eliminating the allergen source. Fluorometholone eye drops are corticosteroid drops that reduce inflammation and can provide relief from itching and redness. Other options like erythromycin or ketotifen may have some effect but would be less effective in addressing this presentation. Warm compresses might help with discharge, but again, would be less effective compared to the impact of corticosteroids. Therefore, the most appropriate treatment is fluorometholone eye drops. The answer is D: Fluorometholone eye drops.  $\times$  The factuality score is 0.625
+
+Candiate Reasoning 3: Given the patient's symptoms of itchy, watery eyes, sneezing, and conjunctival injection, along with a history of similar episodes around springtime, this case is most consistent with seasonal allergic conjunctivitis. The best treatment for mild allergic conjunctivitis involves avoiding triggers when possible, using lubricating artificial tears regularly, and applying a cold compress. Warm compresses are often recommended to help loosen crusts and debris, improving comfort. Therefore, warm compresses would be the most appropriate treatment for this patient. The answer is C: Warm compresses.  $\times$  The factuality score is 0.6
+
+Figure 7: Evaluation of reasoning candidates by the Retrieval-Augmented Factuality Scorer (RAFS). Each candidate provides a reasoning path for a given question, with RAFS assigning a factuality score based on alignment with retrieved evidence.
+
+![](images/c535e467e8ca68b9d29ed6e1238e7c9154637b354cdc63ed64208c5d0243e25c.jpg)  
+Figure 8: Top 10 common paths that lead to correct answer (MedQA)
+
+![](images/daf0bcf4edb7bd28526e3d8f8d4cbe274ce1b4f5b25569e8f233c801f60b78f4.jpg)  
+Figure 9: Top 10 common paths that lead to correct answer (StrategyQA)
+
+# Multi-step reasoning in Medical QA
+
+Question: A 23-year-old pregnant woman at 22 weeks gestation presents with burning upon urination. She states it started 1 day ago and has been worsening despite drinking more water and taking cranberry extract. She otherwise feels well and is followed by a doctor for her pregnancy. Her temperature is  $97.7^{\circ}\mathrm{F}$ $(36.5^{\circ}\mathrm{C})$ , blood pressure is  $122/77\mathrm{mmHg}$ , pulse is  $80/\mathrm{min}$ , respirations are  $19/\mathrm{min}$ , and oxygen saturation is  $98\%$  on room air. Physical exam is notable for an absence of costovertebral angle tenderness and a gravid uterus. Which of the following is the best treatment for this patient? A: Ampicillin, B: Ceftriaxone, C:Ciprofloxacin, D: Doxycycline, E: Nitrofurantoin
+
+Step 1: Identify key clinical features A 23-year-old pregnant woman at 22 weeks gestation presents with burning upon urination (dysuria), worsening symptoms over one day, and no systemic signs of infection (normal temperature, blood pressure, and oxygen saturation). Physical exam shows no costoverterbral angle tenderness, suggesting the absence of pyelonephritis.
+
+Step 2: Determine the likely diagnosis Dysuria without systemic signs in a pregnant woman is most consistent with acute cystitis, a lower urinary tract infection.
+
+Step 3: Evaluate treatment options for cystitis in pregnancy Nitrofurantoin: Safe and effective during the second trimester, commonly used for uncomplicated cystitis. Ampicillin: Can be used but has high resistance rates among common uropathogens like E. coli. Ceftriaxone: Typically used for severe infections like pyelonephritis and administered intravenously, not necessary here. Ciprofloxacin: A fluoroquinolone, contraindicated in pregnancy due to potential cartilage damage to the fetus. Doxycycline: A tetracycline, contraindicated in pregnancy due to risks of fetal bone growth inhibition and teeth discoloration.
+
+Step 4: Select the best treatment Nitrofurantoin is the safest and most effective option for this patient in the second trimester.
+
+Step 5: Conclusion The best treatment for this patient is nitrofurantoin. The answer is E: Nitrofurantoin.
+
+Figure 10: An example of a medical question that requires multi-step reasoning to answer.
+
+# Action 1. Propose an one-step thought.
+
+Instruction: A 46-year-old woman comes to the physician because of a 2-week history of diplopia and ocular pain when reading the newspaper. She also has a 3-month history of amenorrhea, hot flashes, and increased sweating. She reports that she has been overweight all her adult life and is happy to have lost  $6.8\mathrm{kg}$  (15 lb) of weight in the past 2 months. Her pulse is  $110 / \mathrm{min}$ , and blood pressure is  $148 / 98\mathrm{mmHg}$ . Physical examination shows moist palms and a nontender thyroid gland that is enlarged to two times its normal size. Ophthalmologic examination shows prominence of the globes of the eyes, bilateral lid retraction, conjunctival injection, and an inability to converge the eyes. There is no pain on movement of the extraocular muscles. Visual acuity is 20/20 bilaterally. Neurologic examination shows a fine resting tremor of the hands. Deep tendon reflexes are  $3+$  with a shortened relaxation phase. Which of the following is the most likely cause of this patient's ocular complaints? A: Granulomatous inflammation of the cavernous sinus, B: Abnormal communication between the cavernous sinus and the internal carotid artery, C: Glycosaminoglycan accumulation in the orbit, D: Bacterial infection of the orbital contents, E: Sympathetic hyperactivity of levator palpebrae superioris
+
+\*\*\* Response: Let's think step by step.
+
+Step 1: Identify key clinical features: A 46-year-old woman presents with diplopia, ocular pain, weight loss, amenorrhea, and hyperthyroid signs like tachycardia, enlarged thyroid gland, and hyperreflexia.
+
+Step 2: Recognize underlying condition: These findings suggest Graves' disease with thyroid-associated orbitopathy (TAO).
+
+Step 3: Determine cause of ocular symptoms: Prominent globes, lid retraction, and conjunctival injection are caused by immune-mediated glycosaminoglycan accumulation in the orbit, characteristic of TAO.
+
+Step 4: Eliminate other options: Conditions like cavernous sinus inflammation, carotid-cavernous fistula, and bacterial infection do not fit the patient's presentation.
+
+Step 5: The most likely cause of the ocular complaints is glycosaminoglycan accumulation in the orbit.
+
+Step 6: Conclusion: The answer is C: Glycosaminoglycan accumulation in the orbit.
+
+Instruction: {user question}
+
+\*\*\* Response: Let's think step by step.
+
+# Action 2: Propose the remaining thought steps / Action 4: Answer the sub-question again.
+
+Instruction: A 1-year-old boy presents to the emergency department with weakness and a change in his behavior. His parents state that they first noticed the change in his behavior this morning and it has been getting worse. They noticed the patient was initially weak in his upper body and arms, but now he won't move his legs with as much strength or vigor as he used to. Physical exam is notable for bilateral ptosis with a sluggish pupillary response, a very weak sucking and gag reflex, and shallow respirations. The patient is currently drooling and his diaper is dry. The parents state he has not had a bowel movement in over 1 day. Which of the following is the pathophysiology of this patient's condition? A: Antibodies against postsynaptic nicotinic cholinergic ion channels, B: Autoantibodies against the presynaptic voltage-gated calcium channels, C: Autoimmune demyelination of peripheral nerves, D: Blockade of presynaptic acetylcholine release at the neuromuscular junction, E: Lower motor neuron destruction in the anterior horn
+
+\#\#\# Response: Let's think step by step. The infant has rapidly progressing weakness starting from the upper body and moving downward, bilateral ptosis, sluggish pupils, weak sucking and gag reflexes, shallow respirations, drooling, constipation (dry diaper and no bowel movement), and decreased muscle strength. These symptoms are consistent with infant botulism, caused by ingestion of Clostridium botulinum spores (often from honey) that produce botulinum toxin. The toxin inhibits the release of acetylcholine at the presynaptic terminals of the neuromuscular junction by cleaving SNARE proteins necessary for vesicle fusion. This blockade leads to flaccid paralysis and autonomic dysfunction. Therefore, the pathophysiology is blockade of presynaptic acetylcholine release at the neuromuscular junction. The answer is D: Blockade of presynaptic acetylcholine release at the neuromuscular junction.
+
+Instruction: {user question}
+
+Response:
+
+# Action 3: Propose next sub-question along with its answer.
+
+Given a question, please decompose it into sub-questions. For each sub-question, please answer it in a complete sentence, ending with "The answer is". When the original question is answerable, please start the subquestion with "Now we can answer the question: <original question>".
+
+Question 1: A 23-year-old pregnant woman at 22 weeks gestation presents with burning upon urination. She states it started 1 day ago and has been worsening despite drinking more water and taking cranberry extract. She otherwise feels well and is followed by a doctor for her pregnancy. Her temperature is  $97.7^{\circ}\mathrm{F}$ $(36.5^{\circ}\mathrm{C})$ , blood pressure is  $122/77\mathrm{mmHg}$ , pulse is  $80/\mathrm{min}$ , respirations are  $19/\mathrm{min}$ , and oxygen saturation is  $98\%$  on room air. Physical exam is notable for an absence of costovertebral angle tenderness and a gravid uterus. Which of the following is the best treatment for this patient? A: Ampicillin, B: Ceftriaxone, C: Ciprofloxacin, D: Doxycycline, E: Nitrofurantoin
+
+Question 1.1: What is the most likely diagnosis for this patient?
+
+Answer 1.1: The patient is a 23-year-old pregnant woman at 22 weeks gestation presenting with burning upon urination, known as dysuria. She has no fever, normal vital signs, and no costovertebral angle tenderness. This suggests that she is experiencing an acute uncomplicated urinary tract infection (UTI), specifically cystitis.
+
+Question 1.2: Which antibiotics are safe and effective for treating cystitis during pregnancy?
+
+Answer 1.2: During pregnancy, certain antibiotics are preferred due to their safety profiles for both the mother and the fetus. Nitrofurantoin is commonly used and considered safe during the second trimester for treating cystitis. Ampicillin may be less effective due to resistance. Ceftriaxone is typically reserved for more severe infections like pyelonephritis and is administered intravenously. Ciprofloxacin and doxycycline are contraindicated in pregnancy because they can harm fetal development.
+
+Question 1.3: Now we can answer the question: Which of the following is the best treatment for this patient?
+
+Answer 1.3: The best treatment for this pregnant patient with acute cystitis is Nitrofurantoin, as it is safe during the second trimester and effective against common urinary pathogens. Therefore, the answer is E: Nitrofurantoin.
+
+Question 2: {user question}
+
+# Action 5: Rephrase the question/sub-question.
+
+You are an AI assistant to help me rephrase questions by splitting the question context into conditions. In your rephrased question, remember to fully express the information in the original question.
+
+Original Question: A 62-year-old woman comes to the physician because of coughing and fatigue during the past 2 years. In the morning, the cough is productive of white phlegm. She becomes short of breath walking up a flight of stairs. She has hypertension and hyperlipidemia. She has recently retired from working as a nurse at a homeless shelter. She has smoked 1 pack of cigarettes daily for 40 years. Current medications include ramipril and fenofibrate. Her temperature is  $36.5^{\circ}\mathrm{C}$ $(97.7^{\circ}\mathrm{F})$ , respirations are 24/min, pulse is 85/min, and blood pressure is  $140/90\mathrm{mmHg}$ . Scattered wheezing and rhonchi are heard throughout both lung fields. There are no murmurs, rubs, or gallops but heart sounds are distant. Which of the following is the most likely underlying cause of this patient's symptoms? A: Chronic decrease in pulmonary compliance, B: Local accumulation of kinins, C: Mycobacterial invasion of pulmonary parenchyma, D: Progressive obstruction of expiratory airflow, E: Incremental loss of functional residual capacity
+
+Rephrased Question: A 62-year-old woman comes to the physician due to a 2-year history of coughing and fatigue. She has a morning cough productive of white phlegm and becomes short of breath when climbing stairs. She has hypertension and hyperlipidemia and recently retired from nursing at a homeless shelter. She has a 40-pack-year smoking history. Her medications include ramipril and fenofibrate. Vital signs are temperature  $36.5^{\circ}\mathrm{C}$ $(97.7^{\circ}\mathrm{F})$ , respirations 24/min, pulse 85/min, and blood pressure 140/90 mmHg. Physical examination reveals scattered wheezing and rhonchi throughout both lung fields. Heart sounds are distant with no murmurs, rubs, or gallops. Which of the following is the most likely underlying cause of this patient's symptoms? A: Chronic decrease in pulmonary compliance, B: Local accumulation of kinins, C: Mycobacterial invasion of pulmonary parenchyma, D: Progressive obstruction of expiratory airflow, E: Incremental loss of functional residual capacity
+
+Original question: {user question}
+
+Rephrased Question:
+
+# Action 6: Generate multiple search queries and retrieve information.
+
+Given a question, generate a search query that would help gather information to answer it. Your goal is to formulate a query that will retrieve useful evidence or additional details that contribute to answering the question. The query should aim to obtain new information and be specific enough to ensure that the search results are relevant and helpful.
+
+Question 1: A 23-year-old pregnant woman at 22 weeks gestation presents with burning upon urination. She states it started 1 day ago and has been worsening despite drinking more water and taking cranberry extract. She otherwise feels well and is followed by a doctor for her pregnancy. Her temperature is  $97.7^{\circ}\mathrm{F}$ $(36.5^{\circ}\mathrm{C})$ , blood pressure is  $122/77\mathrm{mmHg}$ , pulse is  $80/\mathrm{min}$ , respirations are  $19/\mathrm{min}$ , and oxygen saturation is  $98\%$  on room air. Physical exam is notable for an absence of costovertebral angle tenderness and a gravid uterus. Which of the following is the best treatment for this patient? A: Ampicillin, B: Ceftriaxone, C: Ciprofloxacin, D: Doxycycline, E: Nitrofurantoin
+
+Query 1.1: Common causes and management of dysuria in pregnant women without systemic symptoms.
+
+Document 1.1: Dysuria in pregnant women without systemic symptoms like fever or flank pain is often indicative of acute cystitis, a lower urinary tract infection typically caused by bacteria such as Escherichia coli. Pregnancy increases the risk of urinary tract infections due to hormonal changes that relax ureteral smooth muscle and mechanical compression from the enlarging uterus, leading to urinary stasis. Management involves confirming the diagnosis with a urine culture and initiating antibiotic therapy that is safe for use during pregnancy.
+
+Query 1.2: Antibiotics that are safe and effective for treating cystitis during pregnancy.
+
+Document 1.2: Safe and effective antibiotics for treating cystitis in pregnancy include nitrofurantoin and certain beta-lactam antibiotics like ampicillin and cephalexin. Nitrofurantoin is commonly used during the second trimester and is effective against common urinary pathogens. Ampicillin can be used but may have increased resistance rates. Ceftriaxone is a third-generation cephalosporin administered intravenously and is typically reserved for more severe infections like pyelonephritis. Antibiotics such as doxycycline (a tetracycline) and ciprofloxacin (a fluoroquinolone) are contraindicated in pregnancy due to potential risks to fetal development.
+
+Query 1.3: Antibiotics contraindicated during pregnancy and their associated risks.
+
+Document 1.3: Certain antibiotics are contraindicated during pregnancy due to teratogenic effects or adverse outcomes for the fetus. Doxycycline, a tetracycline antibiotic, is contraindicated because it can inhibit bone growth and cause permanent tooth discoloration in the developing fetus. Ciprofloxacin, a fluoroquinolone, is avoided as it has been associated with cartilage damage in animal studies and potential musculoskeletal risks in humans. These medications should not be used to treat infections in pregnant patients.
+
+Question 2: {user question}
+
+# Action 7: Re-answer the question/sub-question with retrieved information.
+
+Instruction: A 1-year-old boy presents to the emergency department with weakness and a change in his behavior. His parents state that they first noticed the change in his behavior this morning and it has been getting worse. They noticed the patient was initially weak in his upper body and arms, but now he won't move his legs with as much strength or vigor as he used to. Physical exam is notable for bilateral ptosis with a sluggish pupillary response, a very weak sucking and gag reflex, and shallow respirations. The patient is currently drooling and his diaper is dry. The parents state he has not had a bowel movement in over 1 day. Which of the following is the pathophysiology of this patient's condition? A: Antibodies against postsynaptic nicotinic cholinergic ion channels, B: Autoantibodies against the presynaptic voltage-gated calcium channels, C: Autoimmune demyelination of peripheral nerves, D: Blockade of presynaptic acetylcholine release at the neuromuscular junction, E: Lower motor neuron destruction in the anterior horn
+
+Relevant Documents:
+
+Infant Botulism: Caused by ingestion of Clostridium botulinum spores (common sources include honey and soil). Spores germinate in the infant's gut, producing botulinum toxin.
+
+Botulinum Toxin Mechanism: Blocks the release of acetylcholine at presynaptic neuromuscular junctions by cleaving SNARE proteins. Results in flaccid paralysis and autonomic dysfunction.
+
+Clinical Features: Symmetric descending paralysis starting with cranial nerves (ptosis, poor suck, weak gag reflex).
+
+Autonomic symptoms: constipation, dry mouth, drooling. Hypotonia and respiratory difficulties.
+
+\#\#\# Response: Let's think step by step. The infant presents with rapidly progressing weakness starting in the upper body and moving downward, bilateral ptosis with sluggish pupillary response, weak sucking and gag reflexes, shallow respirations, drooling, and constipation (dry diaper, no bowel movement). These symptoms are characteristic of infant botulism, caused by ingestion of Clostridium botulinum spores that produce botulinum toxin. The toxin inhibits the release of acetylcholine at the presynaptic terminals of the neuromuscular junction by cleaving SNARE proteins necessary for neurotransmitter release. This blockade leads to flaccid paralysis and autonomic dysfunction. Therefore, the pathophysiology is blockade of presynaptic acetylcholine release at the neuromuscular junction. The answer is D: Blockade of presynaptic acetylcholine release at the neuromuscular junction.
+
+Instruction: {user question}
+
+Relevant Documents: {relevant documents}
+
+Response:

rareretrievalaugmentedreasoningenhancementforlargelanguagemodels/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47262839dd59fd38600ac6944f54a3f75f106405a524abb6a91543748ad995ad
+size 727022

rareretrievalaugmentedreasoningenhancementforlargelanguagemodels/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:564e63a218d08b44da668e4f694899539de6cbe9ca395da8b2f68b698fd1b202
+size 616085

rationalystpretrainingprocesssupervisionforimprovingreasoning/0d7c1c10-3f4a-40f4-8f09-d45ff8c6bad2_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f827bb04909a45cfff59f51792df79599528061d6468be1aa63150b5a5f3bb7b
+size 135764

rationalystpretrainingprocesssupervisionforimprovingreasoning/0d7c1c10-3f4a-40f4-8f09-d45ff8c6bad2_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8cddfb7b1e5af4a92b85c26641180bf4323d26cafb4f52a599e2670f9d3a1d0
+size 159940

rationalystpretrainingprocesssupervisionforimprovingreasoning/0d7c1c10-3f4a-40f4-8f09-d45ff8c6bad2_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0063889742a355dd068cb99104e0ec71b223ee06c6c5f80c92c6b7ef76782cac
+size 1226769

rationalystpretrainingprocesssupervisionforimprovingreasoning/full.md

@@ -0,0 +1,626 @@
+![](images/d801a41780ff09311e56cbeca7825238fdf855e0a49ce1949338bfe6ec93da6d.jpg)
+
+# RATIONALYST: Pre-training Process-Supervision for Improving Reasoning
+
+Dongwei Jiang\* Guoxuan Wang\* Yining Lu\* Andrew Wang\* Jingyu Zhang\* Chuyu Liu\* Benjamin Van Durme\* Daniel Khashabi\* Johns Hopkins University, \*University of Notre Dame djiang21@jhu.edu
+
+# Abstract
+
+The reasoning steps generated by LLMs might be incomplete, as they mimic logical leaps common in everyday communication found in their pre-training data: underlying rationales are frequently left implicit (unstated). To address this challenge, we introduce RATIONALYST, a model for process-supervision of reasoning based on pre-training on a vast collection of rationale annotations extracted from unlabeled data. We extract 79k rationales from web-scale unlabelled dataset (the Pile) and a combination of reasoning datasets with minimal human intervention. This web-scale pre-training for reasoning allows RATIONALYST to consistently generalize across diverse reasoning tasks, including mathematical, commonsense, scientific, and logical reasoning. Fine-tuned from LLaMa-3-8B-Instruct, RATIONALYST improves the accuracy of reasoning by an average of  $3.9\%$  on 7 representative reasoning benchmarks. It also demonstrates superior performance compared to significantly larger verifiers like GPT-4 and similarly sized models fine-tuned on matching training sets.
+
+# 1 Introduction
+
+Rationales play a crucial role in human reasoning and its accuracy (Rips, 1994; Mercier and Sperber, 2011). In reasoning problems, having accurate rationales often correlates with accurate outcomes (Tversky et al., 1982; Davis, 1984). This importance of rationales extends to Large Language Models (LLMs) as well. Wei et al. (2022) were among the first to show that generating chain-of-thought rationales significantly improves LLMs' reasoning performance. Subsequent research has further refined the methods for eliciting rationales, leading to improved performance (Fu et al., 2023; Zhou et al., 2022).
+
+$^{1}$ Our code, data, and model can be found at this repository: https://anonymous.4open.science/r/RATIONALYST-B5CD
+
+![](images/ef81b43c1a68cdeb0642f4c4fee5743ff4dd45e9938299c576de8ec5c1340ae9.jpg)  
+Figure 1: A simplified example showing how implicit rationales in pre-training data can be leveraged to improve reasoning. ①: Implicit rationales (unstated logical connections) occur frequently in LLM pre-training data. ②: Existing LLMs pre-trained to replicate their pretraining data tend to omit these logical steps as well. ③: However, RATIONALYST learns to generate these rationales at inference time to supervise the chain-of-thought process for better reasoning.
+
+In the context of LLM reasoning, these rationales are typically employed through a chain-of-thought process that makes reasoning explicit by articulating them as plain-text rationales. In this approach, each subsequent rationale is generated conditioned on rationales produced in preceding steps, effectively using them as a form of supervision. However, the generated reasoning chains might be incomplete, containing potential logical leaps while leaving some rationales implicit (or hidden) during the generation process. These gaps in the reasoning chain can weaken the LLM's reasoning ability throughout the problem-solving process.
+
+One reason why chain-of-thought methods might miss implicit steps is that models trained with "next-token prediction" often replicate the omissions present in their training data. Implicit rationales—underlying logical connections that are often not explicitly stated—are frequently missing
+
+![](images/b71be1f435498189e8ef7a398c56fde71df8eb551ac73aa8d6f1e7295ef37712.jpg)  
+Figure 2: An example showing how RATIONALYST works at inference time. RATIONALYST generates implicit rationales given the current reasoning trajectory, which includes both the question and the reasoning steps generated so far ①. Agent LLM generates multiple next-step candidates for reasoning, also based on the current reasoning trajectory ②. Implicit rationale generated by RATIONALYST is used to provide heuristics for choosing the next step candidates proposed by the agent LLM by estimating the probability of the next step candidate given the rationale ③. The reasoning trajectory is updated iteratively with the highest scoring next step candidate ④.
+
+in daily communication and web text. Figure 1  $①$  illustrates this concept using a typical document from LLM pre-training data. In this example, we see a passage from Harry Potter: "Harry used magic outside... He is punished to attend..." The text contains the implicit (unstated) rationale: "When someone breaks the rule, he will be punished!" This implicit rationale is crucial in inferring the causal reasoning that connects the cause (Harry breaking rules) to its effect (punishment), but is also left unstated in the context. As a result, existing LLMs trained to mimic web text will have difficulty surfacing these implicit statements during the reasoning process, which can lead to flawed conclusions, such as erroneously justifying theft as a praiseworthy act when done to support one's family ( $②$  in Figure 1).
+
+This paper presents RATIONALYST, a model tailored for process-supervision of reasoning. RATIONALYST is trained on a vast collection of implicit rationales extracted from a mixture of web-scale unlabeled datasets and existing reasoning datasets. Although existing LLMs may miss crucial details in their reasoning, leading to flawed conclusions (2) in Figure 1), using RATIONALYST provides additional supervision mechanism to guide their reasoning processes, resulting in more robust conclusions (3) in Figure 1).
+
+RATIONALYST is developed and used in three stages: (1) we employ LLMs to extract implicit rationales from unlabeled text corpora without human annotation. These rationales are subsequently
+
+filtered based on their helpfulness in predicting subsequent text (§3.1); (2) we train RATIONALYST to predict those rationales given the preceding context (§3.2); and then (3) as depicted in Figure 2, during inference, we assume reasoning is done incrementally in a chain-of-thought fashion (Wei et al., 2022) by another agent model, and we use RATIONALYST to provide supervision for the agent model at each reasoning step throughout the reasoning process §3.3. By adopting a data-centric approach, RATIONALYST utilizes abundant unlabelled data to provide process supervision (Lightman et al., 2023) across various reasoning tasks without the need for human annotation.
+
+Our method extracts 65k implicit rationales from the web-scale unlabelled dataset The Pile (Gao et al., 2020). To adapt the extracted rationales to our tested domain and stabilize training, we additionally extract a much smaller set of 14k implicit rationales from the question-answer pairs in the training sets of two reasoning datasets: GSM8K (Cobbe et al., 2021a) and ECQA (Aggarwal et al., 2021). Our extraction process controls for answer leakage to prevent artificial amplification of performance. Using this curated set of rationales, RATIO-NALYST is then fine-tuned from LLaMa-3-8B. To assess the effectiveness of our approach, we evaluate RATIO-NALYST on a diverse set of reasoning tasks, including mathematical, commonsense, scientific, and logical reasoning. Our results show that RATIO-NALYST improves the accuracy of reasoning by an average of  $3.9\%$  ( $\S 5.1$ ). To understand the
+
+contribution of different data sources, we conduct an ablation study that demonstrates the utility of rationales from both the large-scale Pile dataset and the smaller, specialized reasoning datasets (§5.2). Notably, RATIONALYST exhibits superior performance when compared to strong general-purpose verifiers like GPT-4 and similar capacity models specifically fine-tuned on matching training sets (§5.4).
+
+Implicit rationales generated by RATIONALYST are also designed to provide supervision in a human-readable form, offering improved interpretability for LLM generation. This added interpretability is particularly beneficial when reasoning over complex domains such as mathematics or coding, where the step-by-step logic can be difficult for humans to follow without explicit explanations. As shown in §5.5, our model is capable of generating human-understandable rationales for unseen data from complex math reasoning.
+
+# 2 Related Work
+
+Supervising reasoning. Supervision-based approaches have been shown to enhance the reasoning abilities of LLMs. Cobbe et al. (2021b) and Snell et al. (2024) demonstrate that training a "verifier" to supervise reasoning can be more parameter-efficient than simply expanding the parameters of the "reasoner" responsible for solving the reasoning task. Ground-truth feedback from interaction with the environment is an effective form of supervision (Wang et al., 2023), but it works only in controlled environments like simulated world. General-purpose verifiers (Dhuliawala et al., 2023; Weir et al., 2024, 2023; Vacareanu et al., 2024) offer broader applicability utilizing principles like compositional reasoning. However, they don't fully capitalize on the vast amount of unlabelled data in the way a data-driven approach might. Process-based supervision (Lightman et al., 2023) offers supervision at each reasoning step rather than just at the final result. While promising, it requires substantial human annotation for the correctness of intermediate steps, making it resource-intensive. Our work aims to address these challenges by proposing a data-centric process-supervision method without the need for human annotation.
+
+Knowledge extraction from unlabelled data. LLMs are conventionally trained on extensive web data using autoregressive next-token prediction. While effective, this approach may not fully har
+
+ness the potential of the pre-training data, as latent information within this data could be better accessed using techniques beyond simple next-token prediction. Recent research has demonstrated several approaches to utilize this latent information to develop more sophisticated language model capabilities (Jiang et al., 2024c). Schick et al. (2023) introduced Toolformer, which autonomously annotates and extracts appropriate positions, names, and inputs for tool use by leveraging supervision from future tokens. Similarly, Cornille et al. (2024) developed a method for learning to plan coherent article writing through self-supervised learning in text. More closely related to our work, Zelikman et al. (2024) proposed Quiet-Star, which applied a comparable technique to uncover underlying rationales in daily communication to enhance reasoning capabilities. Our work adopts a strategy similar to Quiet-Star for extracting rationales in an unsupervised manner. However, our approach diverges in its primary objective: we aim to train a "supervisor" that can utilize these rationales to provide process supervision for any "reasoner." This focus enables us to implement a simpler and more reliable method, as we don't need to directly integrate rationale extraction with "reasoner" training. Our approach thus offers a novel perspective on leveraging latent information in language models to enhance their capabilities.
+
+Rationales as the basis for reasoning. Various studies have focused on improving the use of rationales to elicit reasoning. Fu et al. (2023) refine rationales for more effective reasoning elicitation, while Li et al. (2023) explore different approaches to leveraging rationales to enhance reasoning. Other works, such as Hwang et al. (2024), examine the verification of rationales produced by LLMs during reasoning to improve performance. Additionally, training LLMs on rationale-rich data is a common strategy for enhancing reasoning skills. As highlighted by Lewkowycz et al. (2022) and Jiang et al. (2024a), LLMs trained on science and math data tend to perform better on reasoning tasks, particularly when CoT prompting is used. In this work, we build on this foundation by using rationales as the core of our method to supervise reasoning.
+
+# 3 Building RATIONALYST
+
+We discuss the construction of RATIONALYST and its usage at inference time. First, we describe ex
+
+![](images/cc08c0e096bc92caf7f8ef49ab55cc7ec01108d9e3b18361a3d4e86e9adf5809.jpg)  
+Figure 3: We use LLMs to extract implicit rationales (enclosed by  $\langle \mathtt{BOT} \rangle$  and  $\langle \mathtt{EOT} \rangle$  in bold) that capture reasoning in unlabelled text ( $\S 3.1$  A and B). The sample at the top is taken from unlabelled web-scale pre-training datasets The Pile and the sample at the bottom is taken from existing datasets (GSM8K). These rationales are subsequently filtered based on whether they are useful for predicting future text ( $\S 3.1$  C).
+
+tracting rationales from unlabeled text (§3.1), then use them to train RATIONALYST (§3.2), and finally, employ RATIONALYST to supervise reasoning during inference (§3.3).
+
+Setup. As we will be using multiple LLMs throughout the process, we define them here:  $M_{\mathrm{Ra}}$  is the trained rationale generation model (RATIONALYST) that generates rationales and heuristics during inference.  $M_{\mathrm{Agent}}$  is a general-purpose reasoning agent that produces candidate reasoning steps and incorporates rationales during inference. We use one additional model  $M$  for initial rationale extraction, rationale filtration, and probability estimation of potential next reasoning steps during inference. These LLMs can be implemented using various state-of-the-art models, allowing for adaptability to specific research needs and computational resources.
+
+# 3.1 Large-scale Rationale Extraction
+
+Implicit rationales are often embedded in unlabelled text, reflecting natural thought processes in daily communication. Our extraction process, illustrated in Figure 3, aims to make these rationales explicit. Using an aligned language model  $M$ , we generate rationales from text and then use  $M$  to filter these rationales to retain only those that are useful, akin to the self-supervised "tool" learning approach described by Schick et al. (2023). The same  $M$  is subsequently used to train  $M_{\mathrm{Ra}}$ .
+
+Extracting rationales from pre-training data. We employ  $M$  to generate rationales from the Pile. Due to the size of this dataset, we implement a pre-filtering process to identify reasoning-rich documents by (1) computing the average semantic embedding of representative reasoning training sets
+
+using a paragraph embedding model, and (2) selecting documents from unlabelled datasets that exceed a cosine similarity threshold  $\alpha$  when compared to this average embedding. After pre-filtering, we segment the selected paragraphs into 2000-word segments and instruct  $M$  to generate rationales at the end of each sentence, using prompts with demonstrations. Detailed information on the prompts and in-context learning demonstrations used for rationale extraction can be found in Appendix B.
+
+Extracting rationales from reasoning datasets. In parallel to  $\mathbf{A}$ , we also extract rationales from existing reasoning datasets to adapt the extracted rationales to our tested domain and stabilize training. For a given reasoning dataset with pairs of questions and final answers  $D = \{(q_i, a_i)\}_{i=1}^m$ , we create a prompt  $P$  that instructs  $M$  to generate rationales for each reasoning step in the final answer  $a_i$ . The input of the prompt consists of the entire question and answer, and the output includes implicit rationales that can be inferred from the reasoning process in the answer. Consider the concrete example from existing datasets (bottom) in Figure 3. The solution involves two reasoning steps: "Natalia sold 48 / 2 = 24 clips in May" and "Natalia sold 48 + 24 = 72 chips altogether." Here, the implicit rationale that connects the first and second steps, "Now we should calculate the sum of chips in April and May," is implicit yet helpful for the prediction of the second step. These rationales are subsequently filtered and used to train RATIONALYST.
+
+C Filtering extracted rationales. Generated rationales in A and B may not always be accurate or helpful. In reasoning tasks, our objective is for the extracted rationales to effectively aid in future
+
+reasoning, which means a good rationale should enhance the likelihood of accurately predicting the following text. Let  $i$  be the position of the rationale  $r$  in the sequence  $\mathbf{x} = x_1, \ldots, x_n$ . Given a sequence of weights  $(w_k)_{k \in \mathbb{N}}$ , the cross-entropy loss for future token prediction is defined as:
+
+$$
+L _ {i} (\mathbf {r}) = - \sum_ {j = i} ^ {n} w _ {j - i} \cdot \log p _ {M} (x _ {j} \mid \mathbf {r}, x _ {1: j - 1}),
+$$
+
+where  $M$ , in a different role from its previous use, is employed to estimate the probability over tokens  $x_{i}, \ldots, x_{n}$  prefixed by preceding tokens  $x_{1:i-1}$  and rationale  $\mathbf{r}$ . The weight assigned to each future token decreases exponentially by a factor of 0.9 for each step further away it is from the rationale. We compute  $L_{i} = L_{i}(r_{i}) - L_{i}(\varepsilon)$ , where  $\varepsilon$  represents an empty rationale (i.e., predicting following tokens based only on preceding tokens). A rationale is considered helpful if it makes the prediction of future tokens easier, indicated by  $L_{i} \geq \tau_{f}$ , where  $\tau_{f}$  is a filtering threshold. We retain rationales for which adding the rationale reduces the loss by at least  $\tau_{f}$  compared to having no rationale.
+
+It's crucial to clarify two key aspects of our rationale extraction process. First, while  $M$  extracts rationales from the training sets of reasoning datasets, these training sets are not directly used as targets when training  $M_{\mathrm{Ra}}$ . Instead, the target of training is the rationale generated by  $M$ . Second, we explicitly instruct  $M$  to exclude answers from the extracted rationales. This precaution prevents answer leakage in our prompts.
+
+# 3.2 RATIONALYST Training
+
+The goal of RATIONALYST (denoted by  $M_{\mathrm{Ra}}$ ) training is to develop a model that can generate implicit rationales to guide stepwise problem-solving during inference time. For web-scale datasets like The Pile, the input context consists of a segment of text from a document.  $M_{\mathrm{Ra}}$  learns to generate an implicit rationale that can guide the prediction of the next segment of text in the document's flow. In the case of structured reasoning datasets such as GSM8K or ECQA, the input context includes the question and any preceding reasoning steps toward the answer. Here,  $M_{\mathrm{Ra}}$  learns to generate rationales that could guide the next step in the problem-solving sequence.
+
+Given the appropriate context from either source, the implicit rationales, extracted and filtered as described in §3.1, serve as the target outputs during
+
+training. The overall training objective is to minimize the per-token cross-entropy loss between the generated rationales and their ground truth values from the extracted and filtered rationales.
+
+# 3.3 Inference with the Help of RATIONALYST
+
+During inference, any general-purpose LLM (the "agent model" or  $M_{\mathrm{Agent}}$ ) can be employed for reasoning across various problems. Algorithm 1 outlines the procedure.
+
+$M_{\mathrm{Agent}}$  generates reasoning incrementally in a chain-of-thought fashion, producing multiple candidates for the next reasoning step. These steps and the question form a "reasoning trajectory"  $T$  that aims to solve the problem, which also serves as input to  $M_{\mathrm{Ra}}$ .  $M_{\mathrm{Ra}}$  then generates  $r$ , the implicit rationale (line 3) With the help of implicit rationale, we provide supervision for the next reasoning step. Two supervision methods we considered are:
+
+Implicit supervision. For this supervision,  $M_{\mathrm{Agent}}$  generates the next reasoning steps conditioned on the trajectory  $T$  (line 6). We then use  $M$  to estimate the probability of potential next reasoning steps given rationale  $r$  and reasoning trajectory  $T$  (line 13). This probability-based heuristic aligns with our rationale filtration process used during  $M_{\mathrm{Ra}}$  training: just as we identified rationales that improved the prediction of future text during filtration, here we use rationales to improve the selection of future reasoning steps. By leveraging the probability estimates as a heuristic, we can discriminate between more and less likely next steps in the reasoning process, guiding the overall trajectory towards more accurate conclusions.
+
+Explicit supervision. Another approach is to directly incorporate the implicit rationale into the generation of the next reasoning steps. This method makes the previously implicit rationale an explicit part of the reasoning process. To do that, we ask  $M_{\text{Agent}}$  to generate multiple candidate next steps by temporarily appending  $r$  to the trajectory  $T$ , and then producing potential continuations based on this augmented context (line 8). Then, we estimate the probability of candidate generations according to  $M_{\text{Agent}}$  (line 15). This approach allows  $M_{\text{Agent}}$  to make the final decision on the next reasoning step, as in normal beam search (Snell et al., 2024; Yao et al., 2023), while benefiting from the additional context provided by  $M_{\text{Ra}}$ 's rationales.
+
+Algorithm 1 Inference with RATIONALYST  
+Input: Question  $q$  RATIONALYST  $M_{\mathrm{Ra}}$  Agent model  $M_{\mathrm{Agent}}$  Probability estimation model  $M$  Functions: Heuristic function  $H(M_{\mathrm{Ra}},q,T)$  , stopping condition stop_condition() Hyperparameters: Sampling temperature  $t$  and number of sampled rationales  $N$    
+1:  $T\gets q$  Initialize reasoning trajectory as the question.   
+2: repeat   
+3:  $r\gets M_{\mathrm{Ra}}(T)$  Generate rationale given trajectory.   
+4: heuristic_list  $\equiv \emptyset$    
+5: if supervision  $\equiv =$  implicit then   
+6: next_steps  $\leftarrow M_{\mathrm{Agent}}(T)$    
+7: else if supervision  $\equiv =$  explicit then   
+8: next_steps  $\leftarrow M_{\mathrm{Agent}}(T,r)$    
+9: end if   
+10: for  $n = 1\dots N$  do   
+11:  $x\gets$  next_steps[n] Take next step generation.   
+12: if supervision  $\equiv =$  implicit then   
+13:  $h\gets M(x|T,r)$  Estimate probability of next reasoning step.   
+14: else if supervision  $\equiv =$  explicit then   
+15:  $h\gets M_{\mathrm{Agent}}(x|T,r)$    
+16: end if   
+17: heuristic_list.append(h)   
+18: end for   
+19: m_idx  $\leftarrow$  argmax(heuristic_list)   
+20:  $T\gets T +$  next_steps[m_idx] Extend trajectory with the highest scoring step.   
+21: until stop_condition(T) E.g.,the trajectory contains strings like "The final answer is:"   
+22: return  $T$
+
+After providing heuristics for the next reasoning steps, the step with the highest heuristic (line 19) is selected. The reasoning trajectory is then extended with this highest-scoring step (line 20). The reasoning process concludes when the stop condition is satisfied (line 21), which varies by dataset and often includes cues like "The final answer is:" that can be specified in system prompts for  $M_{\text{Agent}}$  for different tasks.
+
+The computational cost of  $M_{\mathrm{Ra}}$  is comparable to a normal beam search, with the only additional cost being the generation of rationales, which are typically quite short.
+
+# 4 Experimental Setup
+
+# 4.1 Setup for Training RATIONALYST
+
+Rationale extraction. As discussed in §3.1 (A), we perform pre-filtering on The Pile, an unlabelled web-scale dataset, to identify documents with extensive reasoning content before rationale extraction. This is achieved by computing the average semantic embedding from the training sets of the reasoning datasets we test, filtering documents that exceed the cosine similarity threshold  $\alpha$  of 0.3, and keeping only the documents with length under 2000 tokens to fit within LLaMa-3 models' context length. The model we used to calculate these embeddings is MPNet-base (Song et al., 2020).
+
+Following the recipe in §3.1  $\widehat{\mathbb{B}}$ , we also ex
+
+tract rationales from existing reasoning datasets. GSM8K (Cobbe et al., 2021b) and ECQA (Aggarwal et al., 2021) were selected for their complementary coverage of mathematical and commonsense reasoning, respectively. This combination ensures RATIONALYST is trained on diverse reasoning patterns, enhancing its versatility across various tasks.
+
+Rationale annotation and filtration. The model  $M$  used for rationale extraction and rationale filtering are both LLama-3-8B-Instruct (MetaAI, 2024). On GSM8K and ECQA, we manually annotated 100 pairs of {preceding_context, rationale, following_context} to determine an appropriate filtration threshold. The annotations include 50 positive and 50 negative rationale examples. Since it's straightforward to scale up the extraction of rationales from unlabelled data for filtration, we prioritize maximizing the precision of our filtered rationales, even if it means extracting fewer of them. We set the threshold  $\tau_{f}$  to ensure that  $95\%$  of the filtered rationales are accurate. On The Pile, we do not perform rationale annotation due to its diverse composition of corpora with varying characteristics. So the filter threshold  $\tau_{f}$  for the Pile is set to 0 for all of its subdomains. The results of rationale extraction and filtration on GSM8K, ECQA, and The Pile are presented in Table 1. From our extraction and filtration process, we obtained approximately 14k rationales from GSM8K and ECQA combined, and about 65k from The Pile after filtration. Detailed statistics of rationale extraction and filtration results for each dataset, including the number of rationales per document and filtration rates, can be found in Appendix H.
+
+<table><tr><td>Dataset</td><td>Subdomains</td><td colspan="2">#Docs. # Rationales</td><td>Rationales Left (%)</td><td>τf</td></tr><tr><td>GSM8K</td><td>N/A</td><td>7473</td><td>17566</td><td>19.5</td><td>1.2</td></tr><tr><td>ECQA</td><td>N/A</td><td>7600</td><td>19669</td><td>57.6</td><td>0.5</td></tr><tr><td rowspan="6">The Pile</td><td>Pile-CC</td><td>266.6K</td><td>853.2K</td><td>2.9</td><td>0</td></tr><tr><td>StackExchange</td><td>21.8K</td><td>113.6K</td><td>29.8</td><td>0</td></tr><tr><td>Github</td><td>19.9K</td><td>45.8K</td><td>2.6</td><td>0</td></tr><tr><td>HackerNews</td><td>5.8K</td><td>24.4K</td><td>9.4</td><td>0</td></tr><tr><td>PubMed Central</td><td>4.9K</td><td>18.6K</td><td>3.2</td><td>0</td></tr><tr><td>Wikipedia (en)</td><td>4.2K</td><td>23.0K</td><td>7.8</td><td>0</td></tr></table>
+
+Table 1: The statistics on rationale sampling and filtration. We provide the total number of documents and rationales before filtering, and the percentage of leftover rationales after filtering.
+
+RATIONALYST training. RATIONALYST is finetuned with LLaMa-3-8B-Instruct as the base model. We use the default hyperparameters as specified in the LLaMa-3 technical report (MetaAI, 2024) for
+
+fine-tuning. After training, we conducted manual annotation of the model's output and found that the accuracy of rationales generated on unseen test data closely matches the filtration accuracy we specified for training data through our filtration parameters.
+
+# 4.2 Setup for Evaluating RATIONALYST
+
+We evaluate RATIONALYST across diverse reasoning tasks including mathematical (GSM8K (Cobbe et al., 2021b), MATH (Hendrycks et al., 2021), commonsense (ECQA(Aggarwal et al., 2021) and HellaSwag (Zellers et al., 2019)), logical (ProofWriter (Tafjord et al., 2021)), scientific (ARC Clark et al., 2018)), and multi-task reasoning (MMLU-Pro (Wang et al., 2024b)). For inference, we use LLaMa-3-8B-Instruct as our base agent model with temperature 0.7 and topk sampling of 3. We compare RATIONALYST against both process supervision approaches (using LLaMa-3-8B-Instruct and GPT-4) and outcomebased verifiers. Complete experimental details are provided in Appendix A.
+
+# 4.3 Setup for other verifiers.
+
+To evaluate the effectiveness of RATIONALYST's process supervision, we compare it with other approaches. For process supervision with other models, we include LLaMa-3-8B-Instruct and GPT-4 in our comparison. These models are prompted to rerank partial reasoning trajectories as reasoning steps are generated. The prompts and in-context learning demonstrations used for these models on representative datasets are provided in Appendix E. For outcome supervision, we also compare with outcome-based verifiers derived from LLaMa-3-8B-Instruct. These verifiers are fine-tuned on the training sets of each reasoning dataset. Following the approach outlined by Cobbe et al. (2021b), they assess the correctness of the final prediction by directly evaluating the question and final solution. This comparison allows us to assess the performance of RATIONALYST against both process-based and outcome-based supervision methods.
+
+# 5 Empirical Results
+
+# 5.1 Main result: RATIONALYST Improves Performance on Various Tasks
+
+In this section, we train RATIONALYST using a combination of rationales extracted from GSM8K and ECQA, as well as from The Pile, as outlined in Table 1. The baseline does not use any verifier. We
+
+use implicit supervision for this experiment. The main result is shown in Table 2.
+
+<table><tr><td>Reasoning Type</td><td>Dataset</td><td>Baseline Acc.</td><td>RATIONALYST Acc.</td><td>Δ Acc.</td></tr><tr><td rowspan="2">Mathematical</td><td>GSM8K</td><td>77.6</td><td>81.6</td><td>4.0</td></tr><tr><td>Math</td><td>28.0</td><td>32.5</td><td>4.5</td></tr><tr><td rowspan="2">CommonSense</td><td>ECQA</td><td>72.6</td><td>75.2</td><td>2.6</td></tr><tr><td>HellaSwag</td><td>58.2</td><td>60.3</td><td>2.1</td></tr><tr><td>Logical</td><td>ProofWriter</td><td>86.4</td><td>90.7</td><td>4.3</td></tr><tr><td>Scientific</td><td>ARC</td><td>77.6</td><td>80.7</td><td>3.1</td></tr><tr><td>Combined</td><td>MMLU-Pro</td><td>39.6</td><td>45.3</td><td>5.7</td></tr></table>
+
+Table 2: Accuracy and absolute improvement over baseline using RATIONALYST. RATIONALYST generalizes improved performance across different reasoning tasks.
+
+Evaluation of RATIONALYST shows that training with rationales from GSM8K, ECQA, and The Pile improves performance not only on GSM8K and ECQA, but also on other reasoning tasks (e.g. scientific reasoning, logical reasoning, etc) not directly used in rationale extraction. This supports the idea that rationales can be broadly applicable across different reasoning tasks. In addition, since we use the same model (LLaMa-3-8B-Instruct) for rationale extraction, filtering, RATIONALYST training, and inference, our results do not leverage external knowledge from stronger models like LLaMa-3-70B-Instruct or GPT-4. Future work might change  $M$  to stronger models, with the expectation that higher-quality rationales will lead to better performance.
+
+# 5.2 Ablation: Web-scale Rationales Enhance Performance Across Tasks
+
+To assess the benefit of web-scale rationales, we train another model: RATIONALYST w/o Pile solely on rationales extracted from the training sets of GSM8K and ECQA. We re-ran the experiments on the same reasoning datasets using implicit supervision. The results are detailed in Table 3.
+
+We find that training the model on web-scale data results in better performance compared to training only on the rationales extracted from GSM8K and ECQA. This improvement is consistent and particularly significant on MMLU-Pro. Web-scale data likely provides exposure to more diverse reasoning types and content, including specialized knowledge, complex real-world scenarios, and interdisciplinary connections not present in the more focused datasets.
+
+<table><tr><td>Dataset</td><td>RATIONALYST</td><td>RATIONALYST (w/o Pile)</td><td>ΔAcc.</td></tr><tr><td>GSM8K</td><td>81.6</td><td>80.3</td><td>-1.3</td></tr><tr><td>Math</td><td>32.5</td><td>31.4</td><td>-1.1</td></tr><tr><td>ECQA</td><td>75.2</td><td>74.5</td><td>-0.7</td></tr><tr><td>HellaSwag</td><td>60.3</td><td>59.1</td><td>-1.2</td></tr><tr><td>ProofWriter</td><td>90.7</td><td>88.2</td><td>-2.5</td></tr><tr><td>ARC</td><td>80.7</td><td>78.8</td><td>-1.9</td></tr><tr><td>MMLU-Pro</td><td>45.3</td><td>41.2</td><td>-4.1</td></tr></table>
+
+# 5.3 Ablation: Implicit Supervision Works Better than Explicit Supervision
+
+In this section, we conduct ablation studies to test the effectiveness of different supervision methods. To isolate the impact of supervision methods and minimize confounding variables, we focus on GSM8K and ECQA as representative benchmarks for mathematical and commonsense reasoning, respectively. We train two versions of RATIONALYST: one on rationales extracted from the GSM8K training set (RATIONALYST - GSM8K) and another on rationales from the ECQA training set (RATIONALYST - ECQA). These models are used to supervise  $M_{\text{Agent}}$  during inference on their respective tasks.
+
+As shown in Table 4, implicit supervision outperforms explicit supervision. Our manual analysis revealed that implicit supervision's superior performance stems from its greater robustness to errors. When RATIONALYST generates an imperfect rationale, the probability-based heuristic used in implicit supervision can still provide useful guidance even if the rationale itself is not ideal. This approach is less likely to lead  $M_{\text{Agent}}$  to produce incorrect next steps. In contrast, explicit supervision directly incorporates potentially flawed rationales into the reasoning process, which can cause  $M_{\text{Agent}}$  to produce incorrect next steps. Essentially, implicit supervision acts as a softer guide, allowing for some imperfection in rationales, while explicit supervision more strictly adheres to potentially flawed rationales, making it more susceptible to errors.
+
+# 5.4 RATIONALYST Outperforms Other Verifiers
+
+Table 5 presents an analysis of RATIONALYST against various verifiers. Our findings reveal several insights:
+
+Table 3: An ablation study on the benefit of rationales extracted from pre-training data (The Pile). The consistent accuracy drop shows that, utilizing web-scale rationales improves performance on various reasoning datasets.  
+
+<table><tr><td>Heuristic↓ - Evaluation task→</td><td>GSM8K</td><td>ECQA</td></tr><tr><td>Implicit Supervision</td><td>80.3</td><td>74.5</td></tr><tr><td>Explicit Supervision</td><td>77.5</td><td>72.2</td></tr></table>
+
+Table 4: Comparison of implicit and explicit supervision methods on GSM8K and ECQA tasks. Implicit supervision outperforms explicit supervision due to its robustness to errors.
+
+RATIONALYST outperforms vanilla LLaMa-3-8B-Instruct using process supervision: RATIONALYST, even without leveraging The Pile dataset, outperforms process-based verifiers using vanilla LLaMa-3-8B-Instruct. A manual examination of reasoning trajectories suggests that LLaMa-3-8B-Instruct faces difficulties in reranking partial reasoning steps. This challenge likely stems from the model's struggle to differentiate among its own generated outputs, a phenomenon observed in recent studies (Jiang et al., 2024b; Huang et al., 2023).
+
+RATIONALYST shows superior process-supervision performance than much bigger models like GPT-4: We observe consistent superior performance of RATIONALYST compared to GPT-4's process supervision. We hypothesize that this advantage arises from RATIONALYST's specialized design for providing supervision, in contrast to GPT-4's general-purpose training.
+
+RATIONALYST surpasses outcome-based verifiers trained using matching data: Notably, our method surpasses the performance of finetuned outcome-based verifiers on both GSM8K and ECQA datasets, despite these verifiers being trained on matching data. We attribute this success to the richer feedback provided by process-based supervision compared to outcome-based approaches.
+
+We would also like to mention that methods like self-consistency and other prompting techniques (e.g. STEP-BACK Prompting (Zheng et al., 2024)) are complementary to our approach. RATIONALYST can be integrated with these methods to potentially achieve even better results. For more detail, please refer to Appendix I.
+
+# 5.5 RATIONALYST Generates Accurate and Easy-to-understand Rationals
+
+We annotate some samples from the test set of Math (Hendrycks et al., 2021) at inference time, which was not part of the rationale sampling datasets. Through manual observation, we find that our
+
+<table><tr><td>Supervision</td><td>GSM8K</td><td>ECQA</td></tr><tr><td>N/A</td><td>77.6</td><td>72.6</td></tr><tr><td>Process Supervision w/ LLaMa-3</td><td>77.4</td><td>71.5</td></tr><tr><td>Process Supervision w/ GPT-4</td><td>80.0</td><td>74.7</td></tr><tr><td>Outcome Supervision w/ LLaMa-3 + FT</td><td>79.2</td><td>74.3</td></tr><tr><td>RATIONALYST w/o Pile</td><td>80.3</td><td>74.5</td></tr><tr><td>RATIONALYST</td><td>81.6</td><td>76.2</td></tr></table>
+
+Table 5: Comparison of different supervision methods. RATIONALYST outperforms both strong verifiers like GPT-4 and similarly-sized models fine-tuned on matching training data.
+
+model can generate useful rationales that is helpful for understanding LLM's reasoning process on Math (an example is provided in Appendix D). Comparing the rationales generated by RATIONALYST with those generated by Quiet-Star (Zelikman et al., 2024) on the same problems, we find that our method produces more human-understandable rationales. We believe this happens because Quiet-Star optimizes rationales during training using the accuracy of the final prediction as a reward. This approach, while effective for improving task performance, does not explicitly prioritize human interpretability. In addition, this approach might inadvertently develop shortcuts or non-intuitive patterns that optimize for accuracy but not necessarily for clarity or human understanding.
+
+# 6 Conclusion and Future Work
+
+In this paper, we introduced RATIONALYST, a novel self-supervised model designed to enhance the reasoning capabilities of LLMs by leveraging hidden rationales extracted from unlabeled text. While our results demonstrate the effectiveness of this approach, there are several promising directions for future work, including scaling to larger models and datasets, integrating with test-time compute optimization techniques, and exploring the role of implicit rationales in model training. We discuss these future directions in detail in Appendix F. Through this work, we have demonstrated that extracting and utilizing implicit rationales from unlabeled text can meaningfully improve LLMs' reasoning abilities. We hope these insights will inspire further research into leveraging latent knowledge in pre-training data to enhance model capabilities.
+
+# 7 Limitations
+
+One limitation of this work is the comprehensiveness of our experiments. In future research, we plan to extend our experiments to a broader range of reasoning tasks and compare RATIONALYST with
+
+other outcome-based and process-based verifiers. We also plan to adjust the combination of rationales used to train RATIONALYST by (1) sampling from different reasoning tasks and (2) altering the mix of rationales in unlabelled web-scale pre-training data to better understand its generalizability. An additional limitation is that we could potentially improve performance through preference tuning (e.g., DPO) where the model learns to distinguish between valid and invalid rationales, but we have not explored this direction.
+
+# References
+
+Shourya Aggarwal, Divyanshu Mandowara, Vishwajeet Agrawal, Dinesh Khandelwal, Parag Singla, and Dinesh Garg. 2021. *Explanations for Common-senseQA: New Dataset and Models*. In *Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics* and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), Online. Association for Computational Linguistics.  
+Peter Clark, Isaac Cowhey, Oren Etzioni, Tushar Khot, Ashish Sabharwal, Carissa Schoenick, and Oyvind Tafjord. 2018. Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge. arXiv preprint arXiv:1803.05457.  
+Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021a. Training verifiers to solve math word problems. Preprint, arXiv:2110.14168.  
+Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021b. Training verifiers to solve math word problems.  
+Nathan Cornille, Marie-Francine Moens, and Florian Mai. 2024. Learning to plan for language modeling from unlabeled data. Preprint, arXiv:2404.00614.  
+Randall Davis. 1984. Diagnostic reasoning based on structure and behavior. Artificial intelligence, 24(1-3):347-410.  
+Shehzaad Dhuliawala, Mojtaba Komeili, Jing Xu, Roberta Raileanu, Xian Li, Asli Celikyilmaz, and Jason Weston. 2023. Chain-of-verification reduces hallucination in large language models. Preprint, arXiv:2309.11495.  
+Yao Fu, Hao Peng, Ashish Sabharwal, Peter Clark, and Tushar Khot. 2023. Complexity-based prompting for multi-step reasoning. In ICLR. OpenReview.net.  
+Leo Gao, Stella Biderman, Sid Black, Laurence Golding, Travis Hoppe, Charles Foster, Jason Phang, Horace He, Anish Thite, Noa Nabeshima, Shawn Presser, and Connor Leahy. 2020. The Pile: An 800gb dataset of diverse text for language modeling. arXiv preprint arXiv:2101.00027.  
+Suriya Gunasekar, Yi Zhang, Jyoti Aneja, Caio Cesar Teodoro Mendes, Allie Del Giorno, Sivakanth Gopi, Mojan Javaheripi, Piero Kauffmann, Gustavo de Rosa, Olli Saarikivi, Adil Salim, Shital Shah, Harkirat Singh Behl, Xin Wang, Sebastien Bubeck, Ronen Eldan, Adam Tauman Kalai, Yin Tat Lee, and Yuanzhi Li. 2023. Textbooks are all you need. Preprint, arXiv:2306.11644.
+
+Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021. Measuring Mathematical Problem Solving With the MATH Dataset. In NeurIPS, Menlo Park, Calif. AAAI Press.  
+Jie Huang, Xinyun Chen, Swaroop Mishra, Huaixiu Steven Zheng, Adams Wei Yu, Xinying Song, and Denny Zhou. 2023. Large language models cannot self-correct reasoning yet.  
+Hyeonbin Hwang, Doyoung Kim, Seungone Kim, Seonghyeon Ye, and Minjoon Seo. 2024. Self-expllore to avoid the pit: Improving the reasoning capabilities of language models with fine-grained rewards. Preprint, arXiv:2404.10346.  
+Dongwei Jiang, Marcio Fonseca, and Shay B. Cohen. 2024a. Leanreasoner: Boosting complex logical reasoning with lean. Preprint, arXiv:2403.13312.  
+Dongwei Jiang, Jingyu Zhang, Orion Weller, Nathaniel Weir, Benjamin Van Durme, and Daniel Khashabi. 2024b. Self-[in]correct: Llms struggle with refining self-generated responses. Preprint, arXiv:2404.04298.  
+Zhengping Jiang, Yining Lu, Hanjie Chen, Daniel Khashabi, Benjamin Van Durme, and Anqi Liu. 2024c. RORA: Robust free-text rationale evaluation. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1070-1087, Bangkok, Thailand. Association for Computational Linguistics.  
+Aitor Lewkowycz, Anders Andreassen, David Dohan, Ethan Dyer, Henryk Michalewski, Vinay V. Ramasesh, Ambrose Slone, Cem Anil, Imanol Schlag, Theo Gutman-Solo, Yuhuai Wu, Behnam Neyshabur, Guy Gur-Ari, and Vedant Misra. 2022. Solving quantitative reasoning problems with language models. In NeurIPS.  
+Yifei Li, Zeqi Lin, Shizhuo Zhang, Qiang Fu, Bei Chen, Jian-Guang Lou, and Weizhu Chen. 2023. Making large language models better reasoners with step-aware verifier. Preprint, arXiv:2206.02336.  
+Hunter Lightman, Vineet Kosaraju, Yura Burda, Harri Edwards, Bowen Baker, Teddy Lee, Jan Leike, John Schulman, Ilya Sutskever, and Karl Cobbe. 2023. Let's verify step by step. Preprint, arXiv:2305.20050.  
+Liangchen Luo, Yinxiao Liu, Rosanne Liu, Samrat Phatale, Harsh Lara, Yunxuan Li, Lei Shu, Yun Zhu, Lei Meng, Jiao Sun, and Abhinav Rastogi. 2024. Improve mathematical reasoning in language models by automated process supervision. Preprint, arXiv:2406.06592.  
+Hugo Mercier and Dan Sperber. 2011. Why do humans reason? arguments for an argumentative theory. Behavioral and brain sciences, 34(2):57-74.
+
+MetaAI. 2024. Introducing meta llama 3: The most capable openly available llm to date. https://ai.meta.com/blog/meta-llama-3/.  
+Keiran Paster, Marco Dos Santos, Zhangir Azerbayev, and Jimmy Ba. 2023. Openwebmath: An open dataset of high-quality mathematical web text. Preprint, arXiv:2310.06786.  
+Lance J Rips. 1994. The psychology of proof: Deductive reasoning in human thinking. Mit Press.  
+Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, Maria Lomeli, Luke Zettlemoyer, Nicola Cancedda, and Thomas Scialom. 2023. Toolformer: Language models can teach themselves to use tools. arXiv preprint arXiv:2302.04761.  
+Charlie Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. 2024. Scaling llm test-time compute optimally can be more effective than scaling model parameters. Preprint, arXiv:2408.03314.  
+Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, and TieYan Liu. 2020. Mpnet: Masked and permuted pre-training for language understanding. Preprint, arXiv:2004.09297.  
+Oyvind Tafjord, Bhavana Dalvi, and Peter Clark. 2021. Proofwriter: Generating implications, proofs, and abductive statements over natural language. In ACL.  
+Amos Tversky, Daniel Kahneman, and Paul Slovic. 1982. Judgment under uncertainty: Heuristics and biases. Cambridge.  
+Robert Vacareanu, Anurag Pratik, Evangelia Spiliopoulou, Zheng Qi, Giovanni Paolini, Neha Anna John, Jie Ma, Yassine Benajiba, and Miguel Ballesteros. 2024. General purpose verification for chain of thought prompting. Preprint, arXiv:2405.00204.  
+Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi Fan, and Anima Anandkumar. 2023. Voyager: An open-ended embodied agent with large language models. Preprint, arXiv:2305.16291.  
+Peiyi Wang, Lei Li, Zhihong Shao, R. X. Xu, Damai Dai, Yifei Li, Deli Chen, Y. Wu, and Zhifang Sui. 2024a. Math-shepherd: Verify and reinforce llms step-by-step without human annotations. Preprint, arXiv:2312.08935.  
+Yubo Wang, Xueguang Ma, Ge Zhang, Yuansheng Ni, Abhranil Chandra, Shiguang Guo, Weiming Ren, Aaran Arulraj, Xuan He, Ziyan Jiang, Tianle Li, Max Ku, Kai Wang, Alex Zhuang, Rongqi Fan, Xiang Yue, and Wenhu Chen. 2024b. Mmlu-pro: A more robust and challenging multi-task language understanding benchmark. Preprint, arXiv:2406.01574.  
+Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits
+
+reasoning in large language models. Advances in Neural Information Processing Systems (NeurIPS), 35:24824-24837.  
+Nathaniel Weir, Peter Clark, and Benjamin Van Durme. 2023. Nellie: A neuro-symbolic inference engine for grounded, compositional, and explainable reasoning. Preprint, arXiv:2209.07662.  
+Nathaniel Weir, Kate Sanders, Orion Weller, Shreya Sharma, Dongwei Jiang, Zhengping Jiang, Bhavana Dalvi Mishra, Oyvind Tafjord, Peter Jansen, Peter Clark, and Benjamin Van Durme. 2024. Enhancing systematic decompositional natural language inference using informal logic. Preprint, arXiv:2402.14798.  
+Yangzhen Wu, Zhiqing Sun, Shanda Li, Sean Welleck, and Yiming Yang. 2024. An empirical analysis of compute-optimal inference for problem-solving with language models. Preprint, arXiv:2408.00724.  
+Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, and Karthik Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. In Advances in Neural Information Processing Systems (NeurIPS).  
+Eric Zelikman, Georges Harik, Yijia Shao, Varuna Jayasiri, Nick Haber, and Noah D. Goodman. 2024. Quiet-star: Language models can teach themselves to think before speaking. Preprint, arXiv:2403.09629.  
+Rowan Zellers, Ari Holtzman, Yonatan Bisk, Ali Farhadi, and Yejin Choi. 2019. HellaSwag: Can a machine really finish your sentence? In ACL.  
+Huaixiu Steven Zheng, Swaroop Mishra, Xinyun Chen, Heng-Tze Cheng, Ed H. Chi, Quoc V Le, and Denny Zhou. 2024. Take a step back: Evoking reasoning via abstraction in large language models. Preprint, arXiv:2310.06117.  
+Fan Zhou, Haoyu Dong, Qian Liu, Zhoujun Cheng, Shi Han, and Dongmei Zhang. 2022. Reflection of thought: Inversely eliciting numerical reasoning in language models via solving linear systems. Preprint, arXiv:2210.05075.
+
+# A Evaluation Setup Details
+
+# A.1 Task Configuration and Metrics
+
+We evaluate RATIONALYST on a diverse set of reasoning tasks, carefully selected to assess performance across different reasoning domains. Table 6 provides a comprehensive overview of our evaluation configuration.
+
+<table><tr><td>Task</td><td>#Eval</td><td>#Shots</td><td>Reasoning Type</td></tr><tr><td>GSM8K</td><td>1319</td><td>8</td><td>Math</td></tr><tr><td>Math</td><td>5000</td><td>5</td><td>Math</td></tr><tr><td>ECQA</td><td>17944</td><td>6</td><td>CommonSense</td></tr><tr><td>HellaSwag</td><td>10000</td><td>4</td><td>CommonSense</td></tr><tr><td>ProofWriter</td><td>600</td><td>2</td><td>Logical</td></tr><tr><td>ARC</td><td>1172</td><td>4</td><td>Scientific</td></tr><tr><td>MMLU-Pro</td><td>12000</td><td>5</td><td>Mixed</td></tr></table>
+
+Table 6: Detailed configuration of evaluation tasks. For each dataset, we report the training set size (#Train), evaluation set size (#Eval), and number of few-shot demonstrations (#Shots) used during evaluation. We use exact match accuracy across all tasks, following standard evaluation protocols.
+
+For each dataset, we implement specific evaluation protocols:
+
+- GSM8K: We evaluate on the standard test split, using 8-shot demonstrations that showcase step-by-step mathematical reasoning.  
+- MATH: Our evaluation encompasses all difficulty levels and mathematical topics in the dataset, utilizing 5-shot demonstrations calibrated for complexity.  
+- ECQA: We use the validation split due to evaluation server constraints for the test set, maintaining consistency with prior work.  
+- HellaSwag: Evaluation follows the original setup with 4-shot demonstrations focusing on commonsense inference.  
+ProofWriter: We specifically evaluate on proofs with depth more than 5 to assess complex logical reasoning capabilities.  
+- ARC: Our evaluation focuses on the more challenging ARC-Challenge subset, employing 4-shot demonstrations.  
+- MMLU-Pro: Following the original paper's protocol, we evaluate across all reasoning categories using 5-shot demonstrations.
+
+# A.2 Inference Configuration
+
+Our inference setup utilizes LLaMa-3-8B-Instruct as the base agent model ( $M_{\text{Agent}}$ ). Key configuration parameters include:
+
+Sampling Parameters. We employ temperature  $= 0.7$  and top-k  $= 3$  during inference. This configuration allows for diverse reasoning path exploration while maintaining coherence. The temperature setting represents a calculated trade-off between deterministic output (temperature  $= 0$ ) and exploration, showing particular effectiveness on logical reasoning tasks like ProofWriter.
+
+Chain-of-Thought Implementation. For each task category, we implement specialized chain-of-thought prompting:
+
+- Mathematical and Procedural Tasks: Demonstrations break down solutions into atomic reasoning steps, with explicit intermediate calculations.  
+- Multiple-Choice Tasks: Prompts guide systematic analysis of each answer choice, encouraging comparative reasoning.  
+- Logical Reasoning: Step-by-step deductive reasoning chains are demonstrated, emphasizing premise utilization.
+
+# B Prompts used for rationale sampling
+
+In this section, we provide the prompts we used for rationale sampling on GSM8K (Figure 4), ECQA (Figure 5), and The Pile (Figure 6).
+
+# C Prompts used during inference
+
+In this section, we provide the prompts used during inference time to encourage the agent model reason step by step for GSM8K (Figure 7) and ECQA (Figure 8). Note that the input to the agent model appends the last rationale generated by the agent model.
+
+# D Examples of rationales generated at inference time
+
+In this section, we provide rationales generated by RATIONALYST from the test set of MATH during inference time Figure 9, which was not part of the rationale sampling datasets, and observe that our model can still generate useful rationales that help to understand LLM's reasoning process.
+
+# E Prompts used for LLaMa-3 reranking
+
+In this section, we provide the prompts and in-context-learning demonstrations used to instruct
+
+# System Prompt:
+
+Your task is to add rationals to a piece of text. The rationals should help you with predicting future text. You can add rationals by writing "<BOT>rational<EOT>". Here are some examples of rationale generation:
+
+# Example Input 1:
+
+Question: Michael had 58 golf balls. On tueyday, he lost 23 golf balls. On wednesday, he lost 2 more. How many golf balls did he have at the end of wednesday?
+
+Answer: Michael started with 58 golf balls. After losing 23 on tueyday, he had  $58 - 23 = 35$  . After losing 2 more, he had 35-2= 33 golf balls. The answer is 33
+
+# Example Output 1:
+
+Question: Michael had 58 golf balls. On tueyday, he lost 23 golf balls. On wednesday, he lost 2 more. How many golf balls did he have at the end of wednesday?
+
+Answer: <BOT>First, we need to calculate how many golf balls Michael had after losing 23 on tuea<EOt> Michael started with 58 golf balls. After losing 23 on tueaay, he had 58 - 23 = 35. <BOT>Afer losing 2 more on wednesday, we need to calculate how many golf balls he had left<EOt> After losing 2 more, he had 35 - 2 = 33 golf balls. <BOT>We are ready to output the final answer<EOt> The answer is 33
+
+# Example Input 2:
+
+Question: Brennan was researching his school project and had to download files from the internet to his computer to use for reference. After downloading 800 files, he deleted  $70\%$  of them because they were not helpful. He downloaded 400 more files but again realized that 3/5 of them were irrelevant. How many valuable files was he left with after deleting the unrelated files he downloaded in the second round?
+
+Answer: The number of non-valuable files Brennan downloaded in the first round is  $70 / 100 * 800 = <<70 / 100 * 800 = 560 >> 560$  files. The number of valuable files Brennan downloaded in the first round is  $800 - 560 = <<800 - 560 = 240 >> 240$  When he downloaded 400 new files, there were  $3 / 5 * 400 = <<3 / 5 * 400 = 240 >> 240$  non-useful files, which he deleted again. The total number of valuable files he downloaded in the second round is  $400 - 240 = <<400 - 240 = 160 >> 160$  To write his research, Brennan had  $160 + 240 = <<160 + 240 = 400 >> 400$  useful files to reference to write his research. The answer is 400
+
+# Example Output 2:
+
+Question: Brennan was researching his school project and had to download files from the internet to his computer to use for reference. After downloading 800 files, he deleted  $70\%$  of them because they were not helpful. He downloaded 400 more files but again realized that 3/5 of them were irrelevant. How many valuable files was he left with after deleting the unrelated files he downloaded in the second round?
+
+Answer: <BOT>First, we need to calculate how many non-valuable files Brennan downloaded in the first round <EOT> The number of non-valuable files Brennan downloaded in the first round is 70/100*800 = <<70/100*800=560>>560 files. <BOT>Next, we need to calculate how many valuable files Brennan downloaded in the first round <EOT> The number of valuable files Brennan downloaded in the first round is 800-560 = <<800-560=240>>240 <BOT>After downloading 400 new files, we need to calculate how many non-valuable files he downloaded <EOT> When he downloaded 400 new files, there were 3/5*400 = <<3/5*400=240>>240 non-useful files, which he deleted again. <BOT>Finally, we need to calculate how many valuable files he was left with <EOT> The total number of valuable files he downloaded in the second round is 400-240 = <<400-240=160>>160 <BOT>Now we need to calculate the total number of valuable files he has left to write his research <EOT> To write his research, Brennan had 160+240 = <<160+240=400>>400 useful files to reference to write his research. <BOT>We are ready to output the final answer <EOT> The answer is 400
+
+Figure 4: The prompt and in-context learning examples used for sampling rationales for GSM8K. The bolded rationales represent implicit rationales in the document.
+
+LLaMa-3-8B-Instruct and GPT-4 to provide feedback by directly reranking partial reasoning traces given the question (Figure 10).
+
+# F Discussion
+
+Scaling up RATIONALYST. Scaling RATIONALYST with stronger models and increased computational resources is a logical next step. Utilizing stronger models, such as LLaMa-3-70B or GPT4, would enhance the quality of extracted rationales, improve filtration accuracy, and ultimately strengthen RATIONALYST. However, due to computational constraints, we have not pursued this, which remains a limitation of this paper. Additionally, using larger unlabelled datasets with more extensive reasoning content, such as OpenWebMath (Paster et al., 2023), is currently infeasible due to the significant computational and time requirements for pre-filtering and training. These enhancements are planned for future work.
+
+Connection to research on scaling test-time compute. Recent research has focused on extending computational resources at test-time (Snell et al., 2024; Wu et al., 2024), particularly for complex reasoning tasks. In our experiments, we focus on developing heuristics and employ a straightforward approach of sampling multiple candidates and reranking them based on RATIONALYST's guidance. However, RATIONALYST's framework is compatible with more sophisticated test-time compute techniques. Its heuristics can be integrated into existing algorithms like beam-search or lookahead search, potentially enhancing their performance without significantly increasing computational cost.
+
+Is training on extracted rationales necessary? In our approach, we first select a subset of unlabeled data that contains strong reasoning signals, then extract implicit rationales from this data for model fine-tuning. While it has been demonstrated
+
+# System Prompt:
+
+Your task is to add rationals to a piece of text. The rationals should help you with predicting future text. You can add rationals by writing "<BOT>rational"EOT". Here are some examples of rationale generation:
+
+# Example Input 1:
+
+Question: He came across a raw item but his pack was full, he had to abandon something if he was to what the item?  
+Choices: A - join, B - acquire, C - engage, D - maintain, E - remit  
+Answer: Acquiring an item requires one to have space to carry it. As he had box full of items, he had to abandon one of them in order to acquire new one. Join and Engage is not related to item. Maintain an item does not require one to abandon existing item. One cannot remit raw item. The answer is B - acquire.
+
+# Example Output 1:
+
+Question: He came across a raw item but his pack was full, he had to abandon something if he was to what the item?  
+Choices: A - join, B - acquire, C - engage, D - maintain, E - remit  
+Answer: Acquiring an item requires one to have space to carry it. <BOT>We also need to establish relationship between acquire and the question itself<ET> As he had box full of items, he had to abandon one of them in order to acquire new one. <BOT>Let's check whether other answer choices are related to the word item in question<ET> Join and Engage is not related to item. <BOT>Let's check whether other answer choices are contradictory to the word abandon in question<ET> Maintain an item does not require one to abandon existing item. <BOT>Let's check whether other answer choices are related to the word raw item in question<ET> One cannot remit raw item. <BOT>We are ready to make final prediction<ET> The answer is B - acquire.
+
+# Example Input 2:
+
+Question: Where is aberdeen in the US located?  
+Choices: A - washington, B - europe, C - scotland, D - maryland, E - south dakota  
+Answer: Aberdeen is located in Washington state which is in US. Aberdeen is also located in Scotland which is part of Europe. However, Scotland or Europe are not inside US. Aberdeen is not located in Maryland and South Dakota states. The answer is A - washington.
+
+# Example Output 2:
+
+Question: Where is aberdeen in the US located?  
+Choices: A - washington, B - europe, C - scotland, D - maryland, E - south dakota  
+Answer: Aberdeen is located in Washington state which is in US. <BOT>We need to think whether there is another aberdeen with the same name<EOT> Aberdeen is also located in Scotland which is part of Europe. <BOT>The answer choice need to be in the U.S. as mentioned in the question<EOT> However, Scotland or Europe are not inside US. <BOT>Let's check whether aberdeen is located in other answer choices<EOT> Aberdeen is not located in Maryland and South Dakota states. <BOT>We are ready to make final prediction<EOT> The answer is A - washington.
+
+Figure 5: The prompt and in-context learning examples used for sampling rationales for ECQA. The bolded rationales represent implicit rationales in the document.
+
+that training on data with robust reasoning signals can enhance reasoning capabilities on its own (Gunasekar et al., 2023; Jiang et al., 2024a), we believe our method offers additional performance benefits for two reasons. First, many language models have already been trained on datasets like The Pile. The value of fine-tuning on previously encountered text is likely lower than the value of fine-tuning on newly incorporated rationales. Second, implicit rationales encapsulate the reasoning process. Pre-training on these rationales enhances reasoning more effectively than focusing on the whole document.
+
+# G Extended Related Work
+
+Supervising reasoning. Supervision-based approaches have been shown to enhance the reasoning abilities of LLMs. Cobbe et al. (2021b) and Snell et al. (2024) demonstrate that training a "verifier" to supervise reasoning can be more parameter-efficient than simply expanding the parameters of the "reasoner" responsible for solving the reasoning task. Ground-truth feedback from interaction with the environment is an effective form of supervision
+
+(Wang et al., 2023), but it works only in controlled environments like simulated world. Some verifiers employ principles like compositional reasoning to validate the reasoning process (Dhuliawala et al., 2023; Weir et al., 2024, 2023; Vacareanu et al., 2024), these general-purpose approaches do not fully leverage the vast amount of available unlabeled data that a data-driven approach could utilize. Process-based supervision (Lightman et al., 2023; Luo et al., 2024; Wang et al., 2024a) offers supervision at each reasoning step rather than just at the final result. While promising, it requires substantial human annotation for the correctness of intermediate steps or the existence of ground-truth answers for data collection. Our work aims to address these challenges by proposing a data-centric process-supervision method without the need for human annotation.
+
+Knowledge extraction from unlabelled data. LLMs are conventionally trained on extensive web data using autoregressive next-token prediction. While effective, this approach may not fully harness the potential of the pre-training data, as latent information within this data could be better ac
+
+cessed using techniques beyond simple next-token prediction. Recent research has demonstrated several approaches to utilize this latent information to develop more sophisticated language model capabilities. Schick et al. (2023) introduced Toolformer, which autonomously annotates and extracts appropriate positions, names, and inputs for tool use by leveraging supervision from future tokens. Similarly, Cornille et al. (2024) developed a method for learning to plan coherent article writing through self-supervised learning in text. More closely related to our work, Zelikman et al. (2024) proposed Quiet-Star, which applied a comparable technique to uncover underlying rationales in daily communication to enhance reasoning capabilities. Our work adopts a strategy similar to Quiet-Star for extracting rationales in an unsupervised manner. However, our approach diverges in its primary objective: we aim to train a "supervisor" that can utilize these rationales to provide process supervision for any "reasoner." This focus enables us to implement a simpler and more reliable method, as we don't need to directly integrate rationale extraction with "reasoner" training. Our approach thus offers a novel perspective on leveraging latent information in language models to enhance their capabilities.
+
+Rationales as the basis for reasoning. Various studies have focused on improving the use of rationales to elicit reasoning. Fu et al. (2023) refine rationales for more effective reasoning elicitation, while Li et al. (2023) explore different approaches to leveraging rationales to enhance reasoning. Other works, such as Hwang et al. (2024), examine the verification of rationales produced by LLMs during reasoning to improve performance. Additionally, training LLMs on rationale-rich data is a common strategy for enhancing reasoning skills. As highlighted by Lewkowycz et al. (2022) and Jiang et al. (2024a), LLMs trained on science and math data tend to perform better on reasoning tasks, particularly when CoT prompting is used. In this work, we build on this foundation by using rationales as the core of our method to supervise reasoning.
+
+# H The Resulting Data from Extraction/Filteration.
+
+On GSM8K, our method generates an average of 2.34 rationales per document, while on ECQA, it generates 2.58 rationales per document. The filtration process removes  $80.5\%$  of the generated
+
+rationales on GSM8K and  $42.4\%$  on ECQA.
+
+For The Pile, we report the number of rationales per document and the number after filtration for each subdomain. The Pile's documents, being longer than those in GSM8K and ECQA, yield a higher average number of rationales per document. Among the subdomains, StackExchange retains the highest percentage of rationales, likely due to its question-answering format aligning well with our reasoning tasks and containing more inherent reasoning. However, The Pile as a whole contains less reasoning content, making rationale extraction challenging. Setting the threshold to 0 accepts all rationales more helpful than not having them, but the yield remains low. A manual review shows that most filtered rationales describe the preceding context rather than guiding future reasoning.
+
+In total, we extracted approximately 14k rationales from GSM8K and ECQA combined, and about 65k from The Pile after filtration.
+
+# I RATIONALYST in Combination with General Methods for Reasoning Enhancement
+
+We have conducted additional experiments focusing on GSM8K and Math datasets to verify our claim. We specifically chose these arithmetic reasoning benchmarks over multiple-choice questions, as using self-consistency on multiple-choice tasks could artificially inflate performance. Using 64 generations for self-consistency, we observed improved accuracy on both benchmarks with the help of RATIONALYST.
+
+# System Prompt:
+
+Your task is to add rationals to a piece of text. The rationals should help you with predicting future text. You can add rationals by writing "<BOT>rational"EOT>". Here are one example of rationale generation:
+
+# Example Input:
+
+\n\nNot applicable.\n\nJZ, DY and XS are the attending doctors of this patient. QS is responsible for Pathological results and WW is for Microbiological results. YS and XS is responsible for revision of the manuscript. All authors read and approved the final manuscript.\nAuthors' information
+
+{FPar1}n============================================================================================================================================================================================================&& physicians-in-training, and Yi Shi, Qunli Shi, Weiping Wang and Xin Su are attending specialist physicians who dedicate their time to mentoring trainees
+
+# Example Output:
+
+\n\nNot applicable..n\nJZ takes responsibility for drafting the manuscript. Indicates JZ's role, suggesting the next roles to be described will involve other tasks like attending doctors and result analysis. JZ, DY and XS are the attending doctors of this patient. QS is responsible for Pathological results and WW is for Microbiological results. Following the pattern, specific tasks related to results analysis are now described, predicting further details on manuscript revision. YS and XS is responsible for revision of the manuscript. The revision responsibility follows naturally after result analysis, likely leading to a final approval statement. All authors read and approved the final manuscript. Final approval statement indicates the roles and tasks, suggesting a summary or author information follows.  $\backslash$  Authors' information  $\backslash$  Transition to detailed author information section, predicting descriptions of each author's background
+
+\n===================================\n\n<BOT>Starting with the roles of resident physicians-in-training, predicting subsequent details about attending specialist physicians<EOT> and Yi Shi, Qunli Shi, Weiping Wang and Xin Su are attending specialist physicians who dedicate their time to mentoring trainees
+
+# Example Input:
+
+Suppose  $-s^*w = -26^*w$ . Suppose  $-15 = -5*j - w^j$ . Let  $g(b) = b**3 - 3*b**2 - 2*b + 3$ . What is  $g(j)?\backslash n-3\backslash nLet o(w) = -9 - 31214^*w + 31220^*w - 3$ . Determine  $o(1)$ .  $\backslash n-6\backslash nLet y = -11 + 15$ . Suppose  $-4*c + 30 = -y^*r + 50$ ,  $\theta = -5*c - r - 25$ . Let  $q(d) = d + 5$ . Give  $q(c)$ .  $\backslash n\theta\backslash nLet r = 2899 - 2906$ . Let  $i(s) = 2*s^*2 + 12*s - 3$ . Calculate  $i(r)$ .  $\backslash n11\backslash nLet q$  be  $6/(-105) + 4199/595$ . Let  $m(t) = t**3 - 9*t**2 + 5*t + 5$ . Give  $m(q)$ .  $\backslash n-58\backslash nLet d = 440 - 92$
+
+# Example Output:
+
+Suppose  $-5\ast w = -26\ast w$  Suppose  $-15 = -5\ast j - w\ast j$  . Let  $g(b) = b\ast 3 - 3\ast b\ast 2 - 2\ast b + 3$  . What is g(j)?\\n<BOT>First solve the equations to find j, then substitute j into g(b) to calculate g(j)<EOT>\n-3\nLet o(w) = -9 - 31214\*w + 31220\*w - 3. Determine o(1).\\n<BOT>Substitute w = 1 into the function o(w) and simplify to determine its value<EOT>\n-6\nLet y = -11 + 15. Suppose  $-4\ast c + 30 = -y^{\star}r + 50$ $\theta = -5^{*}c - r - 25$  . Let q(d) = d + 5. Give q(c).\\n<BOT>First calculate y, then solve the equations to find c, and finally substitute c into q(d) to calculate q(c)<EOT>\n0\nLet r = 2899 - 2906. Let i(s) = 2\*s\*\*2 + 12\*s - 3. Calculate i(r).\\n<BOT>First calculate r, then substitute r into the function i(s) to calculate i(r)<EOT>\n11\nLet q be  $6 / (-105) + 4199 / 595$  . Let m(t) = t\*\*3 - 9\*t\*\*2 + 5\*t + 5. Give m(q).\\n<BOT>First calculate q, then substitute q into the function m(t) to determine m(q)<EOT>\n-58\nLet d = 440 - 92\n<BOT>Calculate the value of d as 440 - 92<EOT>
+
+Figure 6: The prompt and in-context learning examples used for sampling rationales for The Pile. The bolded rationales represent implicit rationales in the document.
+
+# System Prompt:
+
+You are a smart assistant that solves math word problems. You will only generate one sentence that extends the reasoning trajectory that solves the question given the question and partial answer reasoning trajectory. Please don't repeat your previous generation while you're generating the sentence. If you think you're ready to output the answer, you can finish the response with The answer is:
+
+# Example Input:
+
+Question: Weng earns $12 an hour for babysitting. Yesterday, she just did 50 minutes of babysitting. How much did she earn?
+
+Answer:
+
+# Example Output:
+
+Weng earns 12/60 = $<<12/60=0.2>>0.2 per minute.
+
+# Example Input:
+
+Question: Weng earns $12 an hour for babysitting. Yesterday, she just did 50 minutes of babysitting. How much did she earn?
+
+Answer:Weng earns  $12 / 60 = \$ <<12 / 60 = 0.2>>0.2$ per minute.
+
+# Example Output:
+
+Working 50 minutes, she earned \(0.2 \times 50 = \\)<<0.2*50=10>>10\( . The answer is: 10
+
+Figure 7: The prompt and in-context-learning demonstrations used during inference time to encourage the agent model reason step by step on GSM8K.
+
+# System Prompt:
+
+You are a smart assistant that solves commonsense reasoning problems. You will only generate one sentence that extends the reasoning trajectory that solves the question given the question and partial answer reasoning trajectory. Please don't repeat your previous generation while you're generating the sentence. Please analyze all answer choices before finishing reasoning. If you think you're ready to output the answer, you can finish the response with The answer is:
+
+# Example Input:
+
+Question: He came across a raw item but his pack was full, he had to abandon something if he was to what the item?
+
+Choices: A - join, B - acquire, C - engage, D - maintain, E - remit
+
+Answer:
+
+# Example Output:
+
+Join is not the type of the activity associated with item.
+
+# Example Input:
+
+Question: He came across a raw item but his pack was full, he had to abandon something if he was to what the item?
+
+Choices: A - join, B - acquire, C - engage, D - maintain, E - remit
+
+Answer: Join is not the type of the activity associated with item.
+
+# Example Output:
+
+Acquiring requires space in a pack. To create space in a full pack, one has to abandon an item.
+
+# Example Input:
+
+Question: He came across a raw item but his pack was full, he had to abandon something if he was to what the item?
+
+Choices: A - join, B - acquire, C - engage, D - maintain, E - remit
+
+Answer: Join is not the type of the activity associated with item. Acquiring requires space in a pack. To create space in a full pack, one has to abandon an item.
+
+# Example Output:
+
+Engage is not related to item.
+
+# Example Input:
+
+Question: He came across a raw item but his pack was full, he had to abandon something if he was to what the item?
+
+Choices: A - join, B - acquire, C - engage, D - maintain, E - remit
+
+Answer: Join is not the type of the activity associated with item. Acquiring requires space in a pack. To create space in a full pack, one has to abandon an item. Engage is not related to item.
+
+# Example Output:
+
+Maintain an item does not require one to make space for it.
+
+# Example Input:
+
+Question: He came across a raw item but his pack was full, he had to abandon something if he was to what the item?
+
+Choices: A - join, B - acquire, C - engage, D - maintain, E - remit
+
+Answer: Join is not the type of the activity associated with item. Acquiring requires space in a pack. To create space in a full pack, one has to abandon an item. Engage is not related to item. Maintain an item does not require one to make space for it.
+
+# Example Output:
+
+One cannot remit raw item. The answer is B - acquire.
+
+Figure 8: The prompt and in-context-learning demonstrations used during inference time to encourage the agent model reason step by step on ECQA.
+
+# System Prompt:
+
+You are a smart assistant that solves math problems. You will only generate one sentence that extends the reasoning trajectory that solves the question given the question and partial answer reasoning trajectory. Please don't repeat your previous generation while you're generating the sentence.
+
+# Example Input:
+
+Question: If  $2x - 9y = 14$  and  $\$6x = 42 + y$ , what is the value of the product  $\$xy$ ? Answer:
+
+# Generated Rationale:
+
+Let's first rewrite the second equation to isolate $y$ in terms of $x$
+
+# Example Output:
+
+Rewrite the second equation to express  $y\$$  in terms of  $\$x$ :  $6x = 42 + y$ $\text{implies} y = 6x - 42$
+
+# Example Input:
+
+Question: If  $2x - 9y = 14$  and  $\$6x = 42 + y$ , what is the value of the product  $\$xy$ ? Answer: Rewrite the second equation to express  $\$y$ ' in terms of  $\$x$ :  $\$6x = 42 + y$ \`implies  $y = 6x - 42$
+
+# Generated Rationale:
+
+Substitute the expression for  $\$ y$  into the first equation
+
+# Example Output:
+
+Substitute  $y = 6x - 42$  into the first equation:  $\$ 2x - 9(6x - 42) = 14$
+
+# Example Input:
+
+Question: If  $2x - 9y = 14$  and  $\$6x = 42 + y$ , what is the value of the product  $\$xy$ ? Answer: Rewrite the second equation to express  $\$y$ 's in terms of  $\$x$ :  $\$6x = 42 + y$ implies  $y = 6x - 42$ . Substitute  $\$y = 6x - 42$  into the first equation:  $\$2x - 9(6x - 42) = 14$ \
+
+# Generated Rationale:
+
+We should simplify the equation and solve for $x$
+
+# Example Output:
+
+Simplify and solve for $x:$: $2x - 54x + 378 = 14 \implies -52x + 378 = 14 \implies -52x = -364 \implies x = 7$
+
+# Example Input:
+
+Question: If $2x - 9y = 14$ and $6x = 42 +y$, what is the value of the product $xy$? Answer: Rewrite the second equation to express $y$ in terms of $x:$: $6x = 42 + y \implies y = 6x - 42$. Substitute $y = 6x - 42$ into the first equation: $2x - 9(6x - 42) = 14$. Simplify and solve for $x:$: $2x - 54x + 378 = 14 \implies -52x + 378 = 14 \implies -52x = -364 \implies x = 7$
+
+# Generated Rationale:
+
+Use the value of $x$ to calculate $y$
+
+# Example Output:
+
+Substitute $x = 7$ back into $y = 6x - 42$ to find $y$: $y = 6(7) - 42 - 42 = 0$
+
+# Example Input:
+
+Question: If $2x - 9y = 14$ and $6x = 42 +y$, what is the value of the product $xy$? Answer: Rewrite the second equation to express $y$ in terms of $x$: $6x = 42 + y \implies y = 6x - 42$. Substitute $y = 6x - 42$ into the first equation: $2x - 9(6x - 42) = 14$. Simplify and solve for $x$: $2x - 54x + 378 = 14 \implies -52x + 378 = 14 \implies -364 \implies x = 7$. Substitute $x = 7$ back into $y = 6x - 42$ to find $y$: $y = 6(7) - 42 = 42 - 42 = 0$
+
+# Generated Rationale:
+
+Now we can just multiply the values of  $x$  and  $\$y$ 's to find the product
+
+# Example Output:
+
+Finally, calculate the product $xy$: $xy = 7 \cdot{0} = 0$. The value of the product $xy$ is $\boxed{0}$.
+
+Figure 9: Rationales generated by RATIONALYST for the test set of MATH.
+
+# System Prompt:
+
+Your task is to assign rewards to the reasoning trajectory that solves the question given the question and partial answer reasoning trajectory. The more probable the reasoning trajectory is correct, the higher the reward should be. The reward should be an integer in the range of 0 to 3. End your generation with So the reward is:
+
+# Example Input:
+
+Question: Michael had 58 golf balls. On tueyday, he lost 23 golf balls. On wednesday, he lost 2 more. How many golf balls did he have at the end of wednesday? Answer: Michael started with 58 golf balls. After losing 23 on tueyday, he had  $58 - 23 = 35$
+
+# Example Output:
+
+This reasoning trajectory is correct and reasonable. So the reward is: 3
+
+# Example Input:
+
+Question: Jason had 20 lollipops. He gave Denny some lollipops. Now Jason has 12 lollipops. How many lollipops did Jason give to Denny?
+
+Answer: Jason started with 20 lollipops. After giving some to Denny, he had 12 left. To find out how many he gave away, we need to add  $12 + 20$ , which gives us 32. So, Jason gave Denny 32 lollipops.
+
+# Example Output:
+
+This reasoning trajectory incorrectly adds instead of subtracting the lollipops given to Denny, leading to an illogical result. So the reward is: 1.
+
+Figure 10: The prompt and in-context-learning demonstrations used during process supervision to elicit the feedback by directly reranking partial reasoning trajectory.

rationalystpretrainingprocesssupervisionforimprovingreasoning/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f4178f23a0dcbd1c4dda10d209cf7c2db530b203c3c8bdde491807980425c6b
+size 425183

rationalystpretrainingprocesssupervisionforimprovingreasoning/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d61129772fee5f9e0c3c556137bca8cee42330bbe9fc90e4816ea6e657747d7e
+size 645648

re3synadependencybaseddatasynthesisframeworkforlongcontextposttraining/c5973251-ef67-447a-9904-062d624fffa8_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47d1561dcaedcadaca5c2543770c36a2fb43ac8c618790f4a1eae5a9b2d12481
+size 83477

re3synadependencybaseddatasynthesisframeworkforlongcontextposttraining/c5973251-ef67-447a-9904-062d624fffa8_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e057715c0a3abffc5bfa41e2a56ce919018c307f22d65b72eb77c4252e7fa3f
+size 102008