Add Batch 936e02b2-2f2a-4f46-839e-34b335216990

ACL/2025/Pixel-Level Reasoning Segmentation via Multi-turn Conversations/236b02f2-bb10-4d13-a21a-2ae88f712d15_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db9f02934a918b07a619a328f3ad8a30088743abe0e7de1f804032184cd035f3
+size 146666

ACL/2025/Pixel-Level Reasoning Segmentation via Multi-turn Conversations/236b02f2-bb10-4d13-a21a-2ae88f712d15_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55bb79f14cfbbb204008dc85e2b3dd174a25f3280a99f08357200517e4651fb1
+size 186247

ACL/2025/Pixel-Level Reasoning Segmentation via Multi-turn Conversations/236b02f2-bb10-4d13-a21a-2ae88f712d15_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04bc76cfc47264dafdf98cf201248882d882c39c211f1df8d7f74bac0068d89e
+size 11290398

ACL/2025/Pixel-Level Reasoning Segmentation via Multi-turn Conversations/full.md

@@ -0,0 +1,739 @@
+# Pixel-Level Reasoning Segmentation via Multi-turn Conversations
+
+Dexian Cai $^{1,\ast}$ , Xiaocui Yang $^{1,\ast}$ , Yongkang Liu $^{1}$ , Daling Wang $^{1,\dagger}$ , Shi Feng $^{1}$ , Yifei Zhang $^{1}$ , Soujanya Poria $^{2}$
+
+<sup>1</sup>School of Computer Science and Engineering, Northeastern University, Shenyang, China
+
+$^{2}$ Singapore University of Technology and Design, Singapore
+
+2301840@stu.neu.edu.cn, yangxiaocui@cse.neu.edu.cn, misonsky@163.com
+
+{wangdaling,fengshi,zhangyifei}@cse.neu.edu.cn, sporia@sutd.edu.sg
+
+# Abstract
+
+Existing visual perception systems focus on region-level segmentation in single-turn dialogues, relying on complex and explicit query instructions. Such systems cannot reason at the pixel level and comprehend dynamic user intent that changes over interaction. Our work tackles this issue by introducing a novel task, Pixel-level Reasoning Segmentation (Pixel-level RS) based on multi-turn conversations, tracking evolving user intent via multi-turn interactions for fine-grained segmentation. To establish a benchmark for this novel task, we build a Pixel-level Reasoning Segmentation Dataset Based on Multi-Turn Conversations (PRIST), comprising 24k utterances from 8.3k multi-turn conversational scenarios with segmentation targets. Building on PRIST, we further propose MIRAS, a Multi-turn Interactive ReAsoning Segmentation framework, integrates pixel-level segmentation with robust multi-turn conversation understanding, generating pixel-grounded explanations aligned with user intent. The PRIST dataset and MIRSA framework fill the gap in pixel-level reasoning segmentation. Experimental results on the PRIST dataset demonstrate that our method outperforms current segmentation-specific baselines in terms of segmentation and LLM-based reasoning metrics. The code and data are available at: https://github.com/ccccai239/PixelRIST.
+
+# 1 Introduction
+
+Existing general multimodal large language models (MLLMs) (Bai et al., 2023; Zhu et al., 2023; Liu et al., 2024b) exhibit exceptional visual perception, enabling both image segmentation and textual reasoning, while they primarily rely on explicit human instructions for region-level grounding. Although some segmentation-specific works have explored grounded reasoning responses (Peng et al.,
+
+† Corresponding author.  
+* Equal contribution.
+
+# (a) Reasoning Segmentation
+
+![](images/6ae129b9c7240d6d63d50f9797c6eecf94b2e92efaecfdd64a2531b296d4eb17.jpg)
+
+![](images/4af9fefd045ec10e1172244c0efd6b2063bd90a58ab8f3dea6cc262bb8abbbcd.jpg)  
+Figure 1: RS vs. Pixel-level RS. Pixel-level RS refines intent understanding and segmentation (e.g., "oil bottle") through multi-turn interactions, while RS produces rough segmentation (e.g., "all ingredients") and handles implicit single-turn queries poorly.
+
+2023; You et al., 2023; Pi et al., 2023; Zhang et al., 2023b), they depend on user-provided regions to trigger reasoning. These perception systems still cannot actively comprehend user's nuanced intent in real-world scenarios. To alleviate this problem, Lai et al. (2023) proposes the reasoning segmentation task that aims to achieve segmentation based on a implicit reasoning query. Recent studies (Ren et al., 2024; Xia et al., 2024; Yuan et al., 2024) have extended this region-level task to encompass multi-object segmentation scenarios to advance development. However, these methods have two limitations: 1) They rely on single-turn ambiguous queries and cannot fully understand users' evolving intent. 2) They lack pixel-level segmentation and only achieve region-level segmentation through one-step explanations (e.g., segment roughly all ingredients in Figure 1(a)). In contrast, multi-turn interactions can progressively clarify vague and generalized instructions such as "make a bread". As illustrated in Figure 1(b), the system through multi-turn interactions first guide to clarify the user's desired type of bread, providing targeted responses, and ultimately focuses on the user's specific needs, achieving pixel-level segmentation in final.
+
+To address these challenges, we propose a novel task, Pixel-level Reasoning Segmentation (Pixel-level RS) based on multi-turn conversations, that refines both reasoning and segmentation through multi-turn interactions, requiring the system to understand the evolving user intent and generating pixel-level explanations, and segmentation masks. Given the lack of benchmarks for pixel-level segmentation based on multi-turn reasoning, we build a Pixel-level Reasoning Segmentation Dataset Based on Multi-Turn Conversation (PRIST), consisting of 24k utterances, 8.3k multi-turn conversational scenarios with specific segmented targets, which provides a valuable resource for advancing Pixel-level RS research. PRIST focuses on pixel-level segmentation tasks while introducing new challenges in multi-turn reasoning and evolving intent comprehension. We design a progressive three-step dialogue automatic generation pipeline based on a reasoning tree to iteratively guide and generate dialogue content, inspired by Tree-of-Thought (ToT) (Yao et al., 2024). By integrating a multi-step reasoning chain with a tree structure, this approach facilitates deeper and broader reasoning in pixel-level segmentation training.
+
+To further advance this novel task, we propose a Multi-turn Interactive ReAsoning Segmentation framework, MIRAS, that enables pixel-level segmentation through progressive reasoning. MIRAS incorporates a dual-vision encoder that fuses multiscale features to capture detailed visual information. To improve segmentation performance, we introduce a semantic region alignment strategy to inject semantic information into the mask decoder. Additionally, the framework supports multi-turn interactions to iteratively clarify user intent and ambiguous regions. Given the inherent subjectivity in reasoning tasks, manual assessments can be influenced by personal preferences. To ensure fairness, we develop comprehensive evaluation metrics leveraging Large Language Models (LLMs) to assess multi-turn reasoning segmentation across coherence, consistency, and accuracy dimensions. Our contributions can be summarized as follows:
+
+- We propose a novel task, Pixel-level Reasoning Segmentation, that aims to achieve fine-grained segmentation through multi-turn conversations. Then, we build the PRIST dataset, including 8.3k high-quality multi-turn conversational scenarios and pixel-level segmentation targets.  
+- We develop a multi-turn reasoning segmentation framework, MIRAS, that facilitates pixel
+
+level intentional understanding and segmentation through multi-turn interactions.
+
+- Comprehensive experimental results of our proposed method on different metrics demonstrate both the utility of the PRIST dataset and the effectiveness of the model.
+
+# 2 Related Work
+
+# 2.1 Datasets for Reasoning Segmentation
+
+Current reasoning segmentation datasets (Lai et al., 2023; Rasheed et al., 2024; Yuan et al., 2024) mainly focus on region-level segmentation and single-step reasoning, which fail to meet the comprehensive requirements of the pixel-level RS task. Traditional region-level segmentation datasets (Yu et al., 2016; Krishna et al., 2017) rely on simple and explicit instructions, lacking complex user intents understanding. To bridge this gap, ReasonSeg (Lai et al., 2023) is the first to propose a segmentation dataset based on complex queries, which is a small scale and does not support multi-turn interactions. While recent datasets (Yuan et al., 2024; Rasheed et al., 2024; Ren et al., 2024) have expanded in size and diversity, they remain focused on multi-object segmentation and provide limitations for multi-step reasoning and fine-grained grounding. In contrast, our PRIST dataset utilizes conversation to simulate a human-like multi-step reasoning process, which innovatively combines multi-turn reasoning with pixel-level segmentation. Detailed dataset comparisons are shown in Table 8 in Appendix B.3.
+
+# 2.2 MLLMs for Region-level Segmentation
+
+MLLMs have advanced vision-language perception tasks, with recent works (Peng et al., 2023; You et al., 2023; Zhang et al., 2023b) focusing on image-level visual dialogue. Some (Chen et al., 2023; Peng et al., 2023; Pi et al., 2023) achieve region-level understanding by incorporating positional information and boundary boxes, mainly relying on LLMs for region interpretation. Several models (Lai et al., 2023; Ren et al., 2024; Rasheed et al., 2024; Zhang et al., 2024) integrate segmentation-specific modules with LLMs for end-to-end training, enabling a more comprehensive understanding of regions. While these methods address pixel-level grounding, they still face limitations in complex reasoning. More comparison between models in Appendix A.2. Our proposed MIRAS overcomes these challenges by enhancing segmentation accuracy through interactive reasoning, progressively
+
+![](images/b4fbd3473f700884c586d92fddaf31a1672dd6630f8abe796a0bb7febc3dd0e1.jpg)  
+Figure 2: The Generation Pipeline of PRIST Dataset. i) Step 1 extracts visible elements from images, establishing a semantic foundation for subsequent steps. ii) Step 2-1 generates complex reasoning questions from these elements, while Step 2-2 iteratively refines the questions through a reasoning tree, ensuring rigorous reasoning. iii) Step 3 organizes the nodes in reasoning tree into a multi-turn dialogue format.
+
+refining the boundaries.
+
+# 3 PRIST
+
+The pixel-level RS task takes an image  $\mathcal{I}$  and a multi-turn dialogue  $\mathcal{D}$  as input, then simultaneously generates a target segmentation mask  $\mathcal{M}$  along with a textual reasoning chain  $\{a_1, a_2, \ldots, a_N\}$  that captures the complete dialogue history, where  $a_i$  denotes the system's response in the  $i$ -th turn. It is defined as follows:
+
+$$
+(\mathcal {M}, \left\{a _ {1}, a _ {2}, \dots , a _ {N} \right\}) = \operatorname {M o o l e r} (\mathcal {I}, \mathcal {D}). \tag {1}
+$$
+
+Furthermore, we construct the pixel-level reasoning segmentation dataset based on the multi-turn conversation (PRIST) using a three-step progressive annotation pipeline, capturing fine-grained details through context-aware multi-turn dialogue.
+
+# 3.1 Data Preparation
+
+Given the focus on pixel-level segmentation, we select TextCaps* (Sidorov et al., 2020) as the image source due to its detailed visual information. To ensure a diverse range of scenes, we randomly select 280 images from each of the 10 major categories, resulting in a total of 2.8k images.
+
+# 3.2 Generation Pipeline
+
+We propose a three-step progressive automated annotation pipeline to create the PRIST Dataset, as illustrated in Figure 2. Appendix B.2 details the pipeline's prompts and output formats.
+
+# 3.2.1 Visual Elements Extraction (Step-1)
+
+We first extract  $N$  visible objects  $\mathcal{O} = \{o_i\}_{i=1}^N$ , from the input image  $\mathcal{I}$ . Each object  $o_i$  represents a distinct target for generating a dialogue. Specifically, we automatically identify visible elements with detailed attributes (e.g., color, position) to each object using GPT-4o (Achiam et al., 2023), along with corresponding textual descriptions (see Figure 7a). This step ensures that visual and semantic details are fully represented.
+
+# 3.2.2 Reasoning Process Construction (Step-2)
+
+Pixel-level RS addresses complex scenarios requiring multi-turn reasoning with implicit user instructions. To simulate such scenarios, we implement a "question-first, problem-solving" strategy as shown in Figure 2, where the reasoning process is refined after first forming the reasoning problem and then constructing the reasoning tree. To align with multi-turn reasoning, we propose a hierarchical reasoning tree that recursively decomposes complex questions into smaller subquestions, progressively focusing on segmentation targets. Each reasoning
+
+tree path connects related elements to build a logical chain for pixel-level segmentation.
+
+Reasoning Question Formation (Step 2-1) This step expands a complex reasoning question  $\mathcal{Q}_i$  for each target  $o_i$ , serving as the overall origin for next question decomposition and the theme for multi-turn dialogues in Step-3. To balance processing efficiency, we randomly select  $K$  objects  $(2 \leq K \leq \min(N,4))$  from the element set  $\mathcal{O}$  in Step-1 as targets, as fewer may missing essential interactions while more escalate complexity.
+
+$$
+\mathcal {Q} _ {i} = \text {F o r m a t i o n} (\mathcal {I}, o _ {i}), i = \{1, \dots , K \}. \tag {2}
+$$
+
+Reasoning Tree Construction (Step 2-2) The construction process (see Figure 7b) with GPT-4o initiates by establishing the elements  $o_i$  as leaf nodes, allowing their corresponding reasoning problems  $\mathcal{Q}_i$  to develop a distinct path  $\mathcal{P}_i$  within the reasoning tree  $\mathcal{T}$ . Through iterative decomposition, each  $\mathcal{Q}_i$  evolves into a sequence of progressive QA pairs  $(q_n, a_n)$ , with the tree's depth directly corresponding to the granularity of subquestions. This hierarchical expansion refines the problem-solving framework and progressively narrows pixel-level target localization. The resulting reasoning tree  $\mathcal{T}$  explicitly captures the logical progression of complex questions, providing a structural foundation for the multi-turn dialogues in Step-3.
+
+$$
+\begin{array}{l} \mathcal {P} _ {i} = \operatorname {C o n s t r u c t i o n} \left(\mathcal {Q} _ {i}, o _ {i}\right), \\ = \left\{\left(q _ {1}, a _ {1}\right), \dots , \left(q _ {n}, a _ {n}\right) \right\}, \tag {3} \\ \end{array}
+$$
+
+$$
+\mathcal {T} = \left\{\mathcal {P} _ {1}, \mathcal {P} _ {2}, \dots , \mathcal {P} _ {i} \right\} _ {i = 1} ^ {K},
+$$
+
+where  $n$  is the depth of the reasoning path. To manage computational resources and logical flexibility, we impose a constraint limiting each reasoning tree layer to a maximum of three child nodes.
+
+# 3.2.3 Multi-turn Dialogue Generation (Step-3)
+
+We further build multi-turn dialogue  $\mathcal{D}_i$  based on the hierarchical reasoning tree from Step-2. Specifically, we integrate all nodes in each reasoning path  $\mathcal{P}_i$ , where each node represents a QA pair, to form the progressive multi-turn dialogue  $\mathcal{D}_i$ . Thus, each image can form  $K$  conversations,  $\{\mathcal{D}_1,\dots ,\mathcal{D}_i\}_{i = 1}^K$ . To ensure responses fully integrate contextual information, prompts are designed to incorporate key elements and relationships from Step-1 (see Figure 7c), expanding understanding of landmarks, historical context, and scene interactions. Furthermore, pixel-level RS is designed to guide the model in performing fine-grained segmentation, with the final query in each
+
+dialogue being a segmentation-related instruction (e.g., "Please segment the core objects according to the above dialogue").
+
+<table><tr><td></td><td>G1</td><td>G2</td><td>G3</td><td>G4</td><td>G5</td></tr><tr><td>IoU</td><td>0.82</td><td>0.88</td><td>0.91</td><td>0.86</td><td>0.81</td></tr><tr><td>Kappa</td><td>0.79</td><td>0.82</td><td>0.80</td><td>0.78</td><td>0.75</td></tr></table>
+
+Table 1: Consistency Analysis Results for Mask Annottators.  $Gx$  denotes the  $x$ -th expert group.
+
+# 3.3 Quality Assurance
+
+We employ manual annotation to generate the true segmentation masks for each sample. Ten experts, selected through qualification tests, annotate masks and correct any commonsense errors in the dialogues. To ensure the quality of PRIST, we implement a double-check process. Specifically, experts are organized into five groups, with each group annotating the same set of samples. This set ensures that two annotators independently annotate every sample. More details are in Appendix B.1. The reliability of annotations across the five groups is assessed using IoU (Girshick et al., 2014), which measures the overlap between two annotators, and Cohen's Kappa (Cohen, 1960), which quantifies their consistency. As shown in Table 1, all groups achieve a result of IoU  $>0.80$  and Kappa  $>0.75$ , demonstrating high annotation quality and strong consistency across the groups.
+
+# 3.4 Dataset Analysis
+
+As shown in Table 2, we provide detailed statistics of the PRIST dataset. PRIST contains 24k high-quality utterances, and 8.3k multi-turn conversational scenarios, with each scenario focusing on a single, fine-grained object. The dataset is divided into three subsets: train / validation / test splits, containing 7,312 / 500 / 508 samples, respectively.
+
+To quantify pixel-level segmentation, we adopt the COCO mask size standard for measuring target scales, aligning with existing datasets (Chen et al., 2015; Caesar et al., 2018) and providing a quantitative basis for fine-grained segment evaluation. Statistics in Table 2, fine-grained targets (the scaling factor  $s$  is 1.6) account for  $53\%$  of PRIST, surpassing the  $41\%$  of small targets in COCO (Chen et al., 2015). With a minimum mask area of 304px and a standardized image size of  $1024 \times 1024$ , PRIST meets the fine-grained re
+
+(b) Adjective  
+Figure 3: The focus distribution of PRIST. We analyze focus objects across 3 dimensions: noun, adjective and preposition, which capture fine granularity, diversity, and close spatial relationships between objects.  
+![](images/b5a266e397ef33cc4473d089ae0539498539d5c22783a128ef2fd9bd4d1c8337.jpg)  
+Properties (28.6%) digital, bold, free, multiple, old, vintage ...  
+Position (21.7%) left, right, top, front, central ...  
+Size (16.3%) large, small, mini, tall ... Color (12.2%)  
+red, blue, dark, light, colorful ...  
+Shape (7.5%) circular, round, vertical, floral.  
+Cultural (6.8%) American, Chinese, Japanese ...  
+Material (5.3%)  
+wooden, metallic, plastic ...
+
+![](images/57dd40ae022109b26d346166f3d85b3a283d341a98200874db2e2adf6dd01dd0.jpg)  
+(c) Preposition
+
+<table><tr><td>Statistic</td><td>Number</td></tr><tr><td>Total Images</td><td>2,800</td></tr><tr><td>Total Samples</td><td>8,320</td></tr><tr><td>Segmentation</td><td></td></tr><tr><td>- Focus Classes</td><td>12</td></tr><tr><td>- Granularity (Coarse: Med.: Fine)</td><td>22%: 25%: 53%</td></tr><tr><td>Multi-turn Dialogue</td><td></td></tr><tr><td>- Number of Utterances</td><td>24k</td></tr><tr><td>- Avg. / Max. Turns</td><td>4 / 7</td></tr><tr><td>- Avg. / Max. Dialogue Length</td><td>477.6 / 518</td></tr></table>
+
+Table 2: PRIST Statistics. According to COCO's image standard (640 vs. 1024), mask granularity is categorized as "Fine"  $(< (s \times 32)^{2} \mathrm{px})$ , "Med."  $((s \times 32)^{2}$  to  $(s \times 96)^{2} \mathrm{px})$ , and "Coarse"  $(> (s \times 96)^{2} \mathrm{px})$ .
+
+requirements of pixel-level segmentation. We emphasizes exhibiting high diversity in categories and descriptions to enhance expressiveness. Illustrated the focus distribution of PRIST in Figure 3, the categories include four types: Textual Content, Physical Objects, Visual Elements, and People & Environment, which are further refined into 12 subcategories that cover objects from coarse- to fine-grained levels. At the descriptive level, a combination of Noun, Adjective, and Preposition is employed: nouns provide basic category information (e.g., "tree"), adjectives enrich focus details (e.g., "worn-out chair"), and prepositions describe
+
+spatial relationships (e.g., "a book under the table"). PRIST delivers rich semantic-spatial annotations, establishing a benchmark resource for Pixel-level RS advancement.
+
+# 4 MIRAS
+
+To further research the novel pixel-level RS task, we propose MIRAS, a framework that refines user intent through multi-turn interactions to achieve pixel-grounded explanations and segmentation.
+
+# 4.1 Architecture
+
+The architecture of MIRAS is illustrated in Figure 4, consisting of three core components: Visual Encoder, MLLM  $(\mathcal{F},$  (Liu et al., 2024a)), and Mask Decoder  $(\mathcal{D}_m,$  (Kirillov et al., 2023)). To seamlessly connect reasoning with segmentation, we introduce a special token [SEG] as a placeholder for segmenting regions and enabling end-to-end processing. Two key modules are proposed for pixel-level RS. First, we integrate dual visual encoders (Li et al., 2023) to extract enriched visual features. Then, the semantic region alignment strategy is designed to further refine the model's focus by incorporating target semantic information.
+
+Dual Visual Encoders We train the dual visual encoder using a high-resolution image  $(\mathbf{X}_H, 768$
+
+![](images/e88e2496518420b8480c85ff6b8d0c96054805057994d0ed43581942887f0006.jpg)  
+Figure 4: Overview Architecture of MIRAS. The model integrates MLLM and SAM modules by introducing a special token [SEG]. MIRAS can perform both (a) Multi-turn Response and (b) Segmentation tasks end-to-end.
+
+$\times 768$  pixels) processed by ConvNext-L (Liu et al., 2022) paired with its low-resolution counterpart  $(\mathbf{X}_L,336\times 336$  pixels) processed by CLIP-L/14 Radford et al., 2021), downsampled from  $\mathbf{X}_H$  Then, different resolutions are fused by a crossattention module (Lin et al., 2022) to enhance visual detail capturing. Note that  $\mathbf{X}_H$  equals  $\mathbf{X}_{\mathrm{img}}$
+
+$$
+\mathbf {X} _ {H} ^ {\prime} = \operatorname {C o n v N e x t} (\mathbf {X} _ {H}), \quad \mathbf {X} _ {H} ^ {\prime} \in \mathbb {R} ^ {H \times W \times 3},
+$$
+
+$$
+\mathbf {X} _ {L} ^ {\prime} = \mathbf {C L I P} (\mathbf {X} _ {L}), \quad \mathbf {X} _ {L} ^ {\prime} \in \mathbb {R} ^ {H ^ {\prime} \times W ^ {\prime} \times 3}, \tag {4}
+$$
+
+$$
+\mathbf {E} _ {\mathrm {i m g}} ^ {\prime} = \operatorname {C r o s s A t t e n} (Q = \mathbf {X} _ {L} ^ {\prime}, K = \mathbf {X} _ {H} ^ {\prime}, V = \mathbf {X} _ {H} ^ {\prime}),
+$$
+
+$$
+\mathbf {E} _ {\mathrm {i m g}} = \operatorname {M L P} \left(\mathbf {E} _ {\mathrm {i m g}} ^ {\prime}\right) + \mathbf {E} _ {\mathrm {i m g}} ^ {\prime},
+$$
+
+Semantic Region Alignment We employ the SAM to obtain the pixel-level image features.
+
+$$
+\mathbf {E} _ {\mathrm {s e g}} = \mathcal {V} _ {\mathrm {p i x e l}} \left(\mathbf {X} _ {\mathrm {i m g}}\right), \quad \mathbf {E} _ {\mathrm {s e g}} \in \mathbb {R} ^ {N \times 2 5 6}. \tag {5}
+$$
+
+To provide clear segmentation intent, we design a novel segmentation prompt template [OBJ]{CLASS}[SEG], where {CLASS} is the object description (e.g., [OBJ] a bunch of grapes [SEG] in Figure 4). We further utilize special tokens [OBJ] to extract relevant sub-sequence  $\mathcal{H}_{\mathrm{seg}}$  from  $\mathcal{H}$  for segmentation and employ the cross-attention module to capture sufficient semantic information, which is crucial for efficient fine-grained segmentation. The method resolves potential mismatches in dimensions caused by varying lengths of {CLASS}, denoted as  $N^{sub}$ .
+
+$$
+\mathcal {H} _ {\text {s e g}} ^ {\prime} = \operatorname {C r o s s A t t e n} (Q, K, V | \mathcal {H} _ {\text {s e g}}), \tag {6}
+$$
+
+$$
+\mathcal {H} _ {\text {s e g}} \in \mathbb {R} ^ {N ^ {\text {s u b}} \times 2 5 6}, \mathcal {H} _ {\text {s e g}} ^ {\prime} \in \mathbb {R} ^ {1 \times 2 5 6}, \tag {6}
+$$
+
+The mask decoder  $\mathcal{D}_m$  combines region features from the pixel encoder with the hidden features of [SEG] to produce the final mask.
+
+$$
+\hat {\mathcal {M}} = \mathcal {D} _ {m} \left(\mathbf {E} _ {\text {s e g}}, \mathcal {H} _ {\text {s e g}} ^ {\prime}\right), \tag {7}
+$$
+
+# 4.2 Training Process
+
+We employ a two-stage training process to achieve efficient pixel-level reasoning segmentation. In Stage-1, mask-text alignment pretraining based on various datasets is conducted, followed by instruction fine-tuning using the PRIST dataset in Stage-2 (more details in Appendix C.2). The objectives remain consistent across both stages: the text generation loss  $\mathcal{L}_{\mathrm{t}}$  and a linear combination of per-pixel BCE loss  $\mathcal{L}_{bce}$  and DICE loss  $\mathcal{L}_{dice}$  for segmentation. Only the mask decoder and projection layer are trainable to balance efficiency and performance while keeping the image encoder and MLLM frozen. The training loss is formulated as:
+
+$$
+\mathcal {L} = \lambda_ {t} \mathcal {L} _ {\mathrm {t}} (\mathcal {R}, \hat {\mathcal {R}}) + \lambda_ {b c e} B C E (\mathcal {M}, \hat {\mathcal {M}}) \tag {8}
+$$
+
+$$
++ \lambda_ {d i c e} D I C E (\mathcal {M}, \hat {\mathcal {M}})
+$$
+
+where  $\lambda_{\mathrm{t}}, \lambda_{\mathrm{bce}}$  and  $\lambda_{\mathrm{dice}}$  values 1.0, 2.0 and 0.5 separately, following LISA (Lai et al., 2023).
+
+# 5 Experiment
+
+Experiments evaluate our model's adaptability to the novel pixel-level RS task and its general performance in the classical referring expression segmentation (RES) task (Kazemzadeh et al., 2014; Yu et al., 2016). Appendix C shows backbone selection and more implementation details.
+
+# 5.1 Evaluation Metrics
+
+We establish a benchmark for Pixel-level RS across three evaluation aspects: pixel-level segmentation,
+
+<table><tr><td rowspan="2">Model</td><td rowspan="2">CIoU</td><td colspan="3">Pixel-wise</td><td colspan="5">Response</td></tr><tr><td>Prec.</td><td>Recall</td><td>F1</td><td>BLEU-4</td><td>Dist-1/2</td><td>ROU_L.</td><td>MET.</td><td>BERTS.</td></tr><tr><td colspan="10">Zero-shot</td></tr><tr><td>InternVL2-8B* (Chen et al., 2024b)</td><td>8.26</td><td>7.59</td><td>11.55</td><td>9.16</td><td>1.20</td><td>7.5 / 41.0</td><td>18.44</td><td>23.98</td><td>84.78</td></tr><tr><td>Qwen2-VL-7B* (Wang et al., 2024a)</td><td>10.64</td><td>10.30</td><td>15.88</td><td>12.50</td><td>3.00</td><td>9.3 / 41.2</td><td>26.28</td><td>28.26</td><td>86.73</td></tr><tr><td>LLaVA-v1.5-7B* (Liu et al., 2024b)</td><td>11.25</td><td>11.35</td><td>25.90</td><td>15.68</td><td>3.21</td><td>5.6 / 27.5</td><td>16.02</td><td>18.41</td><td>78.81</td></tr><tr><td>LLaVA-v1.6-7B* (Liu et al., 2024a)</td><td>11.84</td><td>11.90</td><td>34.78</td><td>17.69</td><td>1.07</td><td>5.5 / 27.5</td><td>20.60</td><td>25.00</td><td>84.26</td></tr><tr><td>GPT-4o* (OpenAI, 2024)</td><td>14.13</td><td>17.35</td><td>35.01</td><td>23.18</td><td>4.30</td><td>9.1 / 42.9</td><td>26.35</td><td>28.55</td><td>87.62</td></tr><tr><td>LISA (Lai et al., 2023)</td><td>10.45</td><td>15.33</td><td>43.07</td><td>15.09</td><td>1.97</td><td>6.2 / 28.4</td><td>18.21</td><td>26.59</td><td>85.67</td></tr><tr><td>PixelLM (Ren et al., 2024)</td><td>9.87</td><td>17.21</td><td>35.36</td><td>14.68</td><td>1.34</td><td>6.7 / 30.1</td><td>14.93</td><td>21.26</td><td>85.13</td></tr><tr><td>OMG-LLaVA (Zhang et al., 2024)</td><td>9.67</td><td>16.67</td><td>77.80</td><td>27.46</td><td>8.70</td><td>11.2 / 42.0</td><td>23.47</td><td>27.90</td><td>87.30</td></tr><tr><td>MIRAS (Stage-1)</td><td>13.12</td><td>15.64</td><td>45.11</td><td>23.22</td><td>4.17</td><td>6.9 / 28.4</td><td>25.94</td><td>28.18</td><td>87.54</td></tr><tr><td colspan="10">Fine-tuning</td></tr><tr><td>LISA (Lai et al., 2023)</td><td>11.23</td><td>26.23</td><td>29.22</td><td>27.64</td><td>7.81</td><td>14.2 / 40.7</td><td>27.84</td><td>30.74</td><td>86.75</td></tr><tr><td>PixelLM (Ren et al., 2024)</td><td>10.32</td><td>20.95</td><td>18.84</td><td>11.71</td><td>9.97</td><td>11.6 / 38.0</td><td>30.63</td><td>32.99</td><td>87.80</td></tr><tr><td>OMG-LLaVA (Zhang et al., 2024)</td><td>13.84</td><td>21.54</td><td>49.31</td><td>29.98</td><td>11.21</td><td>12.3 / 35.3</td><td>30.59</td><td>39.18</td><td>88.76</td></tr><tr><td>MIRAS (Stage-2)</td><td>14.72</td><td>24.22</td><td>40.61</td><td>30.34</td><td>8.51</td><td>15.7 / 49.2</td><td>30.82</td><td>40.06</td><td>88.47</td></tr></table>
+
+Table 3: Results on Pixel-level RS. MIRAS employs LLaVA v1.6 as its backbone due to its superior zero-shot performance. * denotes the general MLLMs, others are 7B segmentation-specific MLLMs.
+
+conversation response, and reasoning quality. For segmentation, we employ the CIoU metric (Zheng et al., 2020) and propose pixel-wise mask precision, recall, and F1 metrics. The precision measures the accuracy of segmentation while recall evaluates the coverage. For response, the metrics including BLEU-4 (Papineni et al., 2002), ROUGE-L (Kingma, 2014), Dist-n (Li et al., 2015), and METEOR (Banerjee and Lavie, 2005). Considering the subjectivity of reasoning, we introduce LLM as a scoring tool using four metrics: Progressiveness (PR), Logical Coherence (LC), Content Consistency (CC), and Target Relevance (TR), with higher scores reflecting better performance across aspects. Meanwhile, we employ GPT-4o as a judge to assess dialogue reasoning quality. The model wins when its score surpasses that of the human response, as reflected by the Win Rate (\%) metric. Evaluation criteria are detailed in Appendix C.4.
+
+# 5.2 Baselines
+
+We compare with three types of baselines: 1) General MLLMs. We take advanced close- and open-source MLLMs to evaluate zero-shot on PRIST for pixel-level RS capability. 2) Segmentation-specific MLLMs. LISA (Lai et al., 2023), PixelLM (Ren et al., 2024) and OMG-LLaVA (Zhang et al., 2024) are evaluated under zero-shot and fine-tuning settings to show PRIST's task-specific enhancement. 3) Segmentation-specific Models. We compare three semantic segmentation models (e.g., LVIT (Yang et al., 2022)) on the RES task with
+
+MIRAS to verify its advantage of the architecture.
+
+# 5.3 Experimental Results and Analysis
+
+As shown in Table 3, we comprehensively compare model performance across two critical dimensions: pixel-level segmentation and conversation response. Our proposed MIRAS demonstrates robust performance in pixel-level RS task, surpassing baselines and the closed-source GPT-4o in both segmentation and response metrics, establishing a new benchmark for the PRIST dataset. Notably, the performance improvement from PRIST fine-tuning exhibits universal applicability across different architectures. Table 4 further validates the enhanced reasoning capabilities of our method, with evaluation results approaching human expert proficiency.
+
+# 5.3.1 Pixel-level Segmentation
+
+As illustrated in the left part of Table 3, GPT-4o achieves CIoU 14.13 and precision 17.35, surpassing open-source models. While it surpasses stage-1 MIRAS, it remains below fine-tuned MIRAS (stage-2), indicating our framework can achieve closed-source model competency with efficient resource utilization. To enhance pixel-level performance, all segmentation-specific MLLMs are fine-tuned on the PRIST dataset under a consistent setting. Fine-tuning results in an increase in terms of CIoU and Precision, with OMG-LLaVA's CIoU  $\uparrow 43\%$ , Precision  $\uparrow 29\%$ , and LISA's Precision from 15.33 to 26.23 ( $\uparrow 71\%$ ), respectively. MIRAS (stage-2) establishes new benchmarks with
+
+precision 24.22, F1 30.34, and CIoU 14.72, demonstrating exceptional boundary delineation capabilities. Notably, this performance enhancement accompanies a precision-recall trade-off, i.e., recall decreases (LISA  $\downarrow 32\%$ , OMG-LLaVA  $\downarrow 37\%$ ).
+
+In zero-shot settings, all general MLLMs exhibit limited pixel-level segmentation, with GPT-4o slightly outperforming segmentation-specific MLLMs like PixelLM (Precision 17.35 vs. 17.21). This indicates that general MLLMs emphasize cross-task adaptability, while task-specific improvements rely on design-specific architecture (e.g., Mask Decoder). Additionally, the precision-recall trade-off observed in fine-tuned models reflects a strategic prioritization of segmentation specificity over generalizability in fine-grained tasks, which avoids the overgeneralization issues encountered in zero-shot settings, aligning with the objectives of pixel-level RS. An optimization choice validated by case studies in Appendix D.
+
+# 5.3.2 Conversation Response
+
+In the right of Table 3, GPT-4o demonstrates near-expert dialogue competence across metrics, approaching fine-tuned MLLMs, such as OMGLLaVA, while outperforming open-source general MLLMs like Qwen2-VL-7B. MIRAS achieves the best performance in metrics such as Dist-1/2 (15.7/49.2), ROUGE_L (30.82), and METEOR (40.06), validating its ability to generate high-quality textual responses. GPT-4o's strong baseline performance underscores its inherent dialogic intelligence; however, domain adaptation remains essential for optimal performance. Most fine-tuned models show improvements in text metrics, highlighting PRIST's effectiveness in bridging visual-textual semantic gaps. This demonstrates the framework's dual competence in simultaneously optimizing mask-text alignment and response coherence.
+
+# 5.3.3 Reasoning Quality
+
+Table 4 presents the reasoning quality evaluation based on LLMs (human scores detailed in Appendix C.4.3). Fine-tuning on the PRIST dataset led to improvements across all models, with an average Win Rate increase of approximately  $10\%$ . MIRAS achieved the SOTA with a Win Rate of  $42\%$  and the highest scores in all four reasoning metrics, closely approaching human expert levels. The overall improvement in four fine-tuned MLLMs' reasoning quality shows the substantial potential of PRIST dataset in enhancing reasoning
+
+capabilities, stemming from the extensive conceptual vocabulary it provides during fine-tuning.
+
+<table><tr><td>Model</td><td>PR</td><td>LC</td><td>CC</td><td>TR</td><td>Win Rate(%)</td></tr><tr><td>Human</td><td>4.03</td><td>4.04</td><td>4.26</td><td>4.28</td><td>-</td></tr><tr><td>LISA(ft)</td><td>3.76</td><td>3.69</td><td>3.71</td><td>3.58</td><td>36 (↑11)</td></tr><tr><td>PixelLM(ft)</td><td>3.35</td><td>3.48</td><td>3.32</td><td>3.28</td><td>32 (↑13)</td></tr><tr><td>OMG-LLaVA(ft)</td><td>2.60</td><td>2.48</td><td>2.58</td><td>2.33</td><td>24 (↑8)</td></tr><tr><td>MIRAS(Stage-2)</td><td>3.90</td><td>3.76</td><td>3.83</td><td>3.69</td><td>42 (↑11)</td></tr></table>
+
+# 5.4 Generalization Segmentation
+
+We compare segmentation-specific baselines on classical referring expression segmentation benchmarks to evaluate the generalizability of MIRAS. As detailed in Table 5, MIRAS' base configuration (last row) outperforms segmentation-specific models and demonstrates competitiveness against other MLLMs, even surpassing the latest OMG-LLaVA. Two findings emerge in results: (1) While fine-tuned MIRAS (Stage-2) shows a decline in general performance due to task-specific optimization, it retains advantages over Next-Chat and remains comparable to OMG-LLaVA. (2) The capacity of the base model determines the system's potential, evidenced by consistent improvements in MIRAS when evolving the foundational model from LLaVA-v1 (row 6) to LLaVA-v1.6 (row 9).
+
+Table 4: Comparison of the reasoning quality of domain-specific MLLMs fine-tuned on PRIST.  
+
+<table><tr><td rowspan="2">Model</td><td colspan="3">refCOCO</td><td colspan="3">refCOCO+</td><td colspan="2">refCOCOg</td></tr><tr><td>Val</td><td>TestA</td><td>TestB</td><td>Val</td><td>TestA</td><td>TestB</td><td>Val</td><td>Test</td></tr><tr><td>GRIS (Wang et al., 2022)</td><td>70.5</td><td>73.2</td><td>66.1</td><td>65.3</td><td>68.1</td><td>53.7</td><td>59.9</td><td>60.4</td></tr><tr><td>LAVT (Yang et al., 2022)</td><td>72.7</td><td>75.8</td><td>68.8</td><td>62.1</td><td>68.4</td><td>55.1</td><td>61.2</td><td>62.1</td></tr><tr><td>GRES (Liu et al., 2023)</td><td>73.8</td><td>76.5</td><td>70.2</td><td>66.0</td><td>71.0</td><td>57.7</td><td>65.0</td><td>66.0</td></tr><tr><td>LISA (v1) (Lai et al., 2023)</td><td>74.1</td><td>76.5</td><td>72.3</td><td>65.1</td><td>70.8</td><td>58.1</td><td>67.9</td><td>70.6</td></tr><tr><td>PixelLM (v1) (Ren et al., 2024)</td><td>73.0</td><td>76.5</td><td>68.2</td><td>66.3</td><td>71.7</td><td>58.3</td><td>69.3</td><td>70.5</td></tr><tr><td>MIRAS (Stage-1) (v1)</td><td>75.3</td><td>78.9</td><td>70.2</td><td>66.7</td><td>74.3</td><td>61.5</td><td>73.2</td><td>71.9</td></tr><tr><td>OMG-LLaVA (Zhang et al., 2024)</td><td>78.0</td><td>80.3</td><td>74.1</td><td>69.1</td><td>73.1</td><td>63.0</td><td>72.9</td><td>72.9</td></tr><tr><td>MIRAS (Stage-2) (v1.6)</td><td>76.9</td><td>79.8</td><td>72.8</td><td>68.8</td><td>74.4</td><td>62.5</td><td>71.8</td><td>70.8</td></tr><tr><td>MIRAS (Stage-1) (v1.6)</td><td>78.4</td><td>80.5</td><td>73.4</td><td>72.1</td><td>74.8</td><td>63.4</td><td>72.6</td><td>72.0</td></tr></table>
+
+Table 5: Results on the RES benchmark. "v1/v1.6" indicates LLaVA version.
+
+These impressive results are mainly attributed to MIRAS's convolutional backbone (i.e., ConvNeXt), which supports larger input images and enables semantic-assisted mask generation. This provides a solid foundation for achieving fine-grained segmentation in the next stage. However, this focus on task-specific patterns inherently introduces a trade-off, sacrificing some degree of generalization. We demonstrate in Appendix C.5 that MIRAS exhibits less degradation in generalized segmentation performance compared to other baseline models.
+
+# 5.5 Ablation Study
+
+To validate the effectiveness of the modules in MI-RAS, we conduct the following ablation experiments on the general RES task, ensuring compatibility with the following framework by maintaining the base model as LLaVA-v1. (1) Dual-visual Encoder Table 6 illustrates that the dual-visual encoder improves performance (Val  $\downarrow$ $2.6\%$ , Test  $\downarrow$ $0.9\%$ ) by supporting a higher resolution, which enhances the density of visual features and the ability to capture finer details. (2) Semantic Region Alignment The alignment strategy of injecting semantic information has achieved positive results, as shown in Table 6. When applied to the half dataset, it decreases Val by  $1.4\%$  and Test by  $0.5\%$ . Reducing the application to the full leads to further decline (Val  $\downarrow$ $0.7\%$ , Test  $\downarrow$ $0.4\%$ ), highlighting its effectiveness in enhancing segmentation.
+
+<table><tr><td rowspan="2">Architecture</td><td colspan="2">refCOCOg</td></tr><tr><td>Val (U)</td><td>Test (U)</td></tr><tr><td>MIRAS (v1)</td><td>73.2</td><td>71.9</td></tr><tr><td>w/o Dual-visual Encoder</td><td>70.6</td><td>70.8</td></tr><tr><td>w/o Semantic Region Alignment (50%)</td><td>71.8</td><td>71.4</td></tr><tr><td>w/o Semantic Region Alignment</td><td>71.1 (↓)</td><td>71.0 (↓)</td></tr></table>
+
+Table 6: Ablation. The metric is CIoU.  $50\%$  means half of the samples randomly added semantic information.
+
+# 6 Conclusion
+
+In this paper, we propose a novel task, Pixel-level Reasoning Segmentation, which focuses on fine-grained segmentation. To further advance, we construct a pixel-level reasoning segmentation dataset, PRIST, consisting of 24k utterances and 8.3k pixel-level segmentation targets, generated through a carefully designed three-stage progressive automatic annotation pipeline. Additionally, we present MIRAS, a framework designed for this task that combines segmentation with multi-turn interaction, along with LLM-based reasoning quality evaluation metrics. Comprehensive experiments on segmentation and reasoning demonstrate the effectiveness of the PRIST dataset and the superior performance of MIRAS, which advances research in pixel-level reasoning segmentation meaningfully.
+
+# Limitations
+
+Although our research has achieved certain advancements in the pixel-level RS task, some limitations remain. PRIST is designed only for single-
+
+target segmentation, making it difficult to adapt to more complex scenarios, such as those with empty targets (i.e., no objects requiring segmentation) or multiple targets (i.e., simultaneously involving multiple distinct objects). Further exploring reasoning trees to model relationships among image elements and constructing datasets for multi-object, multilevel segmentation hold research potential. Additionally, we utilize the SAM model to efficiently assist MLLM in integrating text reasoning to achieve pixel-level segmentation. However, their integration of separate visual encoding modules creates structural redundancy, reducing efficiency. Developing a streamlined and efficient model architecture is an important direction for future work.
+
+# Ethics Statement
+
+Pixel-level reasoning segmentation technology is a double-edged sword. On one hand, it demonstrates immense potential in fields (e.g., medical image analysis, autonomous driving, and intelligent surveillance), contributing to technological advancement and societal development. On the other hand, it is crucial to rigorously guard against risks related to privacy infringement and potential misuse. In our research, we meticulously selected key image features for constructing the training dataset to ensure the safety and representativeness of all samples. We employed an annotation process that aligns closely with ethical values for data collection, aiming to eliminate privacy breaches and the generation of harmful content. Before releasing the PRIST dataset, human experts will rigorously review all annotations and filter out inappropriate or risky data. Furthermore, users must agree to strict licensing terms to govern dataset usage. Importantly, although this technology excels in fine-grained visual understanding, it is not a substitute for human judgment. Its applications must always operate under human supervision, balancing innovation with ethical responsibility.
+
+# Acknowledgments
+
+Thanks to all co-authors for their hard work. The work is supported by the National Natural Science Foundation of China (62172086, 62272092), and the Fundamental Research Funds for the Central Universities under Grant (N25XQD004).
+
+# 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 arXiv:2303.08774.  
+Jinze Bai, Shuai Bai, Shusheng Yang, Shijie Wang, Sinan Tan, Peng Wang, Junyang Lin, Chang Zhou, and Jingren Zhou. 2023. Qwen-vl: A versatile vision-language model for understanding, localization, text reading, and beyond. Preprint, arXiv:2308.12966.  
+Satanjeev Banerjee and Alon Lavie. 2005. Meteor: 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.  
+Holger Caesar, Jasper Uijlings, and Vittorio Ferrari. 2018. Coco-stuff: Thing and stuff classes in context. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1209-1218.  
+Keqin Chen, Zhao Zhang, Weili Zeng, Richong Zhang, Feng Zhu, and Rui Zhao. 2023. Shikra: Unleashing multimodal llm's referential dialogue magic. arXiv preprint arXiv:2306.15195.  
+Qiguang Chen, Libo Qin, Jin Zhang, Zhi Chen, Xiao Xu, and Wanxiang Che. 2024a.  $m^3$  cot: A novel benchmark for multi-domain multi-step multi-modal chain-of-thought. arXiv preprint arXiv:2405.16473.  
+Xinlei Chen, Hao Fang, Tsung-Yi Lin, Ramakrishna Vedantam, Saurabh Gupta, Piotr Dollár, and C Lawrence Zitnick. 2015. Microsoft coco captions: Data collection and evaluation server. arXiv preprint arXiv:1504.00325.  
+Zhe Chen, Weiyun Wang, Hao Tian, Shenglong Ye, Zhangwei Gao, Erfei Cui, Wenwen Tong, Kongzhi Hu, Jiapeng Luo, Zheng Ma, et al. 2024b. How far are we to gpt-4v? closing the gap to commercial multimodal models with open-source suites. arXiv preprint arXiv:2404.16821.  
+Jacob Cohen. 1960. A coefficient of agreement for nominal scales. Educational and psychological measurement, 20(1):37-46.  
+Ross Girshick, Jeff Donahue, Trevor Darrell, and Jitendra Malik. 2014. Rich feature hierarchies for accurate object detection and semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 580-587.  
+Sahar Kazemzadeh, Vicente Ordonez, Mark Matten, and Tamara Berg. 2014. ReferItGame: Referring to objects in photographs of natural scenes. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 787-798, Doha, Qatar. Association for Computational Linguistics.
+
+Diederik P Kingma. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.  
+Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C. Berg, Wan-Yen Lo, Piotr Dólar, and Ross Girshick. 2023. Segment anything. arXiv:2304.02643.  
+Ranjay Krishna, Yuke Zhu, Oliver Groth, Justin Johnson, Kenji Hata, Joshua Kravitz, Stephanie Chen, Yannis Kalantidis, Li-Jia Li, David A Shamma, et al. 2017. Visual genome: Connecting language and vision using crowdsourced dense image annotations. International journal of computer vision, 123:32-73.  
+Xin Lai, Zhuotao Tian, Yukang Chen, Yanwei Li, Yuhui Yuan, Shu Liu, and Jiaya Jia. 2023. Lisa: Reasoning segmentation via large language model. arXiv preprint arXiv:2308.00692.  
+Jiwei Li, Michel Galley, Chris Brockett, Jianfeng Gao, and Bill Dolan. 2015. A diversity-promoting objective function for neural conversation models. arXiv preprint arXiv:1510.03055.  
+Yanwei Li, Yuechen Zhang, Chengyao Wang, Zhisheng Zhong, Yixin Chen, Ruihang Chu, Shaoteng Liu, and Jiaya Jia. 2023. Mini-gemini: Mining the potential of multi-modality vision language models. arXiv:2403.18814.  
+Hezheng Lin, Xing Cheng, Xiangyu Wu, and Dong Shen. 2022. Cat: Cross attention in vision transformer. In 2022 IEEE international conference on multimedia and expo (ICME), pages 1-6. IEEE.  
+Chang Liu, Henghui Ding, and Xudong Jiang. 2023. Gres: Generalized referring expression segmentation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 23592-23601.  
+Haotian Liu, Chunyuan Li, Yuheng Li, and Yong Jae Lee. 2024a. Improved baselines with visual instruction tuning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 26296-26306.  
+Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. 2024b. Visual instruction tuning. Advances in neural information processing systems, 36.  
+Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, and Saining Xie. 2022. A convnet for the 2020s. Preprint, arXiv:2201.03545.  
+Pan Lu, Swaroop Mishra, Tanglin Xia, Liang Qiu, Kai-Wei Chang, Song-Chun Zhu, Oyvind Tafjord, Peter Clark, and Ashwin Kalyan. 2022. Learn to explain: Multimodal reasoning via thought chains for science question answering. Advances in Neural Information Processing Systems, 35:2507-2521.
+
+Gerhard Neuhold, Tobias Ollmann, Samuel Rota Bulo, and Peter Kontschieder. 2017. The mapillary vistas dataset for semantic understanding of street scenes. In Proceedings of the IEEE international conference on computer vision, pages 4990-4999.  
+OpenAI. 2024. Hello gpt-4o.  
+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.  
+Zhiliang Peng, Wenhui Wang, Li Dong, Yaru Hao, Shaohan Huang, Shuming Ma, and Furu Wei. 2023. Kosmos-2: Grounding multimodal large language models to the world. arXiv preprint arXiv:2306.14824.  
+Renjie Pi, Jiahui Gao, Shizhe Diao, Rui Pan, Hanze Dong, Jipeng Zhang, Lewei Yao, Jianhua Han, Hang Xu, Lingpeng Kong, et al. 2023. Detgpt: Detect what you need via reasoning. arXiv preprint arXiv:2305.14167.  
+Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning transferable visual models from natural language supervision. Preprint, arXiv:2103.00020.  
+Vignesh Ramanathan, Anmol Kalia, Vladan Petrovic, Yi Wen, Baixue Zheng, Baishan Guo, Rui Wang, Aaron Marquez, Rama Kovvuri, Abhishek Kadian, et al. 2023. Paco: Parts and attributes of common objects. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7141-7151.  
+Hanoona Rasheed, Muhammad Maaz, Sahal Shaji, Abdelrahman Shaker, Salman Khan, Hisham Cholakkal, Rao M Anwer, Eric Xing, Ming-Hsuan Yang, and Fahad S Khan. 2024. Glamm: Pixel grounding large multimodal model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13009-13018.  
+Zhongwei Ren, Zhicheng Huang, Yunchao Wei, Yao Zhao, Dongmei Fu, Jiashi Feng, and Xiaojie Jin. 2024. Pixel reasoning with large multimodal model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 26374-26383.  
+Oleksii Sidorov, Ronghang Hu, Marcus Rohrbach, and Amanpreet Singh. 2020. Textcaps: a dataset for image captioning with reading comprehension. Preprint, arXiv:2003.12462.  
+Peng Wang, Shuai Bai, Sinan Tan, Shijie Wang, Zhihao Fan, Jinze Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Yang Fan, Kai Dang, Mengfei Du, Xuancheng Ren, Rui Men, Dayiheng Liu, Chang Zhou, Jingren Zhou, and Junyang Lin. 2024a.
+
+Qwen2-vl: Enhancing vision-language model's perception of the world at any resolution. arXiv preprint arXiv:2409.12191.  
+Wenhai Wang, Zhe Chen, Xiaokang Chen, Jiannan Wu, Xizhou Zhu, Gang Zeng, Ping Luo, Tong Lu, Jie Zhou, Yu Qiao, et al. 2024b. Visionllm: Large language model is also an open-ended decoder for vision-centric tasks. Advances in Neural Information Processing Systems, 36.  
+Zhaoqing Wang, Yu Lu, Qiang Li, Xunqiang Tao, Yandong Guo, Mingming Gong, and Tongliang Liu. 2022. Cris: Clip-driven referring image segmentation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11686-11695.  
+Zhuofan Xia, Dongchen Han, Yizeng Han, Xuran Pan, Shiji Song, and Gao Huang. 2024. Gsva: Generalized segmentation via multimodal large language models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3858-3869.  
+Zhao Yang, Jiaqi Wang, Yansong Tang, Kai Chen, Hengshuang Zhao, and Philip HS Torr. 2022. Lavt: Language-aware vision transformer for referring image segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18155-18165.  
+Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2024. Tree of thoughts: Deliberate problem solving with large language models. Advances in Neural Information Processing Systems, 36.  
+Haoxuan You, Haotian Zhang, Zhe Gan, Xianzhi Du, Bowen Zhang, Zirui Wang, Liangliang Cao, Shih-Fu Chang, and Yinfei Yang. 2023. Ferret: Refer and ground anything anywhere at any granularity. arXiv preprint arXiv:2310.07704.  
+Licheng Yu, Patrick Poirson, Shan Yang, Alexander C Berg, and Tamara L Berg. 2016. Modeling context in referring expressions. In Computer Vision-ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part II 14, pages 69-85. Springer.  
+Yuqian Yuan, Wentong Li, Jian Liu, Dongqi Tang, Xinjie Luo, Chi Qin, Lei Zhang, and Jianke Zhu. 2024. Osprey: Pixel understanding with visual instruction tuning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 28202-28211.  
+Xiang Yue, Yuansheng Ni, Kai Zhang, Tianyu Zheng, Ruoqi Liu, Ge Zhang, Samuel Stevens, Dongfu Jiang, Weiming Ren, Yuxuan Sun, et al. 2024. Mmmu: A massive multi-discipline multimodal understanding and reasoning benchmark for expert agi. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9556-9567.
+
+Ao Zhang, Liming Zhao, Chen-Wei Xie, Yun Zheng, Wei Ji, and Tat-Seng Chua. 2023a. Next-chat: An lmm for chat, detection and segmentation. arXiv preprint arXiv:2311.04498.
+
+Shilong Zhang, Peize Sun, Shoufa Chen, Min Xiao, Wenqi Shao, Wenwei Zhang, Yu Liu, Kai Chen, and Ping Luo. 2023b. Gpt4roi: Instruction tuning large language model on region-of-interest. arXiv preprint arXiv:2307.03601.
+
+Tao Zhang, Xiangtai Li, Hao Fei, Haobo Yuan, Shengqiong Wu, Shunping Ji, Chen Change Loy, and Shuicheng Yan. 2024. Omg-llava: Bridging image-level, object-level, pixel-level reasoning and understanding. arXiv preprint arXiv:2406.19389.
+
+Zhuosheng Zhang, Aston Zhang, Mu Li, Hai Zhao, George Karypis, and Alex Smola. 2023c. Multimodal chain-of-thought reasoning in language models. arXiv preprint arXiv:2302.00923.
+
+Zhaohui Zheng, Ping Wang, Wei Liu, Jinze Li, Rongguang Ye, and Dongwei Ren. 2020. Distance-iou loss: Faster and better learning for bounding box regression. In Proceedings of the AAAI conference on artificial intelligence, volume 34, pages 12993-13000.
+
+Bolei Zhou, Hang Zhao, Xavier Puig, Tete Xiao, Sanja Fidler, Adela Barriuso, and Antonio Torralba. 2019. Semantic understanding of scenes through the ade20k dataset. International Journal of Computer Vision, 127:302-321.
+
+Deyao Zhu, Jun Chen, Xiaoqian Shen, Xiang Li, and Mohamed Elhoseiny. 2023. Minigpt-4: Enhancing vision-language understanding with advanced large language models. arXiv preprint arXiv:2304.10592.
+
+# A Segmentation Task
+
+# A.1 Referring Expression Segmentation (RES)
+
+The referring expression segmentation (RES) task (Kazemzadeh et al., 2014) involves receiving an image and a natural language expression referring to a specific object in the image (e.g., "Please segment the apple in the image.") as input, and then outputting the segmentation mask of that object. As a classic task in the field of semantic segmentation, it intuitively reflects the model's ability in visual localization. Refcoco, Refcoco+, and Refcocog provide mature evaluation benchmarks for this task. To ensure fairness in the comparison, we choose to compare segmentation-specific baseline models and conduct ablation studies of the model architecture on the RES task. This is because all segmentation-specific baselines support this task and are trained on the aforementioned datasets.
+
+# A.2 Comparisons of Segmentation-specific MLLMs
+
+Table 7 provides a comprehensive comparison of recent segmentation-specific MLLMs in terms of model architecture, pixel-level capabilities, and conversation capabilities. A few works (Zhang et al., 2023b; Lai et al., 2023; Ren et al., 2024; Rasheed et al., 2024) integrate specialized vision modules and LMMs, as indicated by the Region Enc. / Dec. The End-End Model (Pi et al., 2023; Chen et al., 2023; Peng et al., 2023) distinction separates models that leverage LMMs for region-level understanding from those employing external modules.
+
+<table><tr><td>Method</td><td>Region Enc. / Dec.</td><td>Pixel-Level Seg. / Cap.</td><td>Conversation Multi-turn / Reason</td><td>End-End</td></tr><tr><td>VisionLLM (Wang et al., 2024b)</td><td>X/X</td><td>X/X</td><td>X/X</td><td>X</td></tr><tr><td>DetGPT (Pi et al., 2023)</td><td>X/X</td><td>X/X</td><td>✓/✓</td><td>✓</td></tr><tr><td>Shikra (Chen et al., 2023)</td><td>X/X</td><td>X/✓</td><td>X/X</td><td>✓</td></tr><tr><td>Kosmos-2 (Peng et al., 2023)</td><td>X/X</td><td>X/✓</td><td>X/X</td><td>✓</td></tr><tr><td>GPT4RoI (Zhang et al., 2023b)</td><td>✓/X</td><td>X/✓</td><td>✓/X</td><td>✓</td></tr><tr><td>LISA (Lai et al., 2023)</td><td>X/✓</td><td>✓/X</td><td>X/✓</td><td>X</td></tr><tr><td>PixelLM (Ren et al., 2024)</td><td>X/✓</td><td>✓/X</td><td>X/✓</td><td>✓</td></tr><tr><td>GLaMM (Rasheed et al., 2024)</td><td>✓/✓</td><td>✓/✓</td><td>X/✓</td><td>✓</td></tr><tr><td>OMG-LLaVA (Zhang et al., 2024)</td><td>X/✓</td><td>✓/✓</td><td>X/✓</td><td>✓</td></tr><tr><td>MIRAS (Ours)</td><td>X/✓</td><td>✓/✓</td><td>✓/✓</td><td>✓</td></tr></table>
+
+Table 7: Comparison of recent Segmentation-specific MLLMs.
+
+# B More Details about PRIST
+
+# B.1 Data Annotation
+
+We recruit experts from the Computer Science department as annotators for our project, as they are familiar with the task requirements and objectives. Each annotator is compensated at a rate of 30 euros per hour for their work. Before starting annotation, all annotators underwent training and a small subset of data was pre-annotated, which was only considered a pass if the accuracy rate was at least  $80\%$ . The subsequent annotation process was carried out once the pre-annotation results met the required standards. Additionally, We used the open-source tool Labelme<sup>+</sup>, which supports precisely outlining objects in arbitrary shapes and extracting masks, meeting the project's needs for accuracy.
+
+Annotation process: To avoid hallucinations and errors in the model-generated dialogues, we enforced strict quality control on the texts generated by GPT-4o. A total of 10 annotators were recruited, with a two-layer pyramid structure employed to reduce subjective bias and enhance quality. First, six annotators were divided into three pairs, each inde
+
+![](images/4168ad7f46c9e11751672452e0b6307d9b7e2e5d1a33369541fd671f86f2df77.jpg)  
+Figure 5: Wordcloud of the 200 popular focus-related words in PRIST.
+
+pendently annotating different data subsets. Next, an additional annotator per group resolved inconsistencies and checked the quality of consistent labels. Finally, an experienced annotator consolidated and reviewed all data to ensure high quality and consistency. Each annotator's main tasks included generating masks for target objects in images and correcting any commonsense errors in the dialogue content, such as mismatches in time or numbers between text and images.
+
+# B.2 Data Generation Process
+
+We design a fully automated dataset annotation pipeline, leveraging multiple hierarchical levels in the visual domain to construct the PRIST dataset. The pipeline, entirely based on GPT-4o (gpt-4o-2024-08-06), incorporates CoT into a feedback loop to generate relevant multi-turn reasoning dialogues for various images. Each dialogue's final query is a segmentation-related instruction (e.g., "Please segment the core objects according to the above dialogue"), making PRIST suitable for pixel-level segmentation tasks and extending to general VQA (when ignoring the final instruction), providing a versatile foundation for multimodal dialogue research. The data is used to fine-tune MLLMs. We will release the PRIST dataset and the implementation of its automated annotation pipeline to support further research. For detailed information on the implementation of LLM prompts and output formats at three levels, refer to Figure 7.
+
+# B.3 Dataset Statistics and Analysis
+
+As shown in Table 8, PRIST is compared with existing segmentation and multimodal reasoning benchmarks. PRIST combines the double advantages of pixel-level segmentation and multi-turn reasoning. In comparison with segmentation benchmarks (Yu et al., 2016; Krishna et al., 2017; Lai et al., 2023; Rasheed et al., 2024), the focus is on the granularity
+
+and richness of the segmentation targets. Figure 5 illustrates a visual word cloud of the 200 most popular focus-related terms extracted from PRIST. The prominence of each word in the cloud represents its frequency of occurrence in the dataset.
+
+Compared to existing multimodal reasoning benchmarks (Lu et al., 2022; Zhang et al., 2023c; Chen et al., 2024a), PRIST enables a more in-depth and detailed reasoning process for each segmented object. The average text length per sample is 477 tokens, significantly higher than the 294 tokens in  $\mathbf{M}^3\mathrm{CoT}$  (Chen et al., 2024a). Each conversation contains an average of 4 turns, with a maximum of 7, surpassing ScienceQA's average of 2.5 (Lu et al., 2022). PRIST better simulates real-world interactions and human thought processes while ensuring more coherent and detailed reasoning chains for pixel-level reasoning segmentation, presenting a new challenge for the RS field.
+
+# C Experiment Setups
+
+# C.1 Implementation Details
+
+Our experiments are all conducted on 2 NVIDIA A6000 (48G) GPUs. Aligned with region-level segmentation models (Rasheed et al., 2024; Zhang et al., 2024), we adopt LLaVAv1.6-7B as the backbone  $\mathcal{F}$  and ViT-H SAM to instantiate pixel-level encoder  $\nu_{\mathrm{pixel}}$  and mask decoder  $\mathcal{D}_m$ . The visual-language projection layer is implemented through 2-layer MLP and GELU activation following LLaVA-v1.6. The implementation is carried out in PyTorch, with Deepspeed Zero-2 optimization applied during two-stage training, and LoRA fine-tuning with  $r = 8$  for the LLM. Different experimental setups are adopted at each training stage to facilitate model convergence. Detailed training configurations are provided in Table 9.
+
+Backbone Selection In our framework, we adopt LLaVA v1.6 as the backbone due to its superior zero-shot performance compared to LLaVA v1.5\*. However, to ensure a fair comparison with other segmentation-specific MLLMs (Lai et al., 2023; Zhang et al., 2023a, 2024), we also conduct experiments using LLaVA v1.5 as the backbone. As shown in Table 10, the results indicate
+
+<table><tr><td>Benchmark</td><td>#Img.</td><td>#Reg.</td><td>#Samp.</td><td>Caption</td><td>Conversation Single. / Multi.</td><td>Segmentation Region. / Pixel.</td><td>Reasoning</td><td>#Step</td></tr><tr><td colspan="9">Segmentation Benchmark</td></tr><tr><td>RefCOCO (Yu et al., 2016)</td><td>20K</td><td>142K</td><td>142K</td><td>✓</td><td>✓/X</td><td>✓/X</td><td>X</td><td>-</td></tr><tr><td>VG (Krishna et al., 2017)</td><td>82.4K</td><td>3.8M</td><td>3.8M</td><td>✓</td><td>✓/X</td><td>✓/X</td><td>X</td><td>-</td></tr><tr><td>ReasonSeg (Lai et al., 2023)</td><td>1.2K</td><td>1.2K</td><td>1.2K</td><td>✓</td><td>✓/X</td><td>✓/X</td><td>✓</td><td>-</td></tr><tr><td>Osprey (Yuan et al., 2024)</td><td>100K</td><td>503K</td><td>724K</td><td>✓</td><td>✓/X</td><td>✓/✓</td><td>✓</td><td>-</td></tr><tr><td>GranD (Rasheed et al., 2024)</td><td>11M</td><td>810M</td><td>7.5M</td><td>✓</td><td>✓/✓</td><td>✓/✓</td><td>X</td><td>-</td></tr><tr><td>MUSE (Ren et al., 2024)</td><td>246K</td><td>910K</td><td>246K</td><td>✓</td><td>✓/X</td><td>✓/✓</td><td>✓</td><td>-</td></tr><tr><td colspan="9">Multimodal Reasoning Benchmark</td></tr><tr><td>ScienceQA (Lu et al., 2022)</td><td>5.6K</td><td>-</td><td>5.6K</td><td>✓</td><td>✓/✓</td><td>X/X</td><td>✓</td><td>2.5</td></tr><tr><td>MMMU (Yue et al., 2024)</td><td>11.5K</td><td>-</td><td>11.5K</td><td>✓</td><td>✓/X</td><td>X/X</td><td>✓</td><td>1.0</td></tr><tr><td>\( M^3COT \) (Chen et al., 2024a)</td><td>11K</td><td>-</td><td>11K</td><td>✓</td><td>✓/X</td><td>X/X</td><td>✓</td><td>10.9</td></tr><tr><td>PRIST (Ours)</td><td>2.8K</td><td>8.3K</td><td>8.3K</td><td>✓</td><td>✓/✓</td><td>✓/✓</td><td>✓</td><td>4.0</td></tr></table>
+
+Table 8: Comparison of existing segmentation and multimodal reasoning benchmarks. #X: the size of X, Img.: Image; Reg.: Segmentation Regions; Samp.: Conversational Samples; Step: Steps in the reasoning chain, averaged over all samples.  
+
+<table><tr><td>Config</td><td>Stage 1</td><td>Stage 2</td></tr><tr><td>Optimizer</td><td colspan="2">AdamW</td></tr><tr><td>Optimizer momentum</td><td colspan="2">β1=0.9, β2=0.95</td></tr><tr><td>Learning rate schedule</td><td colspan="2">WarmupDecayLR</td></tr><tr><td>Warmup iterations</td><td colspan="2">100</td></tr><tr><td>Weight decay</td><td colspan="2">0</td></tr><tr><td>Gradient accumulation steps</td><td colspan="2">10</td></tr><tr><td>Learning rate</td><td>3e-4</td><td>1e-5</td></tr><tr><td>Batch size</td><td>16</td><td>32</td></tr><tr><td>Training steps</td><td>50k</td><td>20k</td></tr></table>
+
+Table 9: The training settings of MIRAS.
+
+that the performance difference between the two versions is marginal, suggesting that both backbones are equally effective for our task. This consistency allows us to proceed with the latest LLaVA v1.6 while maintaining comparability with existing methods.
+
+# C.2 Two-Stage Training
+
+We employ a two-stage training process to achieve efficient pixel-level reasoning segmentation. The specific details of each stage are as follows:
+
+Stage 1: Mask-Text Alignment Pre-training The objective of this stage is to align mask-based region features with language embeddings. We collect mask-text pairs from various publicly available object-level datasets, including COCO (Chen et al., 2015), Ade20k (Zhou et al., 2019), and Mapillary (Neuhold et al., 2017), as well as region-level datasets like the RefCOCO series (Yu et al., 2016; Kazemzadeh et al., 2014) and PACO (Ramanathan et al., 2023). Additionally, we incorporate VQA
+
+and caption data, such as LLaVA-Instruct-80k (Liu et al., 2024a), COCO Caption(Chen et al., 2015). We mix these data in a 9:6:2:2 ratio and convert them into an instruction-following format for training, thereby enhancing its perceptual and conversational abilities.
+
+Stage 2: Pixel-level Segmentation Fine-tuning At this stage, we maintain fixed model weights while fine-tuning our PRIST dataset to enhance fine-grained segmentation and reasoning capabilities. Subsequently, we integrate VQA data (Liu et al., 2024a) to enable MIRAS to follow user instructions (i.e., PRIST combined with VQA at a 4:1 ratio), enhancing its ability to handle complex pixel-level segmentation with precision.
+
+# Segmentation Prompt
+
+Output Format: <box>(x1, y1), (x2, y2)</box>
+
+Please box out the position of focus and output the detection box in  $<\text{box}>$  (x1, y1), (x2, y2)  $<\text{box}>$  format, with coordinates representing the top-left and bottom-right corners of the detected area. If no focus is detected, return "no detected", and ensure a blank space follows all outputs.
+
+# C.3 General MLLMs Baselines
+
+To evaluate the advanced general MLLMs on our task, we guide them with a unified segmentation prompt. Since these models do not directly generate masks, we first extract the bounding box coordinates and convert them into masks using a con
+
+<table><tr><td rowspan="2">Model</td><td rowspan="2">CIoU</td><td colspan="3">Pixel-wise</td><td colspan="5">Response</td></tr><tr><td>Prec.</td><td>Recall</td><td>F1</td><td>BLEU-4</td><td>Dist-1/2</td><td>ROU_L.</td><td>MET.</td><td>BERTS.</td></tr><tr><td colspan="10">Stage-1</td></tr><tr><td>MIRAS(v1.5)</td><td>12.35</td><td>15.42</td><td>49.93</td><td>22.53</td><td>3.98</td><td>7.0 / 28.6</td><td>26.84</td><td>27.98</td><td>87.32</td></tr><tr><td>MIRAS(v1.6)</td><td>13.12</td><td>15.64</td><td>45.11</td><td>23.22</td><td>4.17</td><td>6.9 / 28.4</td><td>25.94</td><td>28.18</td><td>87.54</td></tr><tr><td colspan="10">Stage-2</td></tr><tr><td>MIRAS(v1.5)</td><td>14.54</td><td>23.01</td><td>40.59</td><td>31.86</td><td>8.47</td><td>14.9 / 49.2</td><td>30.66</td><td>41.15</td><td>88.38</td></tr><tr><td>MIRAS(v1.6)</td><td>14.72</td><td>24.22</td><td>40.61</td><td>30.34</td><td>8.51</td><td>15.7 / 49.2</td><td>30.82</td><td>40.06</td><td>88.47</td></tr></table>
+
+Table 10: The MLLM backbone selection of MIRAS.
+
+sistent function. The generated masks are then compared with ground truth masks. For fairness, we exclude the last round of dialogue in the final evaluation, as the models were not trained with identical segmentation instructions.
+
+# C.4 Evaluation
+
+# C.4.1 Pixel-level Segmentation Metrics
+
+We adopt traditional methods for calculating precision and recall, applying them to the pixel-wise task. Specifically, each pixel is treated as a binary classification (i.e., 0,1) to quantify the model's ability in pixel-level segmentation.
+
+Pixel-wise Precision measures the proportion of true positive samples among the pixels predicted as positive, reflecting the model's prediction accuracy. Its increase indicates the model can predict the pixels of the target region with higher confidence, thereby capturing the target more precisely.
+
+$$
+P r e c i s i o n = \frac {T P}{T P + F P}
+$$
+
+where  $TP$  represents pixels correctly predicted as the target, while  $FP$  represents non-target pixels incorrectly predicted as the target.
+
+Pixel-wise Recall measures the proportion of actual positive samples that are correctly predicted as positive, reflecting the model's coverage of the target area. Its decrease indicates that the model is more strictly segmenting according to the target boundaries, thereby avoiding overgeneralization.
+
+$$
+R e c a l l = \frac {T P}{T P + F N}
+$$
+
+where  $TP$  represents the pixels correctly predicted as the target, while  $FN$  represents the non-target pixels that are incorrectly predicted as the target.
+
+# C.4.2 LLM-based Metrics
+
+Traditional metrics (e.g., BLEU-n, METEOR) often fail to capture deep semantics and logical coherence in dialogue. To address this, we employ GPT-4 as a scoring tool to evaluate the model's reasoning
+
+quality using four key metrics: 1) Progressiveness (PR): Whether the current turn effectively sets up and guides the next. 2) Logical Coherence (LC): Whether there is a smooth logical connection between the current and subsequent turns. 3) Content Consistency (CC): Whether the dialogue revolves around the overall topic. 4) Target Relevance (TR): Whether the dialogue stays focused on the target. We carefully designed prompts for each metric to ensure fairness and consistency. To ensure fairness and consistency, we carefully designed prompts for each metric. Ratings are assigned on a 5-point scale divided into three intervals (0-1, 2-3, 4-5), with clearly defined criteria for each range. The final score is calculated as the average across all four metrics, minimizing randomness and bias. Higher scores indicate better performance, reflecting overall improvement across evaluated aspects. GPT-4 strictly adheres to these guidelines, evaluating each dimension step by step. Notably, when using an LLM for scoring, to eliminate randomness, we repeat the scoring for each metric 5 times and calculate the average to determine the final score. Detailed evaluation prompts are shown in Figure 8.
+
+# C.4.3 Human Response
+
+We randomly select 100 multi-turn conversations from the PRIST test set and invite three experts in computer science to respond masking the original answers. They all possess solid analytical skills and a deep understanding of model reasoning tasks. To ensure fairness in the evaluation, we also used the LLM-based evaluation criteria to score the three responses (as shown in Table 11). Finally, the average score from the three experts is used as the human evaluation value for each metric.
+
+Additionally, we leverage GPT-4 as a judge to design an adversarial evaluation framework, using Win Rate (\%) to assess reasoning quality in multi-turn dialogues. Human responses serve as the standard for model output comparison, the model is considered to achieve a win when its score sur
+
+<table><tr><td></td><td>PR</td><td>LC</td><td>CC</td><td>TR</td></tr><tr><td>Expert 1</td><td>4.06</td><td>4.10</td><td>4.38</td><td>4.29</td></tr><tr><td>Expert 2</td><td>3.98</td><td>4.02</td><td>4.21</td><td>4.31</td></tr><tr><td>Expert 3</td><td>4.05</td><td>4.00</td><td>4.19</td><td>4.24</td></tr><tr><td>Human</td><td>4.03</td><td>4.04</td><td>4.26</td><td>4.28</td></tr></table>
+
+Table 11: The reasoning scores of three human experts based on LLM-based metrics.  
+
+<table><tr><td>User: What text is visible on the wine bottle label?
+MIRAS: The text on the wine bottle label reads
+  \Vegalibrate\&quot;.
+User: How does &#x27;UTIEL-REQUENA&#x27; contribute to
+identifying the wine?
+MIRAS: &#x27;UTIEL-REQUENA&#x27; suggests a specific brand or
+style of the wine.</td><td>Prec. 50.4%
+Recall 98.0%</td></tr><tr><td>User: What text is visible on the wine bottle label?
+MIRAS: The label reads &#x27;UTIEL REQUENA&#x27; and
+&#x27;VEGALIBRE&#x27;.
+User: How does &#x27;UTIEL-REQUENA&#x27; contribute to
+identifying the wine?
+MIRAS: The term &quot;UTIEL-REQUENA&quot; on the label is a
+reference to the winery or vineyard that produced the wine.
+It is a way to identify the source of the wine and to
+showcase the winery&#x27;s brand or identity.</td><td>Prec. 75.1%
+Recall 87.6%</td></tr></table>
+
+Figure 6: Qualitative results of the performance of MIRAS on Pixel-level RS. Top: Stage-1 segments the entire bottle. vs. Bottom: Stage-2 only segments the label and provides more accurate and comprehensive interpretations of the "UTIEL-REQUENA" label.
+
+passes the human response score.
+
+# C.5 Generalization Trade-off
+
+Compared with other baselines, MIRAS exhibits a relatively smaller decline in Tabel 12, which indicates that our method has a better generalization ability than other baselines. This is primarily due to the increased complexity of pixel-level RS compared to traditional referring segmentation tasks, which drives the model to prioritize precise boundary refinement during fine-tuning. General segmentation models typically emphasize detection comprehensiveness, sometimes at the expense of precision (e.g., segmenting a wine bottle label as the entire bottle). In contrast, our PRIST dataset focuses on fine-grained targets, enhancing segmentation accuracy by refining complex boundaries. However, this task-specific fine-tuning slightly reduces the model's generalization ability to broader segmentation tasks. We will further explore this trade-off in future research.
+
+# D Case Study
+
+To reflect the high quality of our constructed dataset, we present examples of PRIST test data in Figure 9, which effectively combine coherent and logically structured multi-turn conversations with fine-grained segmentation. We fine-tuned MIRAS on the PRIST dataset to enhance its understanding
+
+and localization of image details. As shown in Figure 6, the fine-tuned MIRAS (Stage-2) segments only the label, replacing the entire bottle in Stage-1, demonstrating stronger segmentation specificity (with a significant improvement in precision) while effectively mitigating the issue of overgeneralization (with a slight decrease in recall). This result also highlights the potential of PRIST in improving pixel-level reasoning capabilities.
+
+To more intuitively demonstrate the advantages of our model in the pixel-level reasoning segmentation task, we also present several cases of user interactions with the chatbot. As shown in Figure 10, these cases illustrate MIRAS's coherence and consistency in multi-turn interactions and showcase its ability to achieve more precise fine-grained object segmentation.
+
+1. identify and list the elements visible in an image. \  
+2. Provide detailed descriptions of each element, with text, numbers, and other information.  
+3. List at least 3 elements visible in the image.
+
+Role: Visual Extractor  
+Requirements and Goals:  
+Output Format:  
+(a) Illustration of the visual elements extraction (step-1).  
+# Role: Visual Reasoning Expert  
+#Scoring Criteria:  
+```json
+{"elements":[{name": "The name of the element1",text": "Text description of the element1",number": (optional)"Number of the element",caption": "Caption of the element}},...]}
+```
+
+1 Point (Avoid these): Abstract or vague questions, or too simple or too complex (not intuitive).  
+2 Points: Direct questions that can be answered through basic reasoning based on visible elements.  
+3 Points: Clear questions that require reasoning using details and commonsense, focusing on specific objects.
+
+1. Design 3 reasoning questions based on the listed elements and details from the image and caption, with scores of 2-3 points.  
+2. Each question must involve reasoning and use common sense to analyze visible elements.  
+3. Display a reasoning tree structure: Present the reasoning tree structure showing the progression of reasoning for each question. Highlight any overlapping nodes or steps among the three questions to reflect common reasoning pathways.  
+4. Output the corresponding reasoning questions for each of the three paths.
+
+Requirements and Goals:  
+Output Format:  
+(b) Illustration of the reasoning tree construction (step-2).  
+# Role: Multi-Turn Reasoning Dialogue Generator  
+Requirements and Goals:  
+```jsonl
+{"root_node": "Root Node A",  
+"levels": ["First Level: Level 1",  
+"Second Level: Level 2"],  
+"Inference_nodes": {[  
+    "node_name": "Inference Node 1",  
+"overlap": "Overlap of Path 1 and 3", // "Overlap of Path x and y" or "Path x Independent"  
+"reasoning_paths": [{  
+    "path_name": "Path 1",  
+"path_description": "Reasoning Path 1 Description",  
+"reasoning_question": "Reasoning Question for Path 1"}],  
+}，{  
+"node_name": "Inference Node n",  
+"overlap": "Path 2 Independent",  
+"reasoning Paths": [{  
+    "path_name": "Path 2",  
+"path_description": "Reasoning Path 2 Description",  
+"reasoning_question": "Reasoning Question for Path 2"}]  
+}
+```
+
+1. Each dialogue must consist of 4 to 8 turns of questions and answers.  
+2. Each question should focus on a single, specific object visible in the image, with reasoning directly tied to the elements listed in the title and the image details.  
+3. Expand each reasoning path into a multi-round dialogue format according to the provided reasoning tree, demonstrating the logical progression of thought based on the identified paths.
+
+Output Format:  
+(c) Illustration of the multi-turn dialogue generation (step-3).  
+```jsonl
+{"Question1": { "Q1": "<First Question>", "A1": "<Answer to First Question>", "Q2": "<Second Question>", "A2": "<Answer to Second Question>", "Q3": "<Third Question>", "A3": "<Answer to Third Question>", ...}, "Focus": "<Final Focus Object1>", ... }
+```
+
+Figure 7: The prompts and output formats of our dataset annotation pipeline.
+
+<table><tr><td rowspan="2">Model</td><td colspan="3">refCOCO</td><td colspan="3">refCOCO+</td><td colspan="2">refCOCOg</td></tr><tr><td>Val</td><td>TestA</td><td>TestB</td><td>Val</td><td>TestA</td><td>TestB</td><td>Val</td><td>Test</td></tr><tr><td>LISA</td><td>74.1</td><td>76.5</td><td>72.3</td><td>65.1</td><td>70.8</td><td>58.1</td><td>67.9</td><td>70.6</td></tr><tr><td>PixelLM</td><td>73.0</td><td>76.5</td><td>68.2</td><td>66.3</td><td>71.7</td><td>58.3</td><td>69.3</td><td>70.5</td></tr><tr><td>OMG-LLAVA</td><td>78.0</td><td>80.3</td><td>74.1</td><td>69.1</td><td>73.1</td><td>63.0</td><td>72.9</td><td>72.9</td></tr><tr><td>MIRAS</td><td>78.4</td><td>80.5</td><td>73.4</td><td>72.1</td><td>74.8</td><td>63.4</td><td>72.6</td><td>72.0</td></tr><tr><td colspan="9">Fine-tuned</td></tr><tr><td>LISA</td><td>72.9 (↓1.2)</td><td>75.2 (↓1.3)</td><td>72.0 (↓0.3)</td><td>64.3 (↓0.8)</td><td>68.2 (↓2.6)</td><td>56.9 (↓1.2)</td><td>67.5 (↓0.4)</td><td>69.4 (↓1.2)</td></tr><tr><td>PixelLM</td><td>71.2 (↓1.8)</td><td>75.8 (↓0.7)</td><td>66.5 (↓1.7)</td><td>63.2 (↓3.1)</td><td>70.5 (↓1.2)</td><td>57.0 (↓1.3)</td><td>69.0 (↓0.3)</td><td>68.9 (↓1.6)</td></tr><tr><td>OMG-LLAVA</td><td>77.3 (↓0.7)</td><td>79.1 (↓1.2)</td><td>73.5 (↓0.6)</td><td>68.0 (↓1.1)</td><td>72.2 (↓2.6)</td><td>62.7 (↓1.3)</td><td>71.3 (↓1.6)</td><td>70.0 (↓2.9)</td></tr><tr><td>MIRAS</td><td>76.9 (↓1.5)</td><td>79.8 (↓0.7)</td><td>72.8 (↓0.6)</td><td>68.8 (↓3.3)</td><td>74.4 (↓0.4)</td><td>62.5 (↓0.9)</td><td>71.8 (↓0.8)</td><td>70.8 (↓1.2)</td></tr></table>
+
+Table 12: Comparison of baselines's generalization trade-off.
+
+# # Role: A Multiturn Dialogue Reasoning Evaluator
+
+# # Requirements:
+
+1. Please analyze the following multi-turn conversation and score it based on the following dimensions:
+
+- Progression: Does the conversation build logically from one turn to the next? Is the topic being developed effectively?  
+- Logical Coherence: Is there a clear and natural logical connection between each turn in the conversation?  
+- Content Consistency: Does each turn align with the overall topic and conversation goal?  
+- Focal Goal Relevance: Does each turn focus on and contribute towards achieving the focal goal of the conversation?
+
+After providing scores for each dimension, include a summary that captures the strengths and weaknesses of the conversation.
+
+2. Only output the reasoning scores in the above format.  
+3. Please evaluate the entire process to the content of the image.  
+4. Please focus on linking the dimensions between each round of conversation.
+
+# Evaluation Criteria:
+
+1. Progression Score:
+
+4-5 points: Smoothly transitions to the next turn with sufficient setup for subsequent content.  
+2-3 points: Partially advances the dialogue, but the connection feels somewhat awkward.  
+0-1 points: Fails to advance the dialogue or is disconnected from the next turn.
+
+2. Logical Coherence Score:
+
+4-5 points: Perfectly connected with rigorous logic.  
+2-3 points: Generally clear logic, but with minor gaps or weak links.  
+0-1 points: No logical connection or logically incoherent.
+
+3. Content Consistency Score:
+
+4-5 points: Content is fully aligned and tightly focused on the goal.  
+2-3 points: Generally consistent, but with a slight deviation.  
+0-1 points: Deviates from the topic, losing the direction of the dialogue.
+
+4. Focal Goal Relevance Score:
+
+4-5 points: Highly relevant and continuously advances the goal.  
+2-3 points: Somewhat relevant, but with limited contribution.  
+0-1 points: Deviates from the goal, failing to advance it.
+
+# Output in JSON Format:
+
+Your output should be a JSON object containing scores for each dimension, as well as the total score and a brief summary of the evaluation.
+
+```json
+{
+    "scores": [
+        {"name": "progression_score", "score": 4.5},
+        {"name": "logical_coherence_score", "score": 4.0},
+        {"name": "content_consistency_score", "score": 4.2},
+        {"name": "focal_goal_relevance_score", "score": 4.8},
+        {"name": "total_score", "score": 17.5},
+    ],
+    "summary": "The conversation demonstrates excellent progression and relevance to the focal goal. \ However, there are some minor coherence issues and slight deviations from the main topic."
+```
+
+Figure 8: Illustration of the LLM-based evaluation of reasoning quality.
+
+Q1: What information is displayed on the bus's front panel?  
+A1: The front of the bus displays the route number '12' and the destination 'CARISA-HOSPITAL'.  
+Q2: How can the route number and destination help identify the bus's route?  
+A2: The route number '12' and the destination 'CARISA-HOSPITAL' indicate the specific path the bus takes, which is directed towards the Carisa Hospital.  
+Q3: Using the conversation above, create a segmentation that validates the answers.  
+A3: The result is [OBJ]bus front display[SEG].
+
+Name: Bus
+
+Text: A white bus on route number 12 with 'PARQUE PRINCIPADO - CAU SA-HOSPITAL' displayed on the front panel. It has logos for ALSA and BYS on it and is crossing a zebra crossing.
+
+Caption: White bus with route information crossing
+
+Number: 12
+
+![](images/2a04a02092a399d034713dfd7b9f3939277a74893fee5ce0311a7c198d8678bc.jpg)
+
+Name: Traffic cones
+
+Text: Two orange-and-white traffic cones placed on the road near a road construction or manhole.
+
+Caption: Orange-and-white traffic cones at a construction site.
+
+Number: 2
+
+![](images/de9b3361e4d6a334c096fefbf226499ff7bde9e61bb01d43a2d36d37257537f4.jpg)
+
+Q1: What elements are present near the bus that suggest a road safety feature?
+
+A1: Traffic cones are present near the bus, suggesting a road safety feature.  
+Q2: How do these traffic cones indicate an area of maintenance or hazard?  
+A2: Traffic cones are typically used to signal areas under maintenance or potential hazards, warning pedestrians and drivers to be cautious.
+
+Q3: Please use the conversation above to outline a segmentation that aligns with the answers provided.
+
+A3: The result is [OBJ]traffic cones[SEG].
+
+Q1: What text is visible on the bus that indicates its destination?  
+A1: The text  
+'CARISA-HOSPITAL' is visible on the bus, which indicates its destination.  
+Q2: Why is this information important for passengers?  
+A2: This information helps passengers identify the correct bus that will take them to Carisa Hospital.  
+Q3: Segment the provided conversation to match the answers accordingly.  
+A3: The segmentation is [OBJ]text 'CARISA-HOSPITAL' [SEG].
+
+Name: 10, 20 Euro Cent Coin
+
+Text: Two gold colored Euro cent coins with an image of Europe and the values are 10 and 20.
+
+Caption: Two Euro cent coins of 10 and 20.
+
+Number: 2
+
+![](images/51e2cecd5a5e62b1d4c13c2b4967054b07fd07444dfb01c521d6b5824a6f9ad7.jpg)  
+Figure 9: PRIST Samples. The figure displays the PRIST dataset, created through the automated annotation pipeline. It offers a range of semantic labels, multi-turn conversations and attributes for the identified objects.
+
+Name: 2 Euro Cent Coin
+
+Text: Two copper colored 2 Euro cent coins showing the number 2 and an image of Europe.
+
+Caption: 2 Euro Cent Coin
+
+Number: 2
+
+Q1: What is displayed on the larger silver-colored coin?  
+A1: The number 2 is prominently displayed on the largest silver-colored coin.  
+Q2: What does this number indicate regarding the coin's value?  
+A2: It indicates that the coin has a value of 2 Euros.  
+Q3: How does the prominent number assist in determining the value across all coins?  
+A3: The prominent number signifies the denomination directly, aiding in value identification.  
+Q4: Use the above conversation to create a segmentation that justifies the answers.  
+A4: The segmentation is [OBJ] 2 Euro silver-colored coin[SEG].
+
+Q1: What numbers are prominently displayed on each gold-colored coin?  
+A1: The gold-colored coins have the numbers 20 and 10 prominently displayed.  
+Q2: Which unique design element helps identify the value of these coins?  
+A2: The specific numbers and the Euro cent symbol help identify their value.  
+Q3: How does the unique design confirm the coin values?  
+A3: The design and numbers indicate the coins are 20 and 10 Euro cents.  
+Q4: Please segment the conversation above to align with the provided answers.  
+A4: Segmented as [OBJ]gold-colored coins with numbers 20 and 10[SEG].
+
+![](images/4bd67d4cd010147b0ad30aa4ae384bb1fca50c9c0ec61a853e35e1a03594e99c.jpg)
+
+:aequnN
+
+o000000000000000000000000000000000000000000000
+
+e 100
+
+e e e e e e e e e e e e e e e e e e e e e e
+
+![](images/64d3ca12415098b8d9e04fc260640da18d07155e29a6f2de4f2970a6a78ce150.jpg)  
+Figure 10: Pixel-level segmentation and multi-turn conversational interactions facilitated by MIRAS. Left: The model segments only the watch face instead of the entire watch, and identifies the three sub-dials and the "SINCE 1953" logo. Center: The model segments only the emblem on the coin instead of the whole coin, and infers that it is related to Soviet communism. Right: The model segments only the road sign, and accurately recognizes the text "Plaza de San Lorenzo".
+
+![](images/8495dab4826594abb262bf51e9087fa9b4101c4ef2173f53aaef283a58e1776f.jpg)
+
+![](images/b18c3ef9f35e9d6c1bf8748ee2d225bc4faa7d87ee78fb2b4079baa8dc21418a.jpg)

ACL/2025/Pixel-Level Reasoning Segmentation via Multi-turn Conversations/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c12048ee02eabf3cb356c20c288bfc52ab6fabf670283372425a114cf8c54f1
+size 1389797

ACL/2025/Pixel-Level Reasoning Segmentation via Multi-turn Conversations/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:200287a6ddddeb6517ec9a5363ecfce8efc2810cd4a088466ed94fb3e15d01e4
+size 710542

ACL/2025/PlanGenLLMs_ A Modern Survey of LLM Planning Capabilities/40cffb0f-ccac-42db-91ae-d00226630002_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb4c35e3627597881c3c2de53992c223785120e71821e20b9bee586fe5b60258
+size 154310

ACL/2025/PlanGenLLMs_ A Modern Survey of LLM Planning Capabilities/40cffb0f-ccac-42db-91ae-d00226630002_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:562419246c3c20a82424e4626df3a3cdc5d7279e53d4585c86109c2cfa58ddf0
+size 191293

ACL/2025/PlanGenLLMs_ A Modern Survey of LLM Planning Capabilities/40cffb0f-ccac-42db-91ae-d00226630002_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3fa1ab746a0412c0f29443ba23ac57695328a68645701747efc2d81eb7ca8801
+size 474134

ACL/2025/PlanGenLLMs_ A Modern Survey of LLM Planning Capabilities/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcccbcbb4160b1b90b26b33c100fc1601ab11d53812d8ad1d96d8b46ccf062f4
+size 3257792

ACL/2025/PlanGenLLMs_ A Modern Survey of LLM Planning Capabilities/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95e6528f13c6888062cfb9956a14bf647aacd423dba71bcd54f781b89b6ec09b
+size 508613

ACL/2025/Planning with Diffusion Models for Target-Oriented Dialogue Systems/b24cfded-5e60-49f2-bd4f-f86f7532ca3a_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f328936393483b5d1b02d5ce8bd801d86980bceae30f466ff056e077784f83c
+size 173992

ACL/2025/Planning with Diffusion Models for Target-Oriented Dialogue Systems/b24cfded-5e60-49f2-bd4f-f86f7532ca3a_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9781ea8b6e5e208f24caed6fd1dc8cc473d396b778b18376ce868cabfd0d5f75
+size 206916

ACL/2025/Planning with Diffusion Models for Target-Oriented Dialogue Systems/b24cfded-5e60-49f2-bd4f-f86f7532ca3a_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50ac71cb4cf3080315361ed9a7e339b213cfff5959a3f3d68910851b4abdfad9
+size 1359364

ACL/2025/Planning with Diffusion Models for Target-Oriented Dialogue Systems/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4377f63de3e7af20a641d2dfc76dad5068aba296c05a2fe4fc14c31411fd83c9
+size 898298

ACL/2025/Planning with Diffusion Models for Target-Oriented Dialogue Systems/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b518ae2981dec11314d59c2490de960424e862e9647f517f421e87067f2f92d
+size 818236

ACL/2025/Planning-Driven Programming_ A Large Language Model Programming Workflow/e6a16d07-58a4-4795-8ea2-5b6780a5f321_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4a596f17410f29c8038450d1bbb4d63a27fef1ca72613ff95efd0e094599642
+size 224723

ACL/2025/Planning-Driven Programming_ A Large Language Model Programming Workflow/e6a16d07-58a4-4795-8ea2-5b6780a5f321_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1bf55adf82280087690f91b6ad895efbd8a6284079c6ff9d7d3226a4bd8c7c1c
+size 268905

ACL/2025/Planning-Driven Programming_ A Large Language Model Programming Workflow/e6a16d07-58a4-4795-8ea2-5b6780a5f321_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98abd1ab71b45e31b32ccdef2edf4d1ac61cce6efb69492f1ba2380f8b517d0b
+size 8677223

ACL/2025/Planning-Driven Programming_ A Large Language Model Programming Workflow/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d281316b15a3dc0422ea1d1e10bc943739e998aec95f6697d75a3199b50a3e0d
+size 1764378

ACL/2025/Planning-Driven Programming_ A Large Language Model Programming Workflow/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f1fb36b2686ab2eb30588499c4066613afa06d8f7ef2150e9ddca658c511287
+size 1066350

ACL/2025/PlanningArena_ A Modular Benchmark for Multidimensional Evaluation of Planning and Tool Learning/d18e3516-c016-46b7-850b-dc82e0bb821e_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fe4442be0dae52b9b856e7e7c160f71bfb1683c2c46aa81221278b3e5e6f8b9
+size 206471

ACL/2025/PlanningArena_ A Modular Benchmark for Multidimensional Evaluation of Planning and Tool Learning/d18e3516-c016-46b7-850b-dc82e0bb821e_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8c9e7f053f01c2952daa902e3b672ba4802f408a64497042761d66bdbd692e9
+size 241538

ACL/2025/PlanningArena_ A Modular Benchmark for Multidimensional Evaluation of Planning and Tool Learning/d18e3516-c016-46b7-850b-dc82e0bb821e_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8af08eb79c5d5af2f6ddd488c86446331182261c1bc674133f4080b1a32c810a
+size 10330188

ACL/2025/PlanningArena_ A Modular Benchmark for Multidimensional Evaluation of Planning and Tool Learning/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8e35f44a141479db23d70b1cc93ace69cc6d2e6d98d78c82671baa4dc51959e
+size 1527299

ACL/2025/PlanningArena_ A Modular Benchmark for Multidimensional Evaluation of Planning and Tool Learning/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d1fa5df4b1a27e6c5aa69d66cd588d8be4b5649e964940ff57e7c3c7c3bb06b
+size 755742

ACL/2025/Plug-in and Fine-tuning_ Bridging the Gap between Small Language Models and Large Language Models/db05b1c2-b915-4599-8bca-e918603dc3be_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c589e741fff98e6cfce54e30b1a8c188eeb2ff858ee9ee98de2b51bfad9f5259
+size 130254

ACL/2025/Plug-in and Fine-tuning_ Bridging the Gap between Small Language Models and Large Language Models/db05b1c2-b915-4599-8bca-e918603dc3be_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74b7c72a7182ee1cc219bf37c769f58547e4f973bd5643b007b844cf36470c53
+size 157803

ACL/2025/Plug-in and Fine-tuning_ Bridging the Gap between Small Language Models and Large Language Models/db05b1c2-b915-4599-8bca-e918603dc3be_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64128bc6b721a0e700858a74db901e4d5678366d590509dd676795a90bb45a10
+size 761552

ACL/2025/Plug-in and Fine-tuning_ Bridging the Gap between Small Language Models and Large Language Models/full.md

@@ -0,0 +1,484 @@
+# Plug-in and Fine-tuning: Bridging the Gap between Small Language Models and Large Language Models
+
+Kyeonghyun Kim $^{1*}$ , Jinhee Jang $^{1*}$ , Juhwan Choi $^{2*\dagger}$ , Yoonji Lee $^{1}$ , Kyohoon Jin $^{3\dagger}$ , YoungBin Kim $^{1}$
+
+$^{1}$ Chung-Ang University  $^{2}$ AITRICS  $^{3}$ DATUMO
+
+$^{1}\{khyun8072, jinheejiang, pioneer0305, ybkim85\} @cau.ac.kr$
+
+$^{2}$ jhchoi@aitrics.com  $^{3}$ kyohoon.jin@selectstar.ai
+
+# Abstract
+
+Large language models (LLMs) are renowned for their extensive linguistic knowledge and strong generalization capabilities, but their high computational demands make them unsuitable for resource-constrained environments. In contrast, small language models (SLMs) are computationally efficient but often lack the broad generalization capacity of LLMs. To bridge this gap, we propose PiFi, a novel framework that combines the strengths of both LLMs and SLMs to achieve high performance while maintaining efficiency. PiFi integrates a single frozen layer from an LLM into a SLM and fine-tunes the combined model for specific tasks, boosting performance without a significant increase in computational cost. We show that PiFi delivers consistent performance improvements across a range of natural language processing tasks, including both natural language understanding and generation. Moreover, our findings demonstrate PiFi's ability to effectively leverage LLM knowledge, enhancing generalization to unseen domains and facilitating the transfer of linguistic abilities.
+
+# 1 Introduction
+
+Language models (LMs) based on transformer architecture have demonstrated impressive performance across a wide range of natural language processing (NLP) tasks, primarily due to the linguistic knowledge they acquire from training on large-scale datasets (Zhao et al., 2023). In particular, large language models (LLMs), such as GPT-3 (Brown et al., 2020), GPT-4 (Achiam et al., 2023), and Llama-3 (Dubey et al., 2024), stand out for their significantly larger number of parameters compared to small language models (SLMs), which typically contain between 100M and 5B parameters (Lu et al., 2024). Notable examples of SLMs
+
+include BERT (Devlin et al., 2019), RoBERTa (Liu et al., 2019), and BART (Lewis et al., 2020). The extensive parameterization of LLMs allows them to capture a broader range of knowledge, leading to enhanced generalizability across novel tasks, whereas SLMs often require fine-tuning for specific tasks or domains (Brown et al., 2020; Wei et al., 2022; Ye, 2024).
+
+Despite the notable strengths and exceptional performance of LLMs, they come with inherent limitations. A major constraint is the trade-off between performance and inference cost, primarily due to the computational resources required to deploy these large models (Shashidhar et al., 2023). As the number of parameters increases, so does the demand for computational power and memory, making LLMs less suitable for environments with limited resources, such as mobile devices (Nityasya et al., 2020; Lin et al., 2023). Similarly, fine-tuning LLMs for domain-specific applications involves a substantial computational overhead (Dettmers et al., 2023; Zhang et al., 2024a).
+
+In such resource-constrained scenarios, SLMs often provide a more viable solution. Studies have shown that fine-tuned SLMs can outperform LLMs in specific tasks such as sentiment analysis, semantic textual similarity evaluation, and named entity recognition (Yu et al., 2023; Lepagnol et al., 2024). Their lightweight architecture makes them an appealing choice for scenarios where computational efficiency and reduced memory usage are critical (Gao et al., 2023b; Lepagnol et al., 2024).
+
+To address these challenges, we propose the plug-in and fine-tuning (PiFi) framework, which aims to harness the extensive knowledge and strengths of LLMs, such as linguistic ability and domain generalizability, while retaining the computational efficiency of SLMs. PiFi achieves this by extracting a single layer from a designated LLM and integrating it into a SLM, followed by fine-tuning the combined model on the target task. By
+
+incorporating a single LLM layer rather than the full model, PiFi enhances the SLM without compromising its lightweight structure. Furthermore, PiFi optimizes fine-tuning by freezing the extracted LLM layer, thereby minimizing the number of additional parameters to be trained.
+
+To validate the effectiveness of the PiFi framework, we conducted extensive experiments across various datasets, involving diverse natural language understanding (NLU) and natural language generation (NLG) tasks. Additionally, our evaluations in settings such as domain adaptation and multilingual classification demonstrate that PiFi can impart LLMs' benefits to SLMs, such as improved domain generalizability. In particular, our multilingual classification experiments show that PiFi can significantly enhance SLMs performance by leveraging a layer from an LLM pre-trained in the desired language, illustrating that even a straightforward integration of a single LLM layer can substantially assist SLM training through the knowledge transfer from the LLM. We also conducted a comprehensive comparison of different LLMs, evaluated the impact of varying model sizes, and assessed the effect of integrating instruction-tuned LLMs, highlighting the versatility and robustness of the PiFi framework.
+
+# 2 Related Work
+
+# 2.1 Small Language Models
+
+Language models (LMs) have progressed significantly from early models like BERT (Devlin et al., 2019) and GPT-1 (Radford and Narasimhan, 2018), both of which are based on the transformer architecture (Vaswani et al., 2017), to more advanced models such as RoBERTa (Liu et al., 2019), DeBERTa DeBERTa (He et al., 2021), and BART (Lewis et al., 2020). These models utilize larger datasets and advanced strategies, achieving greater performance. With each iteration, these models have enhanced their performance and applicability across diverse tasks and languages.
+
+Currently, open-sourced LMs are available in a range of parameter sizes, from extremely small models with 14.5 million parameters (Jiao et al., 2020) to large models with billions of parameters (Jiang et al., 2023; Almazrouei et al., 2023; Dubey et al., 2024; Yang et al., 2024; Team et al., 2024). In this study, we focus on specifically on SLMs with millions of parameters. Recent research has introduced several strategies to boost the performance
+
+and expand the utility of SLMs while maintaining computational efficiency (Gururangan et al., 2020; Gao et al., 2023b). Building upon these works, we propose the PiFi framework, which leverages the strengths of LLMs to supplement SLMs, effectively bridging the gap between SLMs and their larger counterparts.
+
+# 2.2 Employment of Large Language Models
+
+LLMs have achieved state-of-the-art performance across a broad range of NLP tasks, owing to their vast number of parameters. Models like GPT-4 (Achiam et al., 2023), Llama-3 (Dubey et al., 2024), and Mistral (Jiang et al., 2023) excel in generative tasks, particularly under zero-shot and few-shot settings. Additionally, LLMs demonstrate strong domain generalizability, making them applicable to a variety of domains without additional fine-tuning (Minaee et al., 2024).
+
+These strengths have inspired researchers to explore novel strategies for leveraging LLMs beyond simple downstream task applications. One prominent approach is knowledge distillation, where the knowledge of LLMs is transferred to smaller models. There are two main categories of knowledge distillation methods: parametric and non-parametric. Parametric methods involve using white-box LLMs as teacher models and training student models using the output distribution or intermediate features of the teacher model (Zhong et al., 2024; Timiryasov and Tastet, 2023; Gu et al., 2024). Non-parametric methods, on the other hand, generate synthetic training data using LLMs, which is then used to train smaller student models, thereby achieving knowledge distillation from a data-centric perspective (Ye et al., 2022; West et al., 2022; Gao et al., 2023a; Choi et al., 2024).
+
+In recent years, LLMs have also been integrated into multi-modal tasks, extending their utility beyond traditional NLP applications. For instance, some studies have used LLMs' linguistic capabilities to enhance the pre-training of vision-language models such as CLIP (Radford et al., 2021), by rewriting the textual descriptions in image-text pairs to improve the model's understanding of the relationships between images and text (Fan et al., 2023). Other research has explored using LLMs in computer vision, demonstrating that a combination of a vision transformer (ViT) model (Dosovitskiy et al., 2021) with an LLM layer can enhance performance on image classification and other vision
+
+![](images/edff1932c4f0daca95cad7ef5a36bbbcb02385c6e20d25a2d96501ce8e208e52.jpg)  
+Vanilla Fine-Tuning
+
+![](images/0e1d141fc7fd03a9b0d32cfa0f13e87519e800ce270680919ed42f613d5f92e2.jpg)  
+Plug-in and Fine-Tuning (Ours)  
+Figure 1: The comparison between vanilla fine-tuning of SLMs and our proposed PiFi architecture.
+
+tasks (Pang et al., 2024). The study showed that LLM layers can serve as visual encoders, helping to identify patterns among image tokens and highlight salient features.
+
+Inspired by these findings, we extend the use of LLMs to various NLP tasks. Specifically, we investigate whether key strengths of LLMs—such as generalizability to unseen domains and their rich linguistic knowledge—can be effectively distilled into SLMs, enabling them to benefit from these capabilities. Our work thus aims to bridge the performance gap between small and large models by transferring these strengths through a novel integration and fine-tuning framework.
+
+# 3 Method
+
+This section presents the PiFi framework, which integrates a single layer from an LLM into SLMs to leverage the extensive knowledge of LLMs while maintaining the efficiency of smaller models. We first describe the methodology for incorporating an LLM layer into encoder-based LMs, such as BERT, and encoder-decoder LMs like T5 (Raffel et al., 2020). Subsequently, we detail the fine-tuning process for SLMs with the integrated LLM layer. The overall procedure of the PiFi framework is illustrated in Figure 1.
+
+# 3.1 Plug-in of LLM Layer to Encoder-based LMs
+
+An encoder-based model, denoted as Enc, generates a hidden representation  $h_{enc}$  for an input se
+
+quence  $x$ , which is then passed to a classification head Head to produce the final prediction  $\hat{y}$ . This process is formulated as:
+
+$$
+h _ {e n c} = \operatorname {E n c} (x), \hat {y} = \operatorname {H e a d} \left(h _ {e n c}\right)
+$$
+
+To extend this process, PiFi introduces a single LLM layer, denoted as  $L_{LLM}$ , into the pipeline between Enc and Head. Since the hidden representation size of SLMs (e.g., 768 for BERT) may differ from that of LLMs (e.g., 4096 for Llama-3), we employ two additional transformation layers:  $L_{in}$  and  $L_{out}$ . Specifically,  $L_{in}$  projects  $h_{enc}$  into a compatible dimension for  $L_{LLM}$ , which then processes the projected representation. Subsequently,  $L_{out}$  converts the output of  $L_{LLM}$  back to the original hidden representation size of  $h_{enc}$ . The final transformed feature is then fed into Head to generate the prediction  $\hat{y}$ . The overall process can be expressed as:
+
+$$
+\begin{array}{l} h _ {e n c} = \operatorname {E n c} (x), h _ {L L M} = L _ {L L M} \left(L _ {i n} \left(h _ {e n c}\right)\right) \\ \hat {y} = \operatorname {H e a d} \left(L _ {\text {o u t}} \left(h _ {L L M}\right)\right) \\ \end{array}
+$$
+
+By introducing  $L_{LLM}$ , PiFi enables SLMs to benefit from the rich knowledge encoded within LLMs, thereby improving performance on various downstream tasks.
+
+# 3.2 Plug-in of LLM Layer to Encoder-decoder LMs
+
+Encoder-decoder models are widely used for sequence-to-sequence tasks such as machine translation and text summarization. An encoder-decoder model consists of an encoder, Enc, and a decoder, Dec. The encoder Enc processes the input sequence  $x$  to produce hidden representations  $h_{enc}$ , which are then passed to Dec to generate the target sequence  $\hat{y}$ . This process can be represented as:
+
+$$
+h _ {e n c} = E n c (x), \hat {y} _ {t} = D e c (h _ {e n c}, \hat {y} _ {<   t})
+$$
+
+where  $\hat{y}_{< t}$  denotes the tokens generated in previous time steps. The decoder Dec takes both hence and  $\hat{y}_{< t}$  as inputs to predict the next token  $\hat{y}_t$ .
+
+In the PiFi framework, an LLM layer,  $L_{LLM}$ , is inserted between Enc and Dec. Similar to the procedure for encoder-based models, the hidden representation  $h_{enc}$  is first transformed using  $L_{in}$  to match the input size of  $L_{LLM}$ . After processing by  $L_{LLM}$ , the output is projected back to the original size using  $L_{out}$  before being fed into the decoder Dec. This updated procedure for predicting  $\hat{y}_t$  can be formulated as:
+
+$$
+\begin{array}{l} h _ {e n c} = E n c (x), h _ {L L M} = L _ {L L M} \left(L _ {i n} \left(h _ {e n c}\right)\right) \\ \hat {y} _ {t} = D e c \big (L _ {o u t} (h _ {L L M}), \hat {y} _ {<   t} \big) \\ \end{array}
+$$
+
+By incorporating  $L_{LLM}$  in this manner, PiFi utilizes the LLM's knowledge to improve the performance of encoder-decoder models in generating high-quality target sequences.
+
+# 3.3 Fine-tuning of SLMs with Additional Layer
+
+During the fine-tuning stage, PiFi trains only the parameters of the original SLM,  $L_{in}$ ,  $L_{out}$ , and the classification head. Importantly, the parameters of  $L_{LLM}$  are kept frozen and remain unchanged. This approach offers two key advantages: (1) it minimizes the number of additional parameters to be trained, and (2) it preserves the knowledge encoded in  $L_{LLM}$  during its pre-training stage. If we were to also update the parameters of  $L_{LLM}$ , the model might suffer from catastrophic forgetting, a phenomenon where previously learned knowledge is lost when the model adapts to new data (Luo et al., 2023; Wang et al., 2023). By freezing the LLM layer, PiFi mitigates this risk and maintains the effectiveness of  $L_{LLM}$  throughout fine-tuning.
+
+We evaluate the impact of freezing  $L_{LLM}$  through an ablation study, as presented in Appendix B.4, demonstrating the effectiveness of this strategy in preventing catastrophic forgetting while ensuring optimal performance of the PiFi framework.
+
+# 4 Experiment
+
+In this section, we present a comprehensive evaluation of our proposed PiFi framework through various experiments.
+
+# 4.1 Experimental Setup
+
+We conduct experiments using Llama-3.1-8B (Meta, 2024) as the default LLM to extract  $L_{LLM}$ . Unless specified otherwise, the last layer of Llama-3.1-8B is integrated into a smaller LM as  $L_{LLM}$ .
+
+We performed our experiments on various tasks across NLU tasks and NLG tasks. For NLU tasks, we adopted text classification, natural language inference (NLI), and question answering (QA) tasks. Specifically, we used SST-2 (Socher et al., 2013), IMDB (Maas et al., 2011), Tweet for sentiment classification and offensive language identification (Rosenthal et al., 2017; Zampieri et al., 2019; Barbieri et al., 2020), and CoLA (Warstadt et al., 2019) datasets for text classification tasks. For NLI tasks, we used MNLI (Williams et al., 2018) and SNLI (Bowman et al., 2015). For these two tasks, we measured the performance of the model through accuracy and F1-score. We used SQuAD v1.1 (Rajpurkar et al., 2016) for QA tasks, where the performance was measured by exact match and F1-score.
+
+For NLG tasks, we evaluate each model through a machine translation task with Multi30k dataset (Elliott et al., 2016) and a text summarization task with CNN/DailyMail dataset (Nallapati et al., 2016). For both tasks, we used BLEU (Papineni et al., 2002), ROUGE (Lin, 2004), METEOR (Banerjee and Lavie, 2005), BERTScore (Zhang et al., 2020), and BARTScore (Yuan et al., 2021) as a metric for measuring the performance of each model. All models were trained with five different random seeds, and we report the average performance for each experimental setup.
+
+# 4.2 Performance Improvement of PiFi on NLU Tasks
+
+We assess PiFi's effectiveness on NLU tasks using several SLMs: BERT (Devlin et al., 2019), RoBERTa (Liu et al., 2019), ELECTRA (Clark
+
+<table><tr><td rowspan="2"></td><td colspan="5">Classification</td><td colspan="2">NLI</td><td>QA</td><td rowspan="2">Average</td></tr><tr><td>SST2</td><td>IMDB</td><td>Tweet(Sentiment)</td><td>Tweet(Offensive)</td><td>CoLA</td><td>MNLI</td><td>SNLI</td><td>SQuAD</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.1</td><td>86.9</td><td>83.15</td><td>80.10</td><td>82.00</td><td>89.10</td><td>63.81</td><td>82.45</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.862</td><td>0.7727</td><td>0.7398</td><td>0.8131</td><td>0.8892</td><td>0.7606</td><td>0.7752</td></tr><tr><td rowspan="2">+PiFi(Llama-3.1-8B)</td><td>91.5</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td><td>82.74</td><td>89.48</td><td>66.17</td><td>84.75</td></tr><tr><td>0.9125</td><td>0.4800</td><td>0.9224</td><td>0.8026</td><td>0.7523</td><td>0.8185</td><td>0.8934</td><td>0.7809</td><td>0.7953</td></tr><tr><td rowspan="2">RoBERTabase</td><td>91.65</td><td>87.36</td><td>90.12</td><td>83.60</td><td>80.55</td><td>84.00</td><td>88.73</td><td>68.09</td><td>84.32</td></tr><tr><td>0.9137</td><td>0.4890</td><td>0.894</td><td>0.7682</td><td>0.7192</td><td>0.8332</td><td>0.8856</td><td>0.8006</td><td>0.7879</td></tr><tr><td rowspan="2">+PiFi(Llama-3.1-8B)</td><td>92.54</td><td>88.68</td><td>91.85</td><td>85.35</td><td>82.29</td><td>84.25</td><td>88.49</td><td>68.97</td><td>85.42</td></tr><tr><td>0.9228</td><td>0.4878</td><td>0.9115</td><td>0.8083</td><td>0.7526</td><td>0.8346</td><td>0.8837</td><td>0.8089</td><td>0.7980</td></tr><tr><td rowspan="2">ELECTRAbase</td><td>93.42</td><td>88.31</td><td>90.58</td><td>83.52</td><td>83.99</td><td>85.41</td><td>90.11</td><td>44.44</td><td>82.00</td></tr><tr><td>0.9324</td><td>0.4974</td><td>0.8989</td><td>0.7751</td><td>0.7757</td><td>0.8472</td><td>0.8995</td><td>0.5706</td><td>0.7662</td></tr><tr><td rowspan="2">+PiFi(Llama-3.1-8B)</td><td>94.13</td><td>89.40</td><td>0.9331</td><td>84.99</td><td>86.26</td><td>86.47</td><td>90.48</td><td>67.99</td><td>86.71</td></tr><tr><td>0.9393</td><td>0.4994</td><td>0.9270</td><td>0.7848</td><td>0.8081</td><td>0.8618</td><td>0.9037</td><td>0.8045</td><td>0.8076</td></tr><tr><td rowspan="2">DeBERTabase</td><td>92.60</td><td>87.98</td><td>88.22</td><td>82.67</td><td>80.04</td><td>83.72</td><td>89.61</td><td>67.87</td><td>84.40</td></tr><tr><td>0.9236</td><td>0.4882</td><td>0.8755</td><td>0.7729</td><td>0.7216</td><td>0.8312</td><td>0.8945</td><td>0.8094</td><td>0.7943</td></tr><tr><td rowspan="2">+PiFi(Llama-3.1-8B)</td><td>93.04</td><td>88.85</td><td>92.17</td><td>85.47</td><td>80.87</td><td>84.87</td><td>90.62</td><td>69.65</td><td>85.63</td></tr><tr><td>0.9283</td><td>0.4928</td><td>0.9152</td><td>0.8065</td><td>0.7347</td><td>0.8419</td><td>0.8982</td><td>0.8152</td><td>0.8037</td></tr><tr><td rowspan="2">DeBERTa-V3base</td><td>93.74</td><td>89.45</td><td>91.29</td><td>83.60</td><td>84.75</td><td>87.52</td><td>90.94</td><td>69.40</td><td>86.34</td></tr><tr><td>0.9355</td><td>0.4951</td><td>0.9066</td><td>0.7915</td><td>0.8015</td><td>0.8687</td><td>0.9079</td><td>0.8249</td><td>0.8162</td></tr><tr><td rowspan="2">+PiFi(Llama-3.1-8B)</td><td>95.01</td><td>89.83</td><td>93.80</td><td>85.60</td><td>86.07</td><td>87.98</td><td>91.05</td><td>69.87</td><td>87.40</td></tr><tr><td>0.9481</td><td>0.5014</td><td>0.9325</td><td>0.8056</td><td>0.8118</td><td>0.8932</td><td>0.9095</td><td>0.8283</td><td>0.8287</td></tr></table>
+
+Table 1: Experimental results of PiFi on various NLU tasks and datasets. We used the last layer from Llama-3.1-8B as  $L_{LLM}$  in this experiment. For classification and NLI tasks, we report accuracy and F1-score in the upper row and lower row of each cell. For the QA task, we report the exact match and F1-score in the upper row and lower row of each cell.
+
+et al., 2020), DeBERTa (He et al., 2021), and DeBERTaV3 (He et al., 2023).
+
+Table 1 shows the results, where PiFi consistently outperforms vanilla fine-tuned SLMs across all datasets. For example, integrating PiFi into BERT resulted in a  $2.3\%$  increase in average accuracy compared to standard fine-tuning. Similar gains are observed for other models, demonstrating PiFi's compatibility and effectiveness across a diverse set of NLU tasks and architectures.
+
+# 4.3 Performance Improvement of PiFi on NLG Tasks
+
+To validate PiFi on NLG tasks, we employed encoder-decoder models such as T5<sub>base</sub> (Raffel et al., 2020) and BART<sub>base</sub> (Lewis et al., 2020). These models were evaluated on machine translation and text summarization tasks, thereby validating the advantage of PiFi on NLG tasks beyond NLU tasks.
+
+The experimental results on two NLG tasks are displayed in Table 2. We found that the SLM trained with PiFi exhibits higher performance compared to vanilla fine-tuned SLM in most cases. Particularly, for the text summarization task, both mod
+
+els trained with PiFi achieved higher scores across all evaluation metrics, indicating that PiFi effectively transfers knowledge from the LLM to SLMs, thus enhancing linguistic capabilities in NLG tasks.
+
+# 4.4 Generalizability of PiFi under Domain Shift
+
+Since LLMs are pre-trained on much larger and more diverse corpora compared to SLMs, integrating LLMs into SLMs through PiFi is expected to improve the generalizability of the model, particularly in unseen domains. To test this hypothesis, we conducted an experiment comparing the performance of the PiFi model against a vanilla finetuned model under conditions of domain shift. For this experiment, we used three text classification datasets. Specifically, we used IMDB, CR (Ding et al., 2008), and Tweet for sentiment classification. While all three datasets involve sentiment classification, they represent different domains such as movie review, electronics product review, and tweet messages. We trained the models on each dataset and evaluated their performance across all three datasets.
+
+The results, shown in Table 3, indicate that PiFi
+
+<table><tr><td rowspan="2"></td><td colspan="5">Multi30K (Translation)</td><td colspan="5">CNN/Daily Mail (Summarization)</td></tr><tr><td>BLEU</td><td>ROUGE</td><td>METEOR</td><td>BERTS.</td><td>BARTS.</td><td>BLEU</td><td>ROUGE</td><td>METEOR</td><td>BERTS.</td><td>BARTS.</td></tr><tr><td>T5base</td><td>0.5301</td><td>0.6195</td><td>0.3605</td><td>0.8724</td><td>-4.634</td><td>0.2175</td><td>0.2323</td><td>0.1731</td><td>0.7409</td><td>-5.784</td></tr><tr><td>+ PiFi (Llama-3.1-8B)</td><td>0.5413</td><td>0.6536</td><td>0.3534</td><td>0.8978</td><td>-4.669</td><td>0.2242</td><td>0.2357</td><td>0.1752</td><td>0.7412</td><td>-5.777</td></tr><tr><td>BARTbase</td><td>0.4580</td><td>0.5864</td><td>0.3331</td><td>0.8635</td><td>-4.513</td><td>0.2270</td><td>0.2348</td><td>0.1782</td><td>0.7424</td><td>-5.665</td></tr><tr><td>+ PiFi (Llama-3.1-8B)</td><td>0.4695</td><td>0.5908</td><td>0.3364</td><td>0.8617</td><td>-4.515</td><td>0.2331</td><td>0.2355</td><td>0.1799</td><td>0.7425</td><td>-5.652</td></tr></table>
+
+Table 2: Experimental results of PiFi on machine translation and text summarization tasks. We used the last layer from Llama-3.1-8B as  $L_{LLM}$  in this experiment.  
+
+<table><tr><td rowspan="2">Train Dataset</td><td rowspan="2"></td><td colspan="3">Test Dataset</td></tr><tr><td>IMDB</td><td>Tweet (Sentiment)</td><td>CR</td></tr><tr><td rowspan="6">IMDB</td><td rowspan="2">BERTbase</td><td>85.1</td><td>70.40</td><td>74.56</td></tr><tr><td>0.4773</td><td>0.6918</td><td>0.7301</td></tr><tr><td rowspan="2">+PiFi (Llam-3-8B)</td><td>87.09</td><td>83.68</td><td>79.86</td></tr><tr><td>0.4800</td><td>0.8176</td><td>0.788</td></tr><tr><td rowspan="2">BERTlarge</td><td>86.88</td><td>77.39</td><td>76.10</td></tr><tr><td>0.4836</td><td>0.7647</td><td>0.7419</td></tr><tr><td rowspan="6">Tweet (Sentiment)</td><td rowspan="2">BERTbase</td><td>74.70</td><td>86.90</td><td>85.46</td></tr><tr><td>0.4306</td><td>0.862</td><td>0.8357</td></tr><tr><td rowspan="2">+PiFi (Llam-3-8B)</td><td>77.28</td><td>92.95</td><td>87.25</td></tr><tr><td>0.4351</td><td>0.9224</td><td>0.8596</td></tr><tr><td rowspan="2">BERTlarge</td><td>75.91</td><td>89.26</td><td>86.50</td></tr><tr><td>0.4331</td><td>0.8853</td><td>0.8526</td></tr><tr><td rowspan="6">CR</td><td rowspan="2">BERTbase</td><td>75.72</td><td>82.52</td><td>89.60</td></tr><tr><td>0.4301</td><td>0.8161</td><td>0.8857</td></tr><tr><td rowspan="2">+PiFi (Llam-3-8B)</td><td>77.49</td><td>84.80</td><td>90.90</td></tr><tr><td>0.4362</td><td>0.8365</td><td>0.9015</td></tr><tr><td rowspan="2">BERTlarge</td><td>76.77</td><td>83.90</td><td>90.64</td></tr><tr><td>0.4341</td><td>0.8284</td><td>0.8989</td></tr></table>
+
+consistently outperforms vanilla fine-tuning across all domain shifts. Notably, the PiFi model trained on IMDB showed significant performance improvements when tested on the Tweet and CR datasets, with gains of  $13.28\% \mathrm{p}$  and  $5.3\% \mathrm{p}$ , respectively, over the vanilla fine-tuned model. To verify that this improved generalization is due to leveraging LLM layer knowledge rather than merely a simple increase in parameters, we compared PiFi's fine-tuning performance against  $\mathrm{BERT}_{large}$ , which is slightly larger than the PiFi model. The results showed that PiFi outperformed  $\mathrm{BERT}_{large}$ , confirming that incorporating the LLM layer into an SLM via PiFi allows the model to effectively leverage the extensive knowledge stored in the LLM. This enhancement offers benefits beyond parameter scal
+
+Table 3: Experimental results of PiFi under domain shift. We report accuracy and F1-score in the upper row and lower row of each cell.  
+
+<table><tr><td></td><td>NSMC (Korean)</td><td>Filmstarts (German)</td></tr><tr><td rowspan="2">mBERTbase</td><td>83.62</td><td>86.77</td></tr><tr><td>0.8318</td><td>0.8279</td></tr><tr><td>+ PiFi</td><td>84.04</td><td>87.92</td></tr><tr><td>(Llama-3-8B English)</td><td>0.8362</td><td>0.8362</td></tr><tr><td>+ PiFi</td><td>85.61</td><td>87.09</td></tr><tr><td>(Llama-3-8B Korean)</td><td>0.8522</td><td>0.8337</td></tr><tr><td>+ PiFi</td><td>83.85</td><td>88.11</td></tr><tr><td>(Llama-3-8B German)</td><td>0.8341</td><td>0.8411</td></tr></table>
+
+Table 4: Experimental results on Korean and German text classification datasets. For this experiment, we used Llama-3 model trained in English, Korean, and German. We report accuracy and F1-score in the upper row and lower row of each cell.
+
+ing, thereby improving its ability to generalize to unseen domains.
+
+# 4.5 Transferring of Linguistic Ability of LLMs to SLMs
+
+In this section, we examine how the primary language of large language models (LLMs) influences the performance of the PiFi model when applied to smaller language models (SLMs). Specifically, we explore this effect by training an mBERT model (Pires et al., 2019) on two distinct datasets: NSMC (Park, 2016), a Korean sentiment classification dataset for movie reviews, and Filmstarts (Guhr et al., 2020), a German movie review sentiment dataset. For our experiments, we use Llama-3-8B $^{1}$  along with Llama-3-8B variants further fine-tuned for Korean and German (Lee, 2024; DiscoResearch, 2024) to extract  $L_{LLM}$ , which is then integrated into the mBERT model.
+
+The results of our experiments are presented in
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet(Sentiment)</td><td>Tweet(Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.10</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.862</td><td>0.7727</td><td>0.7398</td></tr><tr><td rowspan="2">+ PiFi-Random(Llama-3.1-8B)</td><td>90.28</td><td>86.43</td><td>89.49</td><td>83.50</td><td>80.01</td></tr><tr><td>0.8997</td><td>0.4879</td><td>0.9095</td><td>0.7564</td><td>0.7046</td></tr><tr><td rowspan="2">+ PiFi-Random-Full(Llama-3.1-8B)</td><td>90.50</td><td>86.46</td><td>90.95</td><td>83.95</td><td>80.39</td></tr><tr><td>0.9017</td><td>0.4767</td><td>0.9024</td><td>0.7812</td><td>0.7376</td></tr><tr><td rowspan="2">+ PiFi(Llama-3.1-8B)</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>0.9125</td><td>0.48</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr></table>
+
+Table 5: Experimental results of PiFi under different configurations of  $L_{LLM}$  initialization and fine-tuning. For this experiment, we compared PiFi-Random (frozen randomly initialized  $L_{LLM}$ ), PiFi-Random-Full (fine-tuned randomly initialized  $L_{LLM}$ ), and PiFi (pre-trained  $L_{LLM}$ ). We report accuracy and F1-score in the upper row and lower row of each cell.
+
+Table 4. We find that while PiFi demonstrates performance improvements on non-English datasets using the default English Llama-3 model, utilizing language-specific Llama-3 variants (e.g., Korean or German) leads to further performance gains for their respective languages. For example, the PiFi model, when combined with the Korean Llama-3 model, achieved an additional  $1.57\%$ p improvement over the English Llama-3 model on the Korean dataset. Conversely, when a PiFi model is trained on a different language than the target downstream task, the performance gains diminish. For instance, training PiFi on the Filmstarts (German) dataset with the Korean Llama-3 model resulted in a performance increase that was  $0.83\%$ p lower than that obtained with the English Llama-3 model and  $1.02\%$ p lower than that with the German Llama-3 model.
+
+These findings indicate that the PiFi model benefits primarily from effectively leveraging knowledge from the LLM through a single layer, rather than simply from an increase in the number of model parameters. In conclusion, aligning the language of the downstream task with the language of the LLM used for  $L_{LLM}$  extraction is critical for maximizing the effectiveness of PiFi, as this alignment ensures optimal utilization of the LLM's linguistic capabilities.
+
+# 4.6 Probing the Role of Intrinsic Knowledge in LLM Layer
+
+To verify that the performance improvement achieved by plugging in  $L_{LLM}$  is not simply due to an increase in the number of parameters but rather a result of leveraging the inherent knowledge embedded in the LLM, we designed a series
+
+<table><tr><td></td><td>Params (M)</td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td><td>Tweet (Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td rowspan="2">110</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.10</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.8620</td><td>0.7727</td><td>0.7398</td></tr><tr><td rowspan="2">+PiFi (Llama-3.1-8B)</td><td rowspan="2">334</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>0.9125</td><td>0.48</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr><tr><td rowspan="2">BERTlarge</td><td rowspan="2">336</td><td>90.73</td><td>86.88</td><td>89.26</td><td>83.95</td><td>80.96</td></tr><tr><td>0.9040</td><td>0.4836</td><td>0.8853</td><td>0.7887</td><td>0.7327</td></tr></table>
+
+Table 6: Experimental results comparing the performance of PiFi and  $\mathrm{BERT}_{\text{large}}$  with similar parameter counts. We report accuracy and F1-score in the upper row and lower row of each cell.
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet(Sentiment)</td><td>Tweet(Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.10</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.862</td><td>0.7727</td><td>0.7398</td></tr><tr><td rowspan="2">+ PiFi(Llama-3.1-8B)</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>0.9125</td><td>0.48</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr><tr><td rowspan="2">+ PiFi(Mistral-7B-v0.1)</td><td>90.89</td><td>86.47</td><td>90.12</td><td>84.99</td><td>81.65</td></tr><tr><td>0.9061</td><td>0.4794</td><td>0.8943</td><td>0.8055</td><td>0.7324</td></tr><tr><td rowspan="2">+ PiFi(Mistral-7B-v0.3)</td><td>91.65</td><td>87.22</td><td>92.57</td><td>85.33</td><td>81.72</td></tr><tr><td>0.9136</td><td>0.4797</td><td>0.9177</td><td>0.802</td><td>0.7502</td></tr><tr><td rowspan="2">+ PiFi(Qwen2 -7B)</td><td>91.17</td><td>86.68</td><td>92.48</td><td>85.70</td><td>81.62</td></tr><tr><td>0.9092</td><td>0.4774</td><td>0.917</td><td>0.809</td><td>0.7475</td></tr><tr><td rowspan="2">+ PiFi(Gemma-2-9B)</td><td>91.39</td><td>87.1</td><td>92.34</td><td>84.29</td><td>80.74</td></tr><tr><td>0.9111</td><td>0.4849</td><td>0.9163</td><td>0.7866</td><td>0.7598</td></tr><tr><td rowspan="2">+ PiFi(Falcon-7B)</td><td>91.44</td><td>86.63</td><td>92.51</td><td>84.85</td><td>80.99</td></tr><tr><td>0.9115</td><td>0.4779</td><td>0.9178</td><td>0.7907</td><td>0.7518</td></tr></table>
+
+Table 7: Experimental results of PiFi across various LLMs, such as Llama-3.1-8B, Mistral-7B-v0.1 and v0.3, Qwen-7B, Gemma-2-9B, and Falcon-7B. We extracted the last layer from each LLM as  $L_{LLM}$  and compared their performance. We report accuracy and F1-score in the upper row and lower row of each cell.
+
+of experiments.
+
+First, Table 5 shows the results of experiments conducted with randomly initialized  $L_{LLM}$  under two configurations. PiFi-Random refers to a model where the randomly initialized  $L_{LLM}$  is frozen during fine-tuning, while PiFi-Random-Full is a model where the entire network, including the randomly initialized  $L_{LLM}$ , is fine-tuned. Although PiFi-Random and PiFi-Random-Full exhibited slight improvements over  $\mathrm{BERT}_{\text{base}}$ , their performance fell short of PiFi, which utilizes the pre-trained weights of the  $L_{LLM}$ . These results demonstrate that the benefits of simply increasing the number of parameters are limited.
+
+In addition to the comparison with randomly initialized baselines, Table 6 compares the performance of PiFi with  $\mathrm{BERT}_{\text {large }}$ , which has a similar number of parameters.  $\mathrm{BERT}_{\text {large }}$  has approximately 336M parameters, while PiFi applied to  $\mathrm{BERT}_{\text {base }}$  has a comparable parameter count of approximately 334M. Despite this similarity in
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td><td>Tweet (Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.1</td><td>86.90</td><td>83.15</td><td>80.10</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.862</td><td>0.7727</td><td>0.7398</td></tr><tr><td>+ PiFi</td><td>90.13</td><td>86.09</td><td>89.67</td><td>84.52</td><td>81.15</td></tr><tr><td>(Qwen2-0.5B)</td><td>0.8977</td><td>0.4843</td><td>0.8899</td><td>0.7959</td><td>0.7453</td></tr><tr><td>+ PiFi</td><td>91.06</td><td>86.29</td><td>92.35</td><td>85.91</td><td>81.54</td></tr><tr><td>(Qwen2-1.5B)</td><td>0.9079</td><td>0.4668</td><td>0.9164</td><td>0.8096</td><td>0.7453</td></tr><tr><td>+ PiFi</td><td>91.17</td><td>86.68</td><td>92.48</td><td>86.04</td><td>81.62</td></tr><tr><td>(Qwen2-7B)</td><td>0.9092</td><td>0.4774</td><td>0.917</td><td>0.8068</td><td>0.7475</td></tr></table>
+
+Table 8: Experimental results of PiFi on different sizes of LM within same model family. For this experiment, we extracted the last layer of 0.5B, 1.5B, 7B version of Qwen2 as  $L_{\mathrm{LM}}$ . We report accuracy and F1-score in the upper row and lower row of each cell.
+
+parameters and native architecture of BERTlarge, BERTbase with PiFi outperformed BERTlarge in terms of performance.
+
+This experimental finding suggests that the performance improvement of PiFi is not merely an outcome of increasing the number of parameters but is primarily due to effectively utilizing the pretrained knowledge embedded in the LLM.
+
+# 4.7 Comparison of PiFi between Various LLMs
+
+LLMs vary significantly in architecture and pretraining data, which can influence their performance as backbones for the PiFi framework. In this section, we evaluate the downstream task performance of PiFi when using different LLMs as sources for extracting  $L_{LLM}$ . Specifically, we incorporate the following LLMs: Llama-3.1-8B (Dubey et al., 2024), Mistral-7B-v0.1 and v0.3 (Jiang et al., 2023), Qwen2-7B (Yang et al., 2024), Gemma-2-9B (Team et al., 2024), and Falcon-7B (Almazrouei et al., 2023). These LLMs serve as backbones for generating  $L_{LLM}$ , which is subsequently integrated into a BERT model for the fine-tuning on text classification tasks.
+
+Table 7 presents the results of these experiments. Our findings reveal several key insights. First, across all LLM variants, PiFi consistently outperforms the BERT model with vanilla fine-tuning, demonstrating the effectiveness of utilizing  $L_{LLM}$ . Second, Llama-3.1-8B and Mistral-7B-v0.3 yield the highest performance across multiple datasets, indicating their strong suitability as backbones for PiFi. Additionally, the comparison between PiFi models using Mistral-7B-v0.1 and v0.3 shows that even incremental improvements within the same model family lead to enhanced downstream per
+
+formance. This result suggests that PiFi is highly responsive to the advancements in LLM capabilities, and we expect that future advancements in LLM development will further augment the effectiveness of PiFi.
+
+# 4.8 Impact on PiFi Depending on LM Size
+
+Currently, LMs are available in various sizes to cater to different computational and performance needs. In this section, we evaluate how the size of the LM used to extract LM layer affects the overall performance of PiFi. For this experiment, we utilize three versions of the Qwen2 model (Yang et al., 2024): Qwen2-0.5B, Qwen2-1.5B, and Qwen2-7B.
+
+Table 8 demonstrates the results. We observe that integrating LM layer from any LM size into a SLM results in improved performance compared to traditional fine-tuning methods. Notably, the usage of 7B model yields the highest performance gains, surpassing both the 0.5B and 1.5B models. This suggests that larger models, with their enhanced capacity for capturing and storing extensive knowledge, provide more informative and effective representations for PiFi, leading to better downstream task performance.
+
+# 5 Conclusion
+
+In this paper, we introduced PiFi, plug-in and fine-tuning, a novel framework designed to leverage the intrinsic knowledge of LLMs while maintaining the efficiency and lightweight nature of SLMs. PiFi incorporates a frozen LLM layer into a SLM and fine-tunes the resulting model, enabling the effective utilization of LLM knowledge without significantly increasing model complexity.
+
+We validated the applicability of PiFi across a diverse range of NLU and NLG tasks, demonstrating its compatibility with various SLMs and LLMs. Notably, the results from our domain shift experiment in Section 4.4 showed that PiFi can significantly improve the performance of SLMs on unseen domains. Similarly, the analysis in Section 4.5 confirmed that PiFi effectively leverages the linguistic capabilities of LLMs by incorporating even a single LLM layer.
+
+In future work, we aim to extend the usability of PiFi to more tasks and languages. We will also explore advanced strategies to further optimize PiFi's effectiveness, such as automatically selecting the number and position of LLM layers to be integrated, allowing for more flexible and task-specific
+
+knowledge transfer.
+
+# Limitations
+
+In this study, we proposed PiFi, an efficient framework for integrating the knowledge of LLMs into SLMs. Despite its effectiveness, there are several limitations in our current approach that warrant further exploration.
+
+One limitation is that we selected  $L_{LLM}$  as the last layer of each LLM based on a heuristic approach. While our analysis in Appendix B.1 indicates that using the last layer generally yields the best performance, it is possible that different layers may be optimal depending on the specific downstream task. Future research could explore methods for automatically selecting the most suitable LLM layer as  $L_{LLM}$ , akin to techniques used in neural architecture search (Elsken et al., 2019; White et al., 2023).
+
+Another limitation is that our experiments primarily focused on relatively straightforward tasks, as this manuscript's primary goal is to propose PiFi and clearly demonstrate its effectiveness. We did not extend our evaluation to more complex benchmarks, such as the massive multitask language understanding dataset (Hendrycks et al., 2021). Testing PiFi on such diverse and challenging tasks in future work could provide deeper insights into its generalizability and highlight areas for further improvement.
+
+# Ethics Statement
+
+We acknowledge the potential for inherent biases in LLMs, and the integration of an LLM layer into a SLM may introduce such biases (Liu et al., 2022). While our experiments did not reveal any explicit evidence of bias, future work will carefully consider the possibility of bias transfer from LLMs to SLMs when employing PiFi, as well as its broader implications.
+
+# Acknowledgements
+
+This work was supported by the Institute of Information Communications Technology Planning Evaluation (IITP) grant funded by the Korea government (MSIT) [RS-2021-II211341, Artificial Intelligent Graduate School Program (Chung-Ang University)] and by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2025-00556246).
+
+# 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 arXiv:2303.08774.  
+Loubna Ben Allal, Anton Lozhkov, Elie Bakouch, Gabriel Martin Blázquez, Guilherme Penedo, Lewis Tunstall, Andrés Marafioti, Hynek Kydlíček, Agustín Piqueres Lajarín, Vaibhav Srivastav, et al. 2025. Smollm2: When smol goes big-data-centric training of a small language model. arXiv preprint arXiv:2502.02737.  
+Ebtesam Almazrouei, Hamza Alobeidli, Abdulaziz Alshamsi, Alessandro Cappelli, Ruxandra Cojocaru, Mérouane Debbah, Étienne Goffinet, Daniel Hesslow, Julien Launay, Quentin Malartic, et al. 2023. The falcon series of open language models. arXiv preprint arXiv:2311.16867.  
+Amit Artzy and Roy Schwartz. 2024. Attend first, consolidate later: On the importance of attention in different llm layers. In Proceedings of EMNLP 2024 BlackboxNLP Workshop, pages 177-184.  
+Satanjeev Banerjee and Alon Lavie. 2005. METEOR: An automatic metric for MT evaluation with improved correlation with human judgments. In Proceedings of ACL 2005 Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and/or Summarization, pages 65-72.  
+Francesco Barbieri, Jose Camacho-Collados, Luis Espinosa Anke, and Leonardo Neves. 2020. TweetEval: Unified benchmark and comparative evaluation for tweet classification. In Findings of EMNLP, pages 1644-1650.  
+Samuel R. Bowman, Gabor Angeli, Christopher Potts, and Christopher D. Manning. 2015. A large annotated corpus for learning natural language inference. In Proceedings of EMNLP, pages 632-642.  
+Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. In Proceedings of NeurIPS, pages 1877-1901.  
+Juhwan Choi, Yeonghwa Kim, Seunguk Yu, Jungmin Yun, and Youngbin Kim. 2024. Unigen: Universal domain generalization for sentiment classification via zero-shot dataset generation. In Proceedings of EMNLP, pages 1-14.  
+Kevin Clark, Minh-Thang Luong, Quoc V. Le, and Christopher D. Manning. 2020. Electra: Pre-training text encoders as discriminators rather than generators. In Proceedings of ICLR.  
+Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. 2023. Qlora: Efficient finetuning
+
+of quantized llms. In Proceedings of NeurIPS, pages 10088-10115.  
+Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of NAACL, pages 4171-4186.  
+Xiaowen Ding, Bing Liu, and Philip S Yu. 2008. A holistic lexicon-based approach to opinion mining. In Proceedings of WSDM, pages 231-240.  
+DiscoResearch. 2024. Llama-3-german-8b. Hugging Face Repository.  
+Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. 2021. An image is worth 16x16 words: Transformers for image recognition at scale. In Proceedings of ICLR.  
+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.  
+Desmond Elliott, Stella Frank, Khalil Sima'an, and Lucia Specia. 2016. Multi30k: Multilingual english-german image descriptions. In Proceedings of ACL 2016 Workshop on Vision and Language, pages 70-74.  
+Thomas Elsken, Jan Hendrik Metzen, and Frank Hutter. 2019. Neural architecture search: A survey. Journal of Machine Learning Research, 20(55):1-21.  
+Lijie Fan, Dilip Krishnan, Phillip Isola, Dina Katabi, and Yonglong Tian. 2023. Improving clip training with language rewrites. In Proceedings of NeurIPS, pages 35544-35575.  
+Jiahui Gao, Renjie Pi, Lin Yong, Hang Xu, Jiacheng Ye, Zhiyong Wu, Weizhong Zhang, Xiaodan Liang, Zhenguo Li, and Lingpeng Kong. 2023a. Self-guided noise-free data generation for efficient zero-shot learning. In Proceedings of ICLR.  
+Ze-Feng Gao, Kun Zhou, Peiyu Liu, Wayne Xin Zhao, and Ji-Rong Wen. 2023b. Small pre-trained language models can be fine-tuned as large models via over-parameterization. In Proceedings of ACL, pages 3819-3834.  
+Yuxian Gu, Li Dong, Furu Wei, and Minlie Huang. 2024. Minillm: Knowledge distillation of large language models. In Proceedings of ICLR.  
+Oliver Guhr, Anne-Kathrin Schumann, Frank Bahrmann, and Hans Joachim Böhme. 2020. Training a broad-coverage German sentiment classification model for dialog systems. In Proceedings of LREC, pages 1627-1632.
+
+Suchin Gururangan, Ana Marasovic, Swabha Swayamdipta, Kyle Lo, Iz Beltagy, Doug Downey, and Noah A. Smith. 2020. Don't stop pretraining: Adapt language models to domains and tasks. In Proceedings of ACL, pages 8342-8360.  
+Pengcheng He, Jianfeng Gao, and Weizhu Chen. 2023. Debertav3: Improving deberta using electra-style pretraining with gradient-disentangled embedding sharing. In Proceedings of ICLR.  
+Pengcheng He, Xiaodong Liu, Jianfeng Gao, and Weizhu Chen. 2021. Deberta: Decoding-enhanced bert with disentangled attention. In Proceedings of ICLR.  
+Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021. Measuring massive multitask language understanding. In Proceedings of ICLR.  
+Edward J Hu, yelong shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. LoRA: Low-rank adaptation of large language models. In Proceedings of ICLR.  
+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.  
+Xiaoqi Jiao, Yichun Yin, Lifeng Shang, Xin Jiang, Xiao Chen, Linlin Li, Fang Wang, and Qun Liu. 2020. TinyBERT: Distilling BERT for natural language understanding. In Findings of EMNLP, pages 4163-4174.  
+Tianjie Ju, Weiwei Sun, Wei Du, Xinwei Yuan, Zhaochun Ren, and Gongshen Liu. 2024. How large language models encode context knowledge? a layer-wise probing study. In Proceedings of LREC-COLING, pages 8235-8246.  
+Diederik P. Kingma and Jimmy Ba. 2015. Adam: A method for stochastic optimization. In Proceedings of ICLR.  
+Junbum Lee. 2024. Llama-3-open-ko. Hugging Face Repository.  
+Pierre Lepagnol, Thomas Gerald, Sahar Ghannay, Christophe Servan, and Sophie Rosset. 2024. Small language models are good too: An empirical study of zero-shot classification. In Proceedings of LREC-COLING, pages 14923-14936.  
+Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In Proceedings of ACL, pages 7871-7880.
+
+Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Proceedings of ACL 2004 Workshop Text Summarization Branches Out, pages 74-81.  
+Zheng Lin, Guanqiao Qu, Qiyuan Chen, Xianhao Chen, Zhe Chen, and Kaibin Huang. 2023. Pushing large language models to the 6g edge: Vision, challenges, and opportunities. CoRR.  
+Ruibo Liu, Chenyan Jia, Jason Wei, Guangxuan Xu, and Soroush Vosoughi. 2022. Quantifying and alleviating political bias in language models. Artificial Intelligence, 304:103654.  
+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.  
+Zhu Liu, Cunliang Kong, Ying Liu, and Maosong Sun. 2024. Fantastic semantics and where to find them: Investigating which layers of generative LLMs reflect lexical semantics. In Findings of ACL, pages 14551-14558.  
+Zhenyan Lu, Xiang Li, Dongqi Cai, Rongjie Yi, Fangming Liu, Xiwen Zhang, Nicholas D Lane, and Mengwei Xu. 2024. Small language models: Survey, measurements, and insights. arXiv preprint arXiv:2409.15790.  
+Yun Luo, Zhen Yang, Fandong Meng, Yafu Li, Jie Zhou, and Yue Zhang. 2023. An empirical study of catastrophic forgetting in large language models during continual fine-tuning. arXiv preprint arXiv:2308.08747.  
+Andrew L. Maas, Raymond E. Daly, Peter T. Pham, Dan Huang, Andrew Y. Ng, and Christopher Potts. 2011. Learning word vectors for sentiment analysis. In Proceedings of ACL, pages 142-150.  
+Meta. 2019. fvcore. GitHub Repository.  
+Meta. 2024. Llama-3.1-8b. Hugging Face Repository.  
+Shervin Minaee, Tomas Mikolov, Narjes Nikzad, Meysam Chenaghlu, Richard Socher, Xavier Amatriain, and Jianfeng Gao. 2024. Large language models: A survey. arXiv preprint arXiv:2402.06196.  
+Ramesh Nallapati, Bowen Zhou, Cicero dos Santos, Caglar Gulçehre, and Bing Xiang. 2016. Abstractive text summarization using sequence-to-sequence RNNs and beyond. In Proceedings of SIGNLL, pages 280-290.  
+Made Nindyatama Nityasya, Haryo Akbario T. Wibowo, Radityo Eko Prasojo, and Alham Fikri Aji. 2020. Costs to consider in adopting nlp for your business. arXiv preprint arXiv:2012.08958.
+
+Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to follow instructions with human feedback. In Proceedings of NeurIPS, pages 27730-27744.  
+Ziqi Pang, Ziyang Xie, Yunze Man, and Yu-Xiong Wang. 2024. Frozen transformers in language models are effective visual encoder layers. In Proceedings of ICLR.  
+Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: A method for automatic evaluation of machine translation. In Proceedings of ACL, pages 311-318.  
+Lucy Park. 2016. Naver sentiment movie corpus. GitHub Repository.  
+Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, Junjie Bai, and Soumith Chintala. 2019a. Pytorch: An imperative style, high-performance deep learning library. In Proceedings of NeurIPS.  
+Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. 2019b. Pytorch: An imperative style, high-performance deep learning library. In Proceedings of NeurIPS.  
+Telmo Pires, Eva Schlinger, and Dan Garrette. 2019. How multilingual is multilingual BERT? In Proceedings of ACL, pages 4996-5001.  
+Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. In Proceedings of ICML, pages 8748-8763.  
+Alec Radford and Karthik Narasimhan. 2018. Improving language understanding by generative pretraining. OpenAI Blog.  
+Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of Machine Learning Research, 21(140):1-67.  
+Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. SQuAD: 100,000+ questions for machine comprehension of text. In Proceedings of EMNLP, pages 2383-2392.
+
+Sara Rosenthal, Noura Farra, and Preslav Nakov. 2017. Semeval-2017 task 4: Sentiment analysis in twitter. In Proceedings of ACL 2017 Workshop on Semantic Evaluation, pages 502-518.  
+Sumuk Shashidhar, Abhinav Chinta, Vaibhav Sahai, Zhenhailong Wang, and Heng Ji. 2023. Democratizing llms: An exploration of cost-performance trade-offs in self-refined open-source models. In *Findings of EMNLP*, pages 9070–9084.  
+Richard Socher, Alex Perelygin, Jean Wu, Jason Chuang, Christopher D. Manning, Andrew Ng, and Christopher Potts. 2013. Recursive deep models for semantic compositionality over a sentiment treebank. In Proceedings of EMNLP, pages 1631-1642.  
+Gemma Team, Morgane Riviere, Shreya Pathak, Pier Giuseppe Sessa, Cassidy Hardin, Surya Bhupatiraju, Léonard Hussenot, Thomas Mesnard, Bobak Shahriari, Alexandre Ramé, et al. 2024. Gemma 2: Improving open language models at a practical size. arXiv preprint arXiv:2408.00118.  
+Inar Timiryasov and Jean-Loup Tastet. 2023. Baby llama: Knowledge distillation from an ensemble of teachers trained on a small dataset with no performance penalty. In Proceedings of CoNLL 2024 BabyLM Challenge, pages 279-289.  
+Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Proceedings of NeurIPS, volume 30.  
+Xiao Wang, Tianze Chen, Qiming Ge, Han Xia, Rong Bao, Rui Zheng, Qi Zhang, Tao Gui, and Xuanjing Huang. 2023. Orthogonal subspace learning for language model continual learning. In Findings of EMNLP, pages 10658-10671.  
+Alex Warstadt, Amanpreet Singh, and Samuel R. Bowman. 2019. Neural network acceptability judgments. Transactions of the Association for Computational Linguistics, 7:625-641.  
+Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, et al. 2022. Emergent abilities of large language models. Transactions on Machine Learning Research.  
+Peter West, Chandra Bhagavatula, Jack Hessel, Jena Hwang, Liwei Jiang, Ronan Le Bras, Ximing Lu, Sean Welleck, and Yejin Choi. 2022. Symbolic knowledge distillation: From general language models to commonsense models. In Proceedings of NAACL, pages 4602-4625.  
+Colin White, Mahmoud Safari, Rhea Sukthanker, Binxin Ru, Thomas Elsken, Arber Zela, Debadeepta Dey, and Frank Hutter. 2023. Neural architecture search: Insights from 1000 papers. arXiv preprint arXiv:2301.08727.
+
+Adina Williams, Nikita Nangia, and Samuel Bowman. 2018. A broad-coverage challenge corpus for sentence understanding through inference. In Proceedings of NAACL, pages 1112-1122.  
+Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumont, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Rémi Louf, Morgan Funtowicz, et al. 2020. Transformers: State-of-the-art natural language processing. In Proceedings of EMNLP (Demo Track), pages 38-45.  
+An Yang, Baosong Yang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Zhou, Chengpeng Li, Chengyuan Li, Dayiheng Liu, Fei Huang, et al. 2024. Qwen2 technical report. arXiv preprint arXiv:2407.10671.  
+Jiacheng Ye, Jiahui Gao, Qintong Li, Hang Xu, Jiangtao Feng, Zhiyong Wu, Tao Yu, and Lingpeng Kong. 2022. Zerogen: Efficient zero-shot learning via dataset generation. In Proceedings of EMNLP, pages 11653-11669.  
+Qinyuan Ye. 2024. Cross-task generalization abilities of large language models. In Proceedings of NAACL 2024 Student Research Workshop, pages 255-262.  
+Hao Yu, Zachary Yang, Kellin Peline, Jean-François Godbout, and Reihaneh Rabbany. 2023. Open, closed, or small language models for text classification? arXiv preprint arXiv:2308.10092.  
+Weizhe Yuan, Graham Neubig, and Pengfei Liu. 2021. Bartscore: Evaluating generated text as text generation. In Proceedings of NeurIPS, pages 27263-27277.  
+Marcos Zampieri, Shervin Malmasi, Preslav Nakov, Sara Rosenthal, Noura Farra, and Ritesh Kumar. 2019. Semeval-2019 task 6: Identifying and categorizing offensive language in social media (offenseval). In Proceedings of NAACL 2019 Workshop on Semantic Evaluation, pages 75-86.  
+Renrui Zhang, Jiaming Han, Chris Liu, Aojun Zhou, Pan Lu, Yu Qiao, Hongsheng Li, and Peng Gao. 2024a. LLaMA-adapter: Efficient fine-tuning of large language models with zero-initialized attention. In Proceedings of ICLR.  
+Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q Weinberger, and Yoav Artzi. 2020. Bertscore: Evaluating text generation with bert. In Proceedings of ICLR.  
+Yang Zhang, Yanfei Dong, and Kenji Kawaguchi. 2024b. Investigating layer importance in large language models. In Proceedings of EMNLP 2024 BlackboxNLP Workshop, pages 469-479.  
+Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, et al. 2023. A survey of large language models. arXiv preprint arXiv:2303.18223.
+
+Ming Zhong, Chenxin An, Weizhu Chen, Jiawei Han, and Pengcheng He. 2024. Seeking neural nuggets: Knowledge transfer in large language models from a parametric perspective. In Proceedings of ICLR.
+
+# A Implementation Detail
+
+We implemented PiFi using the PyTorch (Paszke et al., 2019b) and Transformers (Wolf et al., 2020) libraries. For encoder-based LMs, we used the representation of the CLS token as  $h_{enc}$ . Similarly, for encoder-decoder models, we used the hidden states of the last encoder layer for each token as  $h_{enc}$ . Each model was trained for three epochs with a learning rate of 5e-5 and a batch size of 32 using the Adam optimizer (Kingma and Ba, 2015). Table 20 provides the number of parameters used for each PiFi configuration across various SLMs and LLMs. Please refer to attached source code for further details.
+
+# B Ablation Study
+
+In this section, we present the results of ablation studies to supplement the main experiments conducted in Section 4.
+
+# B.1 Effect of Selection of Layers within LLM
+
+For the experiments in Section 4, we consistently used the last layer of each LLM as  $L_{LLM}$ . However, recent studies have shown that different layers of LLMs capture distinct types of information and serve different roles (Artzy and Schwartz, 2024; Zhang et al., 2024b; Liu et al., 2024). To explore this, we analyzed the effectiveness of PiFi when extracting  $L_{LLM}$  from layers other than the last. Specifically, we experimented with layers 1, 4, 8, 12, 16, 20, 24, 28, 32 of the LLMs used in Section 4.7.
+
+Figure 2 presents the experimental results on the SST-2 dataset. The results clearly show that using the last layer yields the best performance compared to other layers, which aligns with previous findings suggesting that the upper layers of LLMs contain more contextual knowledge (Ju et al., 2024).
+
+In addition to classification tasks, we also evaluated the effectiveness of PiFi in generation tasks. Specifically, we conducted a layer-wise performance analysis on the Multi30K translation task. As shown in Figure 3, the last layer consistently achieved the best performance in generation tasks. However, we acknowledge that for certain tasks,
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td><td>Tweet (Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">SmolLM2-135M</td><td>92.48</td><td>87.70</td><td>90.47</td><td>84.42</td><td>78.93</td></tr><tr><td>0.9224</td><td>0.4817</td><td>0.8980</td><td>0.7673</td><td>0.7111</td></tr><tr><td rowspan="2">+PiFi (Llama-3.1-8B)</td><td>93.42</td><td>88.69</td><td>92.20</td><td>86.04</td><td>81.22</td></tr><tr><td>0.9319</td><td>0.4955</td><td>0.9160</td><td>0.7967</td><td>0.7359</td></tr></table>
+
+Table 9: Experimental results of PiFi on a decoder-based model. For this experiment, we used SmolLM2-135B model with Llama-3.1-8B. We report accuracy and F1-score in the upper row and lower row of each cell.
+
+intermediate or other layers may be more suitable. This highlights the importance of task-specific optimization in layer selection. Therefore, in future research, we plan to explore optimization methods for selecting the most suitable layers for different tasks. Additionally, we aim to extend our work by leveraging automated techniques, such as Neural Architecture Search, to identify the optimal layers more efficiently.
+
+# B.2 Performance of PiFi on Decoder-Based Models
+
+The PiFi framework has primarily been tested on encoder-based and encoder-decoder models. However, it is crucial to verify whether it exhibits the same effectiveness in decoder-based models (e.g., GPT-style architectures). To investigate its applicability and effectiveness further, we evaluated the performance of PiFi on a decoder-based model, SmolLM2-135M (Allal et al., 2025). SmolLM2-135M is a relatively small decoder-based model, comparable in size to BERT-base. In this experiment, we added a classification head at the end of the model to enable it to perform various classification tasks.
+
+As shown in Table 9, the decoder-based PiFi model outperforms the base SmolLM2-135M model across all tasks. This result indicates that PiFi is not limited to encoder-based and encoder-decoder models but can also be effectively applied to decoder-based architectures. These findings highlight the high versatility and scalability of PiFi, providing strong evidence for its applicability across a wide range of language model architectures.
+
+# B.3 Comparison between Text Representation Methods in Encoder-based LMs
+
+In this section, we explore different methods for representing text in encoder-based LMs. In our
+
+<table><tr><td>BERTbase
++ PiFi (Llama-3.1-8B)</td><td>SST-2</td><td>IMDB</td><td>Tweet
+(Sentiment)</td></tr><tr><td rowspan="2">CLS</td><td>91.50</td><td>87.09</td><td>92.95</td></tr><tr><td>0.9125</td><td>0.4800</td><td>0.9224</td></tr><tr><td>Avearge Pooling</td><td>91.06</td><td>86.25</td><td>92.70</td></tr><tr><td>(w/o padding)</td><td>0.9079</td><td>0.4724</td><td>0.9197</td></tr><tr><td>Avearge Pooling</td><td>90.40</td><td>86.56</td><td>92.73</td></tr><tr><td>(w/ padding)</td><td>0.9008</td><td>0.4758</td><td>0.9191</td></tr></table>
+
+Table 10: Experimental results comparing the method for representing the input sequence for encoder-based LMs with PiFi. For this experiment, we used  $\mathrm{BERT}_{\text {base }}$  model with Llama-3.1-8B. We report accuracy and F1-score in the upper row and lower row of each cell.  
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.10</td><td>86.90</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.862</td></tr><tr><td rowspan="2">+ PiFi (Llama-3.1-8B)</td><td>91.50</td><td>87.09</td><td>92.95</td></tr><tr><td>0.9125</td><td>0.4800</td><td>0.9224</td></tr><tr><td rowspan="2">+ PiFi-full (Llama-3.1-8B)</td><td>91.27</td><td>86.00</td><td>90.20</td></tr><tr><td>0.9106</td><td>0.4726</td><td>0.8951</td></tr></table>
+
+original PiFi framework, we used the CLS token representation as  $h_{enc}$ . To evaluate alternative approaches, we conducted experiments with two additional methods: (1) representing  $h_{enc}$  as the average of all token representations in the input sequence, including padding tokens, and (2) representing  $h_{enc}$  as the average of all token representations excluding padding tokens.
+
+Table 10 presents the experimental results. The results reveal distinct differences between these configurations, indicating that using the CLS token representation as  $h_{enc}$  is the most effective choice for ensuring optimal model performance.
+
+# B.4 Full fine-tuning of  $L_{LLM}$  in PiFi
+
+In PiFi framework, we fine-tune the model while keeping  $L_{LLM}$  frozen. To validate the impact of this approach, we conducted an experiment where the entire PiFi model, including  $L_{LLM}$ , was fully fine-tuned.
+
+Table 11 shows the results of this experiment. Freezing  $L_{LLM}$  during fine-tuning leads to better performance compared to fully fine-tuning the
+
+Table 11: Experimental results to validate the effectiveness of fully fine-tuning  $L_{\mathrm{LLM}}$  rather than keeping it frozen. We report accuracy and F1-score in the upper row and lower row of each cell.  
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td><td>Tweet (Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.10</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.8620</td><td>0.7727</td><td>0.7398</td></tr><tr><td>+ PiFi</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>(Llama-3.1-8B)</td><td>0.9125</td><td>0.4800</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr><tr><td>+ PiFi</td><td>91.98</td><td>86.72</td><td>92.81</td><td>85.57</td><td>82.45</td></tr><tr><td>(Llama-3.1-8B-Instruct)</td><td>0.9174</td><td>0.4882</td><td>0.9206</td><td>0.7985</td><td>0.7450</td></tr><tr><td>p-value</td><td>0.7352</td><td>0.3237</td><td>0.4422</td><td>0.2129</td><td>0.5308</td></tr></table>
+
+Table 12: Experimental results comparing the usage of base LLM and instruction-tuned LLM for PiFi. For this experiment, we used Llama-3.1-8B and Llama-3.1-8B-Instruct. The p-value denotes the statistical significance between the distribution of accuracy of PiFi with Llama-3.1-8B and Llama-3.1-8B-Instruct. We report accuracy and F1-score in the upper row and lower row of each cell.
+
+model. We hypothesize that freezing  $L_{LLM}$  preserves the extensive knowledge of the LLM, while full fine-tuning may cause catastrophic forgetting of this knowledge. Thus, it is crucial to keep  $L_{LLM}$  frozen to retain the LLM's vast knowledge, optimize the SLM for the desired task, and reduce training costs.
+
+# B.5 Effectiveness of Instruction-tuned LLM for PiFi
+
+Recent innovations in LLMs involve additional training to enable these models to better follow instructions provided in prompts, leading to improved performance on various downstream tasks (Ouyang et al., 2022). In this section, we explore the impact of using instruction-tuned LLMs for extracting  $L_{LLM}$  within the PiFi framework. Specifically, we compare the performance of PiFi when using Llama-3.1-8B, our default model, against Llama-3.1-8B-Instruct, an instruction-tuned version of the Llama-3.1-8B.
+
+The results are presented in Table 12. Our findings indicate that using instruction-tuned LLMs with PiFi can enhance the performance of SLMs, but the performance gains are marginal. In fact, there is no consistent trend showing that instruction-tuned models outperform their base counterparts across all datasets. To better understand this observation, we conducted a statistical significance test by comparing the results obtained from five different random seeds. The p-values for accuracy distributions between the instruction-tuned and base LLMs were greater than 0.05, indicating that these performance differences are not statistically significant.
+
+This suggests that incorporating instruction
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td><td>Tweet (Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERT-base</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.77</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.8620</td><td>0.7727</td><td>0.7296</td></tr><tr><td>+PiFi</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>(Llama-3.1-8B)</td><td>0.9125</td><td>0.4800</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr><tr><td>+PiFi</td><td>91.49</td><td>87.11</td><td>92.73</td><td>85.00</td><td>82.00</td></tr><tr><td>(Llama-3.1-70B)</td><td>0.9120</td><td>0.4875</td><td>0.9198</td><td>0.7784</td><td>0.761</td></tr><tr><td>+PiFi</td><td>91.05</td><td>86.56</td><td>91.90</td><td>85.14</td><td>81.65</td></tr><tr><td>(Qwen2.5-7B)</td><td>0.9080</td><td>0.4765</td><td>0.9117</td><td>0.8007</td><td>0.7562</td></tr><tr><td>+PiFi</td><td>91.12</td><td>86.27</td><td>91.69</td><td>83.75</td><td>80.74</td></tr><tr><td>(Qwen2.5-32B)</td><td>0.9086</td><td>0.4718</td><td>0.9096</td><td>0.7569</td><td>0.7463</td></tr><tr><td>+PiFi</td><td>91.38</td><td>86.75</td><td>92.21</td><td>84.51</td><td>82.45</td></tr><tr><td>(Qwen2.5-72B)</td><td>0.9115</td><td>0.4788</td><td>0.9150</td><td>0.7796</td><td>0.7450</td></tr></table>
+
+Table 13: Experimental results comparing PiFi performance across various large language models. We evaluated PiFi with Llama-3.1 models (8B, 70B) and Qwen2.5 models (7B, 32B, 72B) across five downstream tasks. We report accuracy and F1-score in the upper row and lower row of each cell.  
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td><td>Tweet (Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.77</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.8620</td><td>0.7727</td><td>0.7296</td></tr><tr><td>+PiFi</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>(Llama-3.1-8B)</td><td>0.9125</td><td>0.4800</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr><tr><td>+PiFi</td><td>91.49</td><td>87.11</td><td>92.73</td><td>85.00</td><td>82.00</td></tr><tr><td>(Llama-3.1-70B)</td><td>0.9120</td><td>0.4875</td><td>0.9198</td><td>0.7784</td><td>0.761</td></tr><tr><td>+PiFi-2 layers</td><td>91.76</td><td>87.20</td><td>92.95</td><td>85.66</td><td>82.22</td></tr><tr><td>(Llama-3.1-70B)</td><td>0.9154</td><td>0.4901</td><td>0.9217</td><td>0.8119</td><td>0.7488</td></tr></table>
+
+tuned LLMs does not necessarily lead to significant improvements for PiFi. Instead, it highlights that the intrinsic knowledge encoded within the LLMs is more crucial for PiFi's success than their capacity to follow human instructions.
+
+# B.6 Impact of Larger-Scale LLM on PiFi Performance
+
+Previously, as analyzed in Section 4.8, we examined the performance of PiFi using the Qwen model across different scales (0.5B, 1.5B, and 7B parameters). However, it is necessary to further examine how performance changes when applying larger-scale LLM layers. To this end, we conducted experiments applying PiFi to various large-scale models, including the 8B and 70B versions of Llama-3.1 and the 7B, 32B, and 72B versions of Qwen2.5. We selected Qwen2.5 as it was newly released at the time of writing this paper, allowing us to evalu
+
+Table 14: Experimental results of PiFi with multiple layers. This experiment compares the performance of using a single layer versus two layers from Llama-3.1-70B. We report accuracy in the upper row and F1-score in the lower row of each cell.  
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td><td>Tweet (Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.10</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.8620</td><td>0.7727</td><td>0.7398</td></tr><tr><td rowspan="2">+PiFi( Llama-3.1-8B)</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>0.9125</td><td>0.4800</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr><tr><td rowspan="2">Single layer: First(Llama-3.1-8B)</td><td>80.09</td><td>77.63</td><td>80.14</td><td>78.59</td><td>68.12</td></tr><tr><td>0.7970</td><td>0.4383</td><td>0.7913</td><td>0.6735</td><td>0.3812</td></tr><tr><td rowspan="2">Single layer: Last(Llama-3.1-8B)</td><td>80.67</td><td>78.31</td><td>80.28</td><td>80.47</td><td>69.31</td></tr><tr><td>0.8010</td><td>0.4392</td><td>0.7939</td><td>0.6983</td><td>0.4076</td></tr></table>
+
+Table 15: Experimental results on the usage of a single frozen  $L_{\mathrm{LLM}}$  without SLM. For this experiment, we extracted the first and last layers of Llama-3.1-8B as  $L_{\mathrm{LLM}}$ . We report accuracy and F1-score in the upper row and lower row of each cell.
+
+ate PiFi's scalability using state-of-the-art models with cutting-edge performance.
+
+As shown in Table 13, when comparing various large language models, performance improvements did not consistently correlate with model size. In particular, the use of the Llama-3.1-70B model layer in PiFi resulted in comparable or slightly lower performance than when using the 8B model layer. Similarly, the performance differences between Qwen2.5-7B, Qwen2.5-32B, and Qwen2.5-72B did not show a clear linear relationship with model size. We interpret this outcome as potentially influenced by a factor we refer to as layer knowledge density. Specifically, the 8B model comprises 32 layers, while the 70B model consists of 80 layers (Dubey et al., 2024). This suggests that a single layer in the 70B model may encapsulate relatively less dense knowledge compared to a single layer in the 8B model.
+
+To mitigate the limitations imposed by this lower density, we conducted additional experiments incorporating the last two layers of the 70B model, with results presented in Table 14. This experiment directly compares the performance when using a single layer versus two layers from the 70B model. However, the observed performance improvement was marginal. This suggests that the selected layer combination may not be optimal, and alternative configurations could lead to greater performance gains. Nevertheless, the large number of possible combinations poses a challenge in identifying the most effective configuration. Future research will explore more efficient methods for identifying effective layer combinations, aiming to enhance PiFi's scalability while optimizing performance.
+
+<table><tr><td></td><td>FLOPs (GFLOPs)</td><td>GPU Memory (GB)</td></tr><tr><td>BERTbase</td><td>272.12</td><td>1.33</td></tr><tr><td>+ PiFi(Llama-3.1-8B)</td><td>279.3</td><td>2.27</td></tr></table>
+
+Table 16: Efficiency-performance trade-off analysis results. This analysis was conducted by measuring FLOPs and GPU memory consumption.
+
+# B.7 Evaluating the Usage of a Single LLM Layer
+
+To evaluate the performance of a single frozen LLM Layer without SLM, we conducted experiments by attaching classification heads to the first and last layers of Llama-3.1-8B.
+
+As shown in the Table 15, there is a clear performance gap between the PiFi framework and the approach using a single LLM Layer. This demonstrates that PiFi effectively enhances performance through integration with an SLM (e.g.,  $\mathrm{BERT}_{\mathrm{base}}$ ).
+
+Notably, the single LLM Layer shows limitations in achieving high performance independently, highlighting the necessity of combining it with an SLM for effective knowledge transfer. Furthermore, despite having fewer parameters than PiFi, the single frozen LLM Layer exhibited significantly degraded performance across all tasks. These findings reaffirm that PiFi can enhance the performance of the SLM by effectively integrating knowledge from a single LLM Layer while preserving the efficiency of the SLM.
+
+# B.8 Analyzing Efficiency-Performance Trade-offs in PiFi
+
+As the number of parameters increases, it is necessary to analyze the trade-offs between efficiency and performance. To validate this, we measured FLOPs during the inference process for each model using the fvcore library (Meta, 2019) and GPU memory consumption using the PyTorch Profiler (Paszke et al., 2019a). The results are presented in Table 16.
+
+The measurements revealed that PiFi adds approximately  $2.6\%$  to the baseline inference cost in terms of FLOPs. By using only the CLS token as input to the  $L_{LLM}$ , PiFi significantly reduces the sequence length processed in the  $L_{LLMs}$  to a single token, thereby greatly reducing computational overhead. Although GPU memory consumption showed a relatively larger increase, this trade-off is deemed reasonable for tasks that require perfor
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td><td>Tweet (Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.10</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.8620</td><td>0.7727</td><td>0.7398</td></tr><tr><td rowspan="2">+PiFi (Llama-3.1-8B)</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>0.9125</td><td>0.4800</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr><tr><td>BERT-base</td><td>79.48</td><td>87.11</td><td>81.05</td><td>27.89</td><td>69.31</td></tr><tr><td>+ ZEROGEN (GPT2-XL)</td><td>0.7885</td><td>0.4875</td><td>0.7748</td><td>0.2146</td><td>0.4076</td></tr><tr><td>BERT-base</td><td>81.56</td><td>68.26</td><td>87.99</td><td>25.29</td><td>30.69</td></tr><tr><td>+ ZEROGEN (Llama-3.1-8B)</td><td>0.8106</td><td>0.4343</td><td>0.8637</td><td>0.2061</td><td>0.2306</td></tr></table>
+
+Table 17: Experimental results comparing PiFi and ZEROGEN for knowledge transfer. For ZEROGEN, 200,000 synthetic examples were generated per domain using GPT-2-XL and Llama-3.1-8B as data generators. We report accuracy and F1-score in the upper row and lower row of each cell.  
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet (Sentiment)</td><td>Tweet (Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.77</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.8620</td><td>0.7727</td><td>0.7296</td></tr><tr><td rowspan="2">+PiFi (Llama-3.1-8B)</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>0.9125</td><td>0.4800</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr><tr><td rowspan="2">BERTbase+LoRA</td><td>82.93</td><td>75.22</td><td>81.37</td><td>73.07</td><td>69.98</td></tr><tr><td>0.8207</td><td>0.4287</td><td>0.7677</td><td>0.4490</td><td>0.4252</td></tr></table>
+
+Table 18: Experimental results comparing PiFi and LoRA for parameter-efficient fine-tuning. For LoRA, we used rank=8, scaling factor  $\alpha = 16$ , and dropout=0.1 with the same BERT<sub>base</sub> backbone. We report accuracy and F1-score in the upper row and lower row of each cell.
+
+mance improvement in environments where additional memory resources are available. Moreover, PiFi's modular design allows for flexible scalability according to system constraints, making it adaptable even in resource-constrained environments.
+
+# B.9 Comparison with Knowledge Distillation and Parameter-Efficient Fine-tuning Methods
+
+We also conducted a comparison with ZEROGEN (Ye et al., 2022), which leverages synthetic data generated by an LLM to enhance the SLM as an alternative method for distilling the inherent knowledge of an LLM. Following the settings of prior studies, we generated 200,000 synthetic examples per domain using GPT-2-XL. Additionally, to account for the fact that ZEROGEN's performance depends on the quality of the data generator, we conducted more comprehensive experiments by generating additional synthetic examples using the latest model, Llama-3.1-8B.
+
+As shown in the Table 17, ZEROGEN proved less effective than PiFi in most tasks, which we attribute to noise introduced during the synthetic
+
+<table><tr><td></td><td>SST-2</td><td>IMDB</td><td>Tweet(Sentiment)</td><td>Tweet(Offensive)</td><td>CoLA</td></tr><tr><td rowspan="2">BERTbase</td><td>89.41</td><td>85.10</td><td>86.90</td><td>83.15</td><td>80.77</td></tr><tr><td>0.8907</td><td>0.4733</td><td>0.8620</td><td>0.7727</td><td>0.7296</td></tr><tr><td rowspan="2">+PiFi(Llama-3.1-8B)</td><td>91.50</td><td>87.09</td><td>92.95</td><td>86.03</td><td>82.07</td></tr><tr><td>0.9125</td><td>0.4800</td><td>0.9224</td><td>0.8026</td><td>0.7523</td></tr><tr><td></td><td>+2.09%p</td><td>+1.99%p</td><td>+6.05%p</td><td>+2.88%p</td><td>+1.97%p</td></tr><tr><td rowspan="2">BERTbase-10%</td><td>87.32</td><td>82.32</td><td>85.40</td><td>81.07</td><td>75.14</td></tr><tr><td>0.8691</td><td>0.4541</td><td>0.8463</td><td>0.7420</td><td>0.6866</td></tr><tr><td rowspan="2">+PiFi-10%(Llama-3.1-8B)</td><td>89.96</td><td>84.12</td><td>91.21</td><td>84.31</td><td>77.20</td></tr><tr><td>0.8967</td><td>0.4688</td><td>0.9030</td><td>0.7807</td><td>0.6527</td></tr><tr><td></td><td>+2.64%p</td><td>+1.80%p</td><td>+5.81%p</td><td>+3.24%p</td><td>+2.06%p</td></tr></table>
+
+Table 19: Experimental results of PiFi under limited training data conditions. We used only  $10\%$  of the total training data for each dataset, and report accuracy in the upper row of each cell and F1-score in the lower row.
+
+data generation process. These results suggest that directly and efficiently utilizing the inherent knowledge of an LLM, rather than indirectly leveraging it through synthetic data generation, offers a more advantageous approach.
+
+In addition to knowledge distillation methods, we compared PiFi with parameter-efficient finetuning using LoRA (Hu et al., 2022), which has gained popularity for adapting language models with minimal computational overhead. For our experiments, we applied LoRA to the BERT-base backbone using standard settings (rank=8, scaling factor  $\alpha = 16$ , dropout=0.1) and evaluated its performance on the same set of downstream tasks.
+
+As shown in Table 18, PiFi consistently outperforms LoRA across all downstream tasks. This demonstrates that PiFi achieves superior performance improvements by effectively integrating LLM layers, enabling direct transfer of rich linguistic knowledge and generalization capabilities encoded in large-scale models to SLMs.
+
+# B.10 Performance Evaluation of PiFi under Limited Training Data
+
+To assess whether the PiFi framework operates effectively in a limited training data environment, experiments were conducted using only  $10\%$  of the total training data for each dataset. Specifically, the following sample counts were used: SST-2 (6,920  $\rightarrow$  692), IMDB (20,000  $\rightarrow$  2,000), Tweet for sentiment classification (45,615  $\rightarrow$  4,562), Tweet for offensive classification (11,916  $\rightarrow$  1,192), and CoLA (5,536  $\rightarrow$  554).
+
+As shown in Table 19, even when trained on restricted data, the model with PiFi consistently outperformed the BERT<sub>base</sub>. In particular, the SST-2, Tweet for offensive classification, and CoLA
+
+datasets exhibited relatively greater performance improvements compared to when the full dataset was used, indicating that PiFi can deliver significant benefits even under constrained data conditions. In contrast, the IMDB and Tweet for sentiment classification datasets have a considerably larger amount of total data (20,000 and 45,615 samples, respectively), so even with only  $10\%$  of the data, a sufficient number of training samples was available. Consequently, the relative performance gains were smaller; however, this can be viewed as a positive example of PiFi's flexible applicability even without large-scale data. These results underscore that PiFi offers stable and consistent performance advantages in limited training data environments, thereby enhancing its potential for real-world low-resource scenarios.
+
+![](images/b963f7f036f665016f6cfea62ffeea4c7986f53e395a8cf646614ddf5fb5851d.jpg)  
+Figure 2: Comparison of the performance of PiFi models with different position of  $L_{LLM}$  on SST-2 dataset.
+
+![](images/8069c81bdba4e587232281130365fdcae51cd28ddd9ec87f94950c1aaba1a9e7.jpg)  
+Figure 3: Comparison of the performance of PiFi models with different position of  $L_{LLM}$  on Multi30K dataset.
+
+<table><tr><td></td><td>SLM Params(M)</td><td>\(L_{LLM}\) Params(M)</td><td>\(L_{in}\) Params(M)</td><td>\(L_{out}\) Params(M)</td><td>Head Params(M)</td></tr><tr><td>\(BERT_{base}\) + PiFi (Llama-3.1-8B)</td><td>109.482</td><td rowspan="7">218.112</td><td rowspan="7">3.146</td><td rowspan="7">3.146</td><td rowspan="13">0.592</td></tr><tr><td>\(RoBERTa_{base}\) + PiFi (Llama-3.1-8B)</td><td>124.646</td></tr><tr><td>\(ELECTRA_{base}\) + PiFi (Llama-3.1-8B)</td><td>108.892</td></tr><tr><td>\(DeBERTa_{base}\) + PiFi (Llama-3.1-8B)</td><td>138.602</td></tr><tr><td>\(DeBERTa-V3_{base}\) + PiFi (Llama-3.1-8B)</td><td>183.832</td></tr><tr><td>\(BERT_{base}\) + PiFi (Mistral-7B-v0.1)</td><td>109.482</td></tr><tr><td>\(BERT_{base}\) + PiFi (Mistral-7B-v0.3)</td><td>109.482</td></tr><tr><td>\(BERT_{base}\) + PiFi (Qwen2-0.5B)</td><td rowspan="6">109.482</td><td>14.912</td><td>0.688</td><td>0.688</td></tr><tr><td>\(BERT_{base}\) + PiFi (Qwen2-1.5B)</td><td>46.798</td><td>1.180</td><td>1.180</td></tr><tr><td>\(BERT_{base}\) + PiFi (Qwen2-7B)</td><td>233.058</td><td>3.146</td><td>3.146</td></tr><tr><td>\(BERT_{base}\) + PiFi (Gemma-2-9B)</td><td>198.195</td><td>3.146</td><td>3.146</td></tr><tr><td>\(BERT_{base}\) + PiFi (Falcon-7B)</td><td>207.070</td><td>3.146</td><td>3.146</td></tr><tr><td>\(BERT_{base}\) + PiFi (Llama-3.1-70B)</td><td>855.654</td><td>6.291</td><td>6.291</td></tr><tr><td>\(T5_{base}\) + PiFi (Llama-3.1-8B)</td><td>222.904</td><td rowspan="2">218.112</td><td rowspan="2">3.146</td><td rowspan="2">3.146</td><td rowspan="2">-</td></tr><tr><td>\(BART_{base}\) + PiFi (Llama-3.1-8B)</td><td>139.420</td></tr></table>
+
+Table 20: The parameters of each module for PiFi, across various setup regarding SLMs and LLMs.

ACL/2025/Plug-in and Fine-tuning_ Bridging the Gap between Small Language Models and Large Language Models/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63833e4fc94d1c2957a34cf2a4dd1ebc767df3670a1e0c3f47d3361e1421106b
+size 983056

ACL/2025/Plug-in and Fine-tuning_ Bridging the Gap between Small Language Models and Large Language Models/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ed191dea41e6fd1808150b931ef81db12ff96f88314d115e70fba77c8d2dcb9
+size 581890

ACL/2025/Polishing Every Facet of the GEM_ Testing Linguistic Competence of LLMs and Humans in Korean/d6fae89e-df30-4307-98b5-29828a043d27_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:227ce97923d0002d5bd8f8ea017f461d0b837935f8829b3cfd309356e7f336ef
+size 211475

ACL/2025/Polishing Every Facet of the GEM_ Testing Linguistic Competence of LLMs and Humans in Korean/d6fae89e-df30-4307-98b5-29828a043d27_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:846767730d94cdedd9cc94f4b7539278ad90e6144bfcec225a236a480ef93f43
+size 269204

ACL/2025/Polishing Every Facet of the GEM_ Testing Linguistic Competence of LLMs and Humans in Korean/d6fae89e-df30-4307-98b5-29828a043d27_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a19e99d9b16d982ea0bd41ad264d0a312a82d789813852261f0448e14402b58f
+size 7212302

ACL/2025/Polishing Every Facet of the GEM_ Testing Linguistic Competence of LLMs and Humans in Korean/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b53af5ca1017762c6c7e00cfd80a5adacacd602ae023da10ca5ee8ae323f7132
+size 1535827

ACL/2025/Polishing Every Facet of the GEM_ Testing Linguistic Competence of LLMs and Humans in Korean/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8bd15449be4e8a88106546f5a22372c9390e5830313401febb0ff70b0f74cd30
+size 975696

ACL/2025/PolyNarrative_ A Multilingual, Multilabel, Multi-domain Dataset for Narrative Extraction from News Articles/2e671b1e-ea48-400d-ac9b-a60dac2b849d_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3226eb51c4a23de1449a6736714262d1f544c2b03f0cd66f17814f449b281731
+size 173235

ACL/2025/PolyNarrative_ A Multilingual, Multilabel, Multi-domain Dataset for Narrative Extraction from News Articles/2e671b1e-ea48-400d-ac9b-a60dac2b849d_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e256c02b68fc37796f7902ddf2cc8360f48f178e2501c1e741379bb86bad554
+size 208800

ACL/2025/PolyNarrative_ A Multilingual, Multilabel, Multi-domain Dataset for Narrative Extraction from News Articles/2e671b1e-ea48-400d-ac9b-a60dac2b849d_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f67d712fc718ecf268f2e4ff6d099e32f3c1f1bda6fe883061c43e7d6d9d25c0
+size 1032028

ACL/2025/PolyNarrative_ A Multilingual, Multilabel, Multi-domain Dataset for Narrative Extraction from News Articles/full.md

@@ -0,0 +1,809 @@
+# PolyNarrative: A Multilingual, Multilabel, Multi-domain Dataset for Narrative Extraction from News Articles
+
+Nikolaos Nikolaidis<sup>1</sup>, Nicolas Stefanovitch<sup>2</sup>, Purificação Silvano<sup>3</sup>, Dimitar Dimitrov<sup>4</sup>, Roman Yangarber<sup>5</sup>, Nuno Guimarães<sup>6,3</sup>, Elisa Sartori<sup>7</sup>, Ion Androutsopoulos<sup>1,8</sup>, Preslav Nakov<sup>9</sup>, Giovanni Da San Martino<sup>7</sup>, Jakub Piskorski<sup>10</sup>,
+
+$^{1}$ Department of Informatics, Athens University of Economics and Business,
+
+$^{2}$ European Commission Joint Research Centre,  $^{3}$ University of Porto,
+
+$^{4}$ Sofia University "St. Kliment Ohridski",  $^{5}$ University of Helsinki, INESC TEC, Portugal,  $^{7}$ University of Padova,
+
+$^{8}$ Archimedes, Athena Research Center, Greece,  $^{9}$ MBZUAI,
+
+$^{10}$ Institute of Computer Science, Polish Academy of Science,
+
+nnikon@aueb.gr, jpiskorski@gmail.com
+
+# Abstract
+
+We present PolyNarrative, a new multilingual dataset of news articles, annotated for narratives. Narratives are overt or implicit claims, recurring across articles and languages, promoting a specific interpretation or viewpoint on an ongoing topic, often propagating mis/disinformation. We developed two-level taxonomies with coarse- and fine-grained narrative labels for two domains: (i) climate change and (ii) the military conflict between Ukraine and Russia. We collected news articles in four languages (Bulgarian, English, Portuguese, and Russian) related to the two domains and manually annotated them at the paragraph level. We make the dataset publicly available, along with experimental results of several strong baselines that assign narrative labels to news articles at the paragraph or the document level. We believe that this dataset will foster research in narrative detection and enable new research directions towards more multi-domain and highly granular narrative related tasks.
+
+# 1 Introduction
+
+As online news and social networks have radically altered the media landscape, understanding the dynamics of news propagation and the narratives that news convey is more crucial than ever. Online news outlets make it possible for recurring narratives to appear, often with very different phrasings, in multiple articles and propagate with very high velocity across audiences, languages, and countries. This is especially problematic with manipulative narratives, potentially containing strong biases, mis/disinformation, propaganda, or harmful content. Such a risk is even more pervasive when it is related to divisive issues and can have large societal implications, including destabilization, conflict, hate, or incitement to violence.
+
+![](images/10eb26a41d6849971f72a9c426a80284ca6662c8a3716eb79f3e248ee0acf2b5.jpg)  
+Figure 1: Two articles on different topics (finance vs. car enthusiasts), conveying three different narratives. The three narratives found in the articles are in the round shapes in the middle and the paragraphs are color-coded with the color of the narrative that they contain.
+
+Various definitions of narrative can be found in the literature, depending on the context and the goal of the analysis. In this paper, we define narrative as a recurring, repetitive (across and within articles), overt or implicit claim that presents and promotes a specific interpretation or viewpoint on an ongoing (and frequently dynamic) news topic.
+
+Narratives may be embedded in thousands of articles across multiple topics, themes, languages, and news genres (e.g., news reporting or opinion). They can manifest using very different vocabularies, media frames (Entman, 2007), and tones, which makes their automatic detection and extraction very challenging. Moreover, multiple narratives can be present in the same article, or even in the same paragraph or in the same sentence.
+
+An example is shown in Fig. 1, where the narrative "Downplaying Climate Change" occurs in two articles from very different domains: finance vs. auto enthusiasts.
+
+Thus, the effective automatic detection, extraction, and analysis of narratives in news articles is an important challenge and remains an open problem (§2). Such capacity is crucial for analyzing the news landscape, identifying media bias, and detecting attempts to influence readers. It can also be used to warn and educate media consumers to improve their media literacy. Large Language Models (LLMs) have made the need for such capacity even more urgent, as they are vulnerable to (Xu et al., 2024) and prone to generating manipulative narratives (Vykopal et al., 2024) with modest effort.
+
+In this work, we focus on the problem of narrative detection, treating it as multi-label paragraph-level classification problem (§3). The biggest challenge is the lack of annotated datasets (Nunes et al., 2024), especially for long news articles (as opposed to micro-blogs or snippets). Existing narrative classification datasets for long articles (§2) have been annotated at the document level and assign only a single label per annotation. Narratives, however, are often conveyed by small parts of the articles, and thus it is important to pinpoint those parts, which is why we model the problem as a paragraph-level classification.
+
+We consider two news domains of current relevance ( $\S 3.1$ ): climate change (CC) and the military conflict between Ukraine and Russia (URW). Unlike previous work on narrative classification, where the labels are flat and coarse (Li et al., 2023) ( $\S 2$ ), we developed a two-level taxonomy with coarse- and fine-grained narrative labels (hereafter narratives and sub-narratives) for each topic ( $\S 3.2$ ). We collected news articles in four languages—English, Bulgarian, Portuguese, and Russian—related to the two domains ( $\S 4.1$ ), and manually annotated them at the paragraph level with labels from the corresponding taxonomies ( $\S 4.2$ ). We report statistics about the resulting dataset ( $\S 4.3$ ) and discuss its quality ( $\S 4.4$ ). We make the dataset, dubbed PolyNarrative, publicly available, along with experimental results ( $\S 5$ ) and code of several strong baselines that assign narrative labels to news articles at the paragraph or the document level; in the latter case, we use the union of the paragraph-level ground-truth labels of each document as the correct prediction.
+
+Our contributions are as follows:
+
+- We propose a new two-level hierarchical taxonomy for narrative extraction. We further develop a novel dataset<sup>1</sup> that uses this taxonomy with news article annotations from two levels (coarse- and fine-grained), at the paragraph level (previous work was document level), using multi-label annotation (previous work was mostly single-label) across two current domains (previous work was single-domain), and four languages (previous work was for a single language).  
+- We provide a comprehensive overview of the data acquisition and annotation process, and highlight the challenges and measures taken to address them.  
+- We perform evaluation using several strong multi-label classification methods, and draw conclusions that can benefit future work.
+
+# 2 Related Work
+
+# 2.1 Narratives in News
+
+The term narrative in the context of news articles has been used in a variety of different formulations. This includes modeling them as descriptions of spatiotemporal events (Norambuena et al., 2023; Silvano et al., 2024) and capturing narrative components in the form of semantic relationships between entities (Nunes et al., 2024).
+
+Related but distinct is the problem of Opinion Mining (Liu and Zhang, 2012; Cortis and Davis, 2021), where the task is to extract the opinion of an entity (opinion holder) on an entity or issue (or group thereof), based on an aspect and determine its polarity. Although a restricted subset of narratives can be expressed as opinions, the two concepts serve very different purposes.
+
+In an effort to synthesize various formulations of the concept, Dennison (2021) proposed a refined definition of narratives as "selective depictions of reality across at least two points that can include one or more causal claims, and are generally generalizable and can be applied to multiple situations, as opposed to specific stories". This formulation, although well structured from a theoretical perspective, was not sufficient to properly operationalise the concept for our intention to provide a concrete taxonomy.
+
+Denmark Punishing Farmers for Cow 'Emissions' to 'Fight Global Warming'
+
+Denmark has become the first country to force farmers to comply with the goals of the World Economic Forum's (WEF) "Net Zero" agenda, to supposedly "fight global warming." ...
+
+Criticism of institutions and authorities: Criticism of national governments
+
+Although carbon dioxide is typically blamed for causing "climate change," globalists claim that methane traps about 87 times more heat on a 20-year timescale but top scientists have debunked these claims as a hoax.
+
+A recent peer-reviewed study provided conclusive scientific evidence proving that carbon dioxide (CO2) emissions in Earth's atmosphere cannot cause "global warming."
+
+Questioning the measurements and science: Methodologies/metrics used are unreliable/faulty
+
+Questioning the measurements and science: Greenhouse effect/carbon dioxide do not drive climate change
+
+However, not only did they find that higher levels of CO2 made no difference, but they also proved that it simply isn't possible for increases in carbon dioxide to cause temperatures to rise.
+
+Questioning the measurements and science: Greenhouse effect/carbon dioxide do not drive climate change
+
+Nevertheless, unelected globalists at the United Nations Environment Program claim that livestock accounts for about  $32\%$  of "human-caused methane emissions." ...
+
+Criticism of institutions and authorities: Criticism of international entities
+
+Figure 2: Example annotated article from the Climate Change domain; the paragraph-level annotations are in bold.
+
+Thus, our definition of narrative (§1), while informed by the above general formulation, is based on previous efforts to create taxonomies and datasets for practical use (Kotseva et al., 2023; Li et al., 2023; Coan et al., 2021).
+
+# 2.2 Automatic Narrative Detection
+
+To apply modern NLP methods, as highlighted by Santana et al. (2023), extraction of narratives has been approached as hierarchical event detection (Glavaš et al., 2014), as frame bias detection (Zaghouani et al., 2024), as spatiotemporal entity relationship detection (Nunes et al., 2024), and as multi-class classification (Li et al., 2023).
+
+Li et al. (2023) presented a social media dataset based on short texts from microblogs and trained a BERT-based narrative classifier detecting COVID anti-vaccine claims from and using a flat taxonomy, while Weigand et al. (2022) focused on conspiracy narrative detection by applying a binary classifier to detect the presence of conspiracy theories and applied a topic modeling segmentation without a pre-defined taxonomy. Coan et al. (2021) released an English-only dataset using a two-level taxonomy on climate change denial, annotated at the snippet level from website fragments and using a single label per snippet, while Kotseva et al. (2023) provided a multi-lingual dataset annotated at document-level using a single label per document.
+
+While datasets with individual characteristics (long-articles, hierarchical taxonomies, multilinguality) have been released before, to the best of our knowledge, there has been no other dataset that provides a multi-lingual corpus, annotated at
+
+paragraph level using a hierarchical taxonomy in a multi-label multi-class fashion.
+
+# 2.3 Narrative Taxonomies
+
+Several narrative taxonomies have been created to classify recurring argumentation in online news. Kotseva et al. (2023) created a three-level narrative taxonomy on COVID-19 and used it to classify and to analyze trends over time. Hughes et al. (2021) presented a taxonomy of common anti-vax narratives, organized on several common tropes and rhetorical strategies. Coan et al. (2021) presented a two-level taxonomy for common cases of climate change denial in short snippets. Amanatullah et al. (2023) presented a flat taxonomy of common pro-Russian narratives found in the alleged pro-Kremlin influence campaigns related to the war in Ukraine. In this work, we used the last two taxonomies as a starting point for the two domains of interest.
+
+# 3 Task Description
+
+Hereafter, we use Narrative (capitalized) to refer to a top-level (coarse) label of our taxonomies. A sub-Narrative is a lower-level (fine) label of our taxonomies, representing a narrower claim within a Narrative. For example, “ $CO_2$  is beneficial” is a sub-Narrative of the “Climate Change is beneficial” Narrative of the Climate Change (CC) taxonomy.
+
+A (sub-)Narrative can appear in a variety of contexts and topics, and can manifest using subtle and indirect argumentation or framing, and thus its detection can be very challenging.
+
+For example, the sub-Narrative "Western sanctions will backfire" could be expressed throughout a financial reporting article, or in a small paragraph at the end of an opinion article commenting on military matters.
+
+Thus, in order to better tackle this challenge, our approach focuses on highlighting specific segment(s) of each news article that expresses each (sub-)Narrative. In this sense, we consider a paragraph as a minimal segment. If a (sub-)Narrative is expressed in multiple paragraphs of an article (or the entire article), all the corresponding paragraphs need to be annotated accordingly. Moreover, one paragraph can express more than one Narrative and more than one level of Narratives (Narratives and sub-Narratives). In these scenarios, the paragraph needs to be annotated with all the applicable (sub-)Narratives. Given this formulation, the task becomes a hierarchical multi-label multi-class classification problem at the paragraph level.
+
+# 3.1 Domains
+
+We selected two domains (topics) that currently receive extensive news coverage, are presented through several different perspectives, and are highly susceptible to manipulation through the creation of artificial narratives. The first domain is Climate Change (CC) and the second one is the Ukraine-Russia War (URW). In both cases, we have numerous accounts of repeated argumentation in the form of narratives. We are particularly interested in potentially manipulative narratives, and this is reflected in the choice of the (sub-)Narratives of our taxonomies. However, we do not assume that every claim corresponding to a (sub-)Narrative of our taxonomies is necessarily a case of mis/disinformation. This is important to highlight, as mixing legitimate with mis/disinformative claims is a frequent manipulative practice (Goel et al., 2023).
+
+# 3.2 Taxonomies
+
+To create the two taxonomies, we started from existing narrative taxonomies for URW (Amanatullah et al., 2023) and CC (Coan et al., 2021). We asked media analysts to suggest modifications based on their media monitoring experience. These modifications included additions (new (sub-)Narratives), elaborations (splitting existing (sub-)Narratives into multiple different ones), modifications (refining definitions and names) and merging (grouping into common (sub-)Narratives). We asked, when possible, to phrase each (sub-)Narrative as a con
+
+crete claim (e.g., "Renewable energy is dangerous" or "Ukraine is a puppet of the West"). When the argumentation regarding a (sub-)Narrative was very diverse or fragmented, we gave a more general descriptive label (e.g., "Criticism of Institutions" or "Speculating war outcomes"). We ended up with 38 and 36 sub-Narratives, grouped into 11 and 10 Narratives for URW and CC, respectively.
+
+The top level of the two taxonomies (Narratives) is shown in Figure 3. The full taxonomies, including sub-Narratives and definitions, can be found in Appendix A. We note that the two taxonomies are not complete; in fact, they are not intended to be complete, rather they are representative of the practical experience of the media analysts, and different valid formulations are possible. We stress again that although our taxonomies focus on potentially manipulative narratives, we do not assume that every claim corresponding to a (sub-)Narrative of our taxonomies is necessarily mis/disinformation or indicates strong bias.
+
+# UKRAINE-RUSSIA WAR (URW)
+
+Blaming the war on others rather than the invader
+
+Discrediting Ukraine
+
+Russia is the Victim
+
+Praise of Russia
+
+Overpraising the West
+
+Speculating war outcomes
+
+Discrediting the West, Diplomacy
+
+Negative Consequences for the West
+
+Distrust towards Media
+
+Amplifying war-related fears
+
+Hidden plots by secret schemes of powerful groups
+
+# CLIMATE CHANGE (CC)
+
+Criticism of climate policies
+
+Criticism of institutions and authorities
+
+Climate change is beneficial
+
+Downplaying climate change
+
+Questioning the measurements and science
+
+Criticism of climate movement
+
+Controversy about green technologies
+
+Hidden plots by secret schemes of powerful groups
+
+Amplifying Climate Fears
+
+Green policies are geopolitical instruments
+
+Figure 3: Coarse-grained Narratives for the Ukraine-Russia War (URW) and Climate Change (CC) domains. The sub-Narratives are given in Appendix A.
+
+# 4 The PolyNarrative Dataset
+
+# 4.1 Data Acquisition
+
+To obtain news articles for annotation, we used an in-house online news scraping and indexing tool to retrieve articles on the two selected domains in the four target languages. We considered both mainstream and "alternative" news sources (that is, sources known to repeatedly publish content
+
+with misinformation). We also developed ad-hoc scoring methods, outlined below, to heuristically evaluate the relevance of each article to the corresponding domain. More specifically:
+
+1. We created keyword-based queries for the two domains in all the languages, and used them to retrieve a large number of articles from the news article index of our in-house tool. For the URW domain, we included documents from 2021 to 2024. For the CC domain, the documents ranged from 2015 to 2024.  
+2. To assess the relevance of the resulting articles, we formulated another set of key phrases, corresponding to the (sub-)Narratives of the taxonomies (e.g., "Ukraine is corrupt", "Science is debunked") or other aspects we wished to emphasize or de-emphasize. We then used bart-large-mnli<sup>3</sup> to perform zero-shot classification with the title and the first 300 characters of each article, resulting in a classification score per article and key phrase.  
+3. We used an XLM-RoBERTa-based  $^{4}$  multi-label classifier, trained on the Persuasion Techniques dataset (Piskorski et al., 2023a,b), and used the approach of Nikolaidis et al. (2024) to produce Persuasiveness Score metrics per article. For the CC domain, we also used a climate change denial classifier, based on the work of Coan et al. (2021), to further filter and reduce the number of articles to review.  
+4. We used a linear combination of the resulting scores (relevance score per key phrase, Persuasiveness Score metrics and the score from the climate change denial classifier) to automatically rank articles from most to least likely to contain relevant (sub-)Narratives.
+
+A member of the team who was familiar with the news landscape of the language, then manually inspected the resulting set of articles, focusing on articles that seemed relevant to the defined taxonomies. Since manual inspection was time-consuming, we used the different scores computed above to reorder the articles and focus mainly on highly ranked articles.
+
+# 4.2 Data Annotation
+
+In order to clarify the annotation process, including the meaning of key concepts such as Narrative,
+
+sub-Narrative, the precise meaning of each label of the taxonomies, and good practices, we created a document with detailed annotation guidelines. This included specific definitions and examples for each label, together with a detailed description of the annotation process to be followed. The definitions and examples of the taxonomies can be found in Appendix A.
+
+As we wanted to produce a multilingual dataset with four languages, we had multiple annotator teams, each focused on one language. We assigned two annotators per news article and a curator who handled consolidation and quality assurance before the final version. We held regular calls in which hard cases were highlighted, comparisons across annotations from different languages were made, and issues with ambiguities of the guidelines were discussed. To annotate the documents, we used the Inception annotation tool (Klie et al., 2018).
+
+Each language was assigned a coordinator who had the responsibility to sort through the collected articles, assign annotators and curators, and create annotation batches incrementally. The coordinator was also responsible for training the annotators.
+
+We instructed the annotators to read each news article paragraph by paragraph, and to annotate each of them with all the applicable Narratives and sub-Narratives before moving to the next one. When no Narrative or sub-Narrative was found, the annotator simply moved to the next paragraph. Because the sub-Narrative labels were many and difficult to memorize, we configured Inception so that for each paragraph, the annotator would first select all the applicable Narratives, and then all the applicable sub-Narratives from a menu displaying the sub-Narratives of the selected Narratives only. When a paragraph was deemed to contain a Narrative, but none of its sub-Narratives, the "[Narrative]: Other" label was selected at the sub-Narrative level. Similarly, the "Other" label was selected when no Narrative was applicable.
+
+During the annotation process, several issues were highlighted. First, the distribution of labels was very uneven and varied noticeably across languages (Table 2). This reflects the reality of the news discourse, as some Narratives are more frequent in some countries than others. After each annotation batch, the collected label statistics were used by the coordinator to optimize the selection of articles for the next batch. Second, some labels were semantically very close, which led to frequent confusion. We used frequency metrics and cura
+
+tor feedback to discuss such cases in the regular meetings, where we reiterated and further refined the guidelines. For example, the sub-Narratives "Ukraine is a puppet of the West" and "The West does not care about Ukraine, only their interests" were frequently confused. To resolve this ambiguity, we added guidelines instructing the annotators to select the former when Ukraine was the subject and the latter when the West was.
+
+# 4.3 Dataset Statistics
+
+Table 1 shows statistics about the documents of the PolyNarrative dataset, broken down per language, and aggregated over both domains. We can see that, except for Russian, all languages have approximately 400 annotated training documents; the test documents are approximately 35 for each language. We aimed at articles of relatively long text with a median length of 500 words and a minimum length of 250 words.
+
+Even after the rigorous data selection process, the final label distribution after annotation is highly imbalanced (see Table 2 and Appendix B for a breakdown per language). Additionally, the distribution of labels varies noticeably from language to language, due to differences in media interest across countries, as already discussed. For example, for URW, Russian is a notable outlier in terms of distribution, with "Praise of Russia" being one of the most common Narratives. One note to highlight, is that after the article selection process, the proportion of mainstream and alternative news varies widely across languages and could have an impact on label distribution.
+
+Moreover,  $44\%$  and  $54\%$  of the documents in CC and URW, respectively, are assigned more than one label; for paragraphs, the corresponding percentages are  $12\%$  and  $15\%$ , respectively. This observation justifies our choice to perform multi-label annotation.
+
+# 4.4 Dataset Quality
+
+Table 3 reports the inter-annotator agreement scores for Narratives (coarse labels) and sub-Narratives (fine-grained labels), per language, collectively for both domains, measured as Krippendorff's  $\alpha$  at paragraph level using the simplifiedorff $^5$  library. The agreement is under the recommended value of 0.667 but is higher than the IAA of tasks of similar complexity (Piskorski et al., 2023b, 2024).
+
+<table><tr><td rowspan="2">Language</td><td colspan="4">TRAIN</td></tr><tr><td>#documents</td><td>#paragraphs</td><td>#sent.</td><td>#char.</td></tr><tr><td>BG</td><td>400</td><td>3,951</td><td>6,728</td><td>937,150</td></tr><tr><td>EN</td><td>394</td><td>3,683</td><td>8,012</td><td>1,176,513</td></tr><tr><td>PT</td><td>395</td><td>3,877</td><td>5,739</td><td>971,356</td></tr><tr><td>RU</td><td>133</td><td>583</td><td>1,630</td><td>206,124</td></tr></table>
+
+<table><tr><td rowspan="2">Language</td><td colspan="4">TEST</td></tr><tr><td>#documents</td><td>#paragraphs</td><td>#sent.</td><td>#char.</td></tr><tr><td>BG</td><td>35</td><td>325</td><td>542</td><td>70,291</td></tr><tr><td>EN</td><td>41</td><td>584</td><td>1,106</td><td>144,401</td></tr><tr><td>PT</td><td>35</td><td>330</td><td>578</td><td>86,132</td></tr><tr><td>RU</td><td>32</td><td>154</td><td>478</td><td>56,872</td></tr><tr><td>Total</td><td>1,465</td><td>13,487</td><td>24,813</td><td>3,649k</td></tr></table>
+
+Table 1: Document statistics per language, collectively for both domains (CC, URW).  
+
+<table><tr><td>Narrative</td><td>#para</td><td>#doc</td></tr><tr><td>CC: Amplifying Climate Fears</td><td>1,024</td><td>238</td></tr><tr><td>CC: Climate change is beneficial</td><td>29</td><td>12</td></tr><tr><td>CC: Controversy about green technologies</td><td>101</td><td>31</td></tr><tr><td>CC: Criticism of climate movement</td><td>230</td><td>65</td></tr><tr><td>CC: Criticism of climate policies</td><td>234</td><td>102</td></tr><tr><td>CC: Criticism of institutions and authorities</td><td>467</td><td>171</td></tr><tr><td>CC: Downplaying climate change</td><td>135</td><td>49</td></tr><tr><td>CC: Green policies are geopolitical instruments</td><td>14</td><td>11</td></tr><tr><td>CC: Hidden plots by secret schemes of powerful groups</td><td>138</td><td>50</td></tr><tr><td>CC: Questioning the measurements and science</td><td>83</td><td>29</td></tr><tr><td>URW: Amplifying war-related fears</td><td>486</td><td>185</td></tr><tr><td>URW: Blaming the war on others rather than the invader</td><td>282</td><td>149</td></tr><tr><td>URW: Discrediting Ukraine</td><td>1,069</td><td>366</td></tr><tr><td>URW: Discrediting the West, Diplomacy</td><td>862</td><td>331</td></tr><tr><td>URW: Distrust towards Media</td><td>80</td><td>39</td></tr><tr><td>URW: Hidden plots by secret schemes of powerful groups</td><td>53</td><td>20</td></tr><tr><td>URW: Negative Consequences for the West</td><td>152</td><td>79</td></tr><tr><td>URW: Overpraising the West</td><td>34</td><td>23</td></tr><tr><td>URW: Praise of Russia</td><td>602</td><td>271</td></tr><tr><td>URW: Russia is the Victim</td><td>318</td><td>167</td></tr><tr><td>URW: Speculating war outcomes</td><td>161</td><td>79</td></tr><tr><td>Total annotations</td><td>6,554</td><td>2,467</td></tr></table>
+
+Table 2: Distribution statistics for coarse-grained labels (Narratives), assigned to paragraphs and documents, in both TRAIN and TEST splits. Statistics for fine-grained (sub-Narratives) are shown in Appendix B.
+
+As one would expect, the agreement on finer labels is lower, because of their subtler differences.
+
+We noticed that the CC domain caused more confusion between the annotators than URW. In both cases, we could see that the confusion was skewed by a small set of under-agreed labels (5 for URW and 7 for CC) that achieved disagreement above  $40\%$  and  $60\%$ , respectively. If we excluded these labels, the IAA for all languages rises to 0.567 and 0.560 for the coarse and 0.452 and 0.516 for fine-grained, for CC and URW respectively.
+
+There were some sub-Narratives that were commonly confused. For example, in the URW subset "Discrediting the West, Diplomacy: West is tired of Ukraine", "Discrediting the West, Diplomacy: The West does not care about Ukraine, only about its
+
+<table><tr><td>granularity</td><td>lang. domain</td><td>BG</td><td>EN</td><td>PT</td><td>RU</td><td>all</td></tr><tr><td rowspan="3">coarse</td><td>ALL</td><td>0.736</td><td>0.499</td><td>0.461</td><td>0.427</td><td>0.571</td></tr><tr><td>CC</td><td>0.652</td><td>0.375</td><td>0.465</td><td>-</td><td>0.524</td></tr><tr><td>URW</td><td>0.700</td><td>0.558</td><td>0.362</td><td>0.427</td><td>0.533</td></tr><tr><td rowspan="3">fine</td><td>ALL</td><td>0.642</td><td>0.388</td><td>0.385</td><td>0.415</td><td>0.480</td></tr><tr><td>CC</td><td>0.541</td><td>0.283</td><td>0.331</td><td>-</td><td>0.408</td></tr><tr><td>URW</td><td>0.626</td><td>0.457</td><td>0.349</td><td>0.415</td><td>0.479</td></tr></table>
+
+Table 3: Krippendorff's  $\alpha$  for different granularities, languages and domains on paragraph level.
+
+interests”, and “Discrediting Ukraine: Ukraine is a puppet of the West”.
+
+Overall, on a positive note, the majority of the disagreement between the two annotators in the sub-Narrative labels was between labels of the same Narrative (e.g. sub-Narratives under "Discrediting the West, Diplomacy" were frequently confused with one another) On average, out of all the individual annotator disagreements (paragraphs where the two annotators picked a different sub-Narrative),  $67\%$  was with sub-Narratives of the same Narrative. A detailed breakdown of the disagreements statistics for each label can be found in Appendix E.
+
+# 5 Experiments and Evaluation
+
+# 5.1 Experimental Setup
+
+To provide baseline results for the new PolyNarrative dataset, we performed experiments with commonly used multi-label classification methods. We assessed the following configurations:
+
+- Label granularity: We assessed the performance in both coarse (predicting Narratives) and fine-grained (predicting sub-Narratives) settings. Additionally, we assessed how the extra information about the fine-grained predictions impacted coarse-grained predictions.  
+- Paragraph vs. document: We assessed the performance when the models were trained and evaluated to classify paragraphs vs. entire documents. The document-level labels were produced by taking the union of the labels of all of the paragraphs contained in each document.  
+- Prompting vs. fine-tuning: We also assessed the capability of (open-source) LLMs to annotate (sub-)Narratives without further training, given instructions similar to the guidelines given to the human annotators, comparing performance to fine-tuned models.
+
+We experimented with the following models:
+
+- XLM-RoBERTaLarge (Conneau et al., 2020), in two versions:  
+- XLM-RoBERTa-coarse: trained using only the coarse labels.  
+- XLM-RoBERTa-fine: trained using the fine-grained labels.  
+- Llama3.1-70b (Grattafori et al., 2024), without fine-tuning, in four versions:  
+- Llama3.1-70b-0shot-labels: Prompted with the names of the labels of the taxonomy only, in a zero-shot fashion.  
+- Llama3.1-70b-guidelines: Prompted with the full annotation guidelines, which include hand-picked examples for each (sub-)Narrative.  
+- Llama3.1-70b-guidelines-labels: Prompted with the names of the labels, followed by the full annotation guidelines.  
+- Llama3.1-70b-hierarchical: two-step hierarchical prompting, where the model is instructed to pick the coarse-grained labels first, and then given these coarse-grained predictions, decide on the fine-grained label; full guidelines are also provided.
+
+When using Llama, after receiving the model's response, we performed a series of post-processing operations, where the generated output labels were filtered and normalized to remove hallucinated labels and to handle cases where the model generated labels in a slightly different form (e.g., different capitalization, punctuation errors).
+
+To extract paragraph-level annotations from Llama, we inserted extra "[paragraph_N:] tags at the beginning of each paragraph,  $N$  being the number of the paragraph, and modified the prompts to instruct the LLM to output each extra tag followed by the relevant labels of the corresponding paragraph. After the call to the LLM, we recombined the outputs with the input paragraphs in a post-processing phase, using regular expressions. The exact prompts used are given in Appendix C.
+
+The XLM-RoBERTa (Conneau et al., 2020) models were trained at the token level. For each token, a multi-hot vector equal in size to the number of labels was provided as ground-truth, and each
+
+<table><tr><td rowspan="3">Domain</td><td rowspan="3">Model</td><td colspan="4">document</td><td colspan="4">paragraph</td></tr><tr><td colspan="2">coarse</td><td colspan="2">fine</td><td colspan="2">coarse</td><td colspan="2">fine</td></tr><tr><td>F1</td><td>F1_sample</td><td>F1</td><td>F1_sample</td><td>F1</td><td>F1_sample</td><td>F1</td><td>F1_sample</td></tr><tr><td rowspan="7">CC</td><td>XLM-RoBERTa-fine</td><td>0.457</td><td>0.523</td><td>0.198</td><td>0.347</td><td>0.282</td><td>0.161</td><td>0.146</td><td>0.116</td></tr><tr><td>XLM-RoBERTa-coarse</td><td>0.423</td><td>0.550</td><td>-</td><td>-</td><td>0.266</td><td>0.189</td><td>-</td><td>-</td></tr><tr><td>Llama3.1-70b-0shot-labels</td><td>0.542</td><td>0.577</td><td>0.144</td><td>0.198</td><td>0.349</td><td>0.149</td><td>0.076</td><td>0.043</td></tr><tr><td>Llama3.1-70b-guidelines</td><td>0.503</td><td>0.493</td><td>0.151</td><td>0.142</td><td>0.326</td><td>0.128</td><td>0.083</td><td>0.044</td></tr><tr><td>Llama3.1-70b-guidelines-labels</td><td>0.573</td><td>0.505</td><td>0.168</td><td>0.199</td><td>0.352</td><td>0.138</td><td>0.085</td><td>0.050</td></tr><tr><td>Llama3.1-70b-hierarchical</td><td>0.534</td><td>0.548</td><td>0.130</td><td>0.226</td><td>0.302</td><td>0.146</td><td>0.055</td><td>0.053</td></tr><tr><td>Stratified Random Baseline</td><td>0.106</td><td>0.223</td><td>0.016</td><td>0.090</td><td>0.062</td><td>0.071</td><td>0.008</td><td>0.026</td></tr><tr><td rowspan="7">URW</td><td>XLM-RoBERTa-fine</td><td>0.410</td><td>0.486</td><td>0.210</td><td>0.380</td><td>0.257</td><td>0.143</td><td>0.128</td><td>0.105</td></tr><tr><td>XLM-RoBERTa-coarse</td><td>0.428</td><td>0.524</td><td>-</td><td>-</td><td>0.261</td><td>0.185</td><td>-</td><td>-</td></tr><tr><td>Llama3.1-70b-0shot-labels</td><td>0.475</td><td>0.425</td><td>0.210</td><td>0.245</td><td>0.297</td><td>0.124</td><td>0.126</td><td>0.071</td></tr><tr><td>Llama3.1-70b-guidelines</td><td>0.359</td><td>0.370</td><td>0.155</td><td>0.182</td><td>0.232</td><td>0.121</td><td>0.069</td><td>0.046</td></tr><tr><td>Llama3.1-70b-guidelines-labels</td><td>0.432</td><td>0.373</td><td>0.190</td><td>0.249</td><td>0.291</td><td>0.125</td><td>0.117</td><td>0.078</td></tr><tr><td>Llama3.1-70b-hierarchical</td><td>0.381</td><td>0.420</td><td>0.172</td><td>0.173</td><td>0.205</td><td>0.115</td><td>0.073</td><td>0.048</td></tr><tr><td>Stratified Random Baseline</td><td>0.148</td><td>0.179</td><td>0.033</td><td>0.046</td><td>0.044</td><td>0.042</td><td>0.013</td><td>0.018</td></tr></table>
+
+Table 4: Test set results when classifying documents or paragraphs, for Narratives (coarse) and sub-Narratives (fine), averaged over all languages, for Climate Change (CC) and Ukraine-Russia War (URW), respectively.
+
+token inherited the ground-truth labels of its paragraph. At inference time, we obtained a similar predicted multi-hot vector from each token, and each label was assigned (or not) based on the majority vote of the paragraph's tokens. To bypass the 512 token limitation of XLM-R, we used a sliding window with a  $50\%$  overlap.
+
+We also report the results for a naive stratified random baseline, where we randomly draw labels from a multinomial distribution, while respecting the frequency statistics in the training set.
+
+# 5.2 Experimental Results
+
+Since our task is multi-label multi-class classification with a large number of classes, and noticeable class imbalance, we opted to use suitable metrics that are robust in these conditions. Table 4 presents the  $\mathrm{F_1}$  macro and sample-averaged  $\mathrm{F_1}$  scores for all models in both domains (CC, URW). Sample-average  $\mathrm{F_1}$  score is calculated by calculating the  $\mathrm{F_1}$  score on each instance and subsequently averaging over all instances. We observe that the XLM-RoBERTalarge classifier outperforms Llama3.1-70B in the fine-grained case both for document- and paragraph-level predictions, especially in sample  $\mathrm{F_1}$ , where the difference is very pronounced. The Llama3.1 prompted model seems to be more competitive in coarse-grain evaluation.
+
+Interestingly, including the full guidelines in the prompt generally achieves worse results than using the taxonomy labels only. When we add both the taxonomy labels and the guidelines, performance improves, compared to using only the guidelines,
+
+but it is unclear whether it is better to include only the taxonomy labels. Hierarchical prompting does not seem to yield competitive results.
+
+In terms of per-label performance, the weakest model performance was measured on the sub-Narratives under "Downplaying climate change" and "Green policies are geopolitical instruments" Narratives in the CC domain and "Russia is the Victim" and "Blaming the war on others rather than the invader" in the URW domain.
+
+One key observation is that models exhibited relatively lower performance for the “[Narrative]: Other” labels, that is, the paragraphs that contained a Narrative but none of the listed sub-Narratives, that may include a very diverse set of argumentation, coming from a set of heterogeneous sub-Narratives that our taxonomy does not cover.
+
+Performance also seemed to vary noticeably per language. In case of Russian, Llama3.1-70b models seemed to perform rather poorly on a per-sample level. On XLM-RoBERTa, English consistently exhibited poorer performance compared to other languages, with Bulgarian and Portuguese showing different performance per domain. We provide a detailed breakdown of the performance (by label and by language) of two models (XLM-RoBERTa  $f_{ine}$  and Llama3.1-70b-guidelines-labels) on Appendix D.
+
+# 6 Conclusion and Future work
+
+We present PolyNarrative, a new multi-lingual, multi-label, multi-domain dataset for extraction of narratives at the paragraph level from long news
+
+articles. The dataset comprises 1,476 articles totaling 13,625 paragraphs in 4 languages (Bulgarian, English, Portuguese, and Russian), annotated using two expert-refined Narrative hierarchical taxonomies with 38 and 36 fine-grained and 10 and 11 coarse-grained narratives for each domain. We describe in detail the data acquisition and annotation process and highlighted noteworthy issues.
+
+We present preliminary experimental results using both fine-tuned encoder (XLM-R) and prompting-based decoder (Llama 3.1) Language Models in multiple granularities (fine, coarse), levels (paragraph, document), and configurations (prompting strategies). We also highlight interesting findings regarding model performance.
+
+We hope that this dataset will catalyze new research directions in narrative detection and extraction, and stimulate the development of new methods and techniques. We highlight that the taxonomies are based on expert real-world experience and they are not meant to be perceived as complete or to capture the whole breadth of the media discourse around the two domains. We encourage the research community to propose extensions that potentially capture more diverse perspectives. We encourage the direction of developing methods that are domain- and potentially taxonomy-agnostic, able to perform narrative detection in a completely unsupervised fashion, removing the dependence on human subjectivity. We believe that a multi-domain and highly granular dataset such as this one can facilitate experimentation towards this goal.
+
+Regarding future work, we plan to experiment with ways to perform generation of new narrative labels and automatic revision of taxonomy and guidelines. One way to approach this is by analyzing the structure of narrative elements in the text relate them with narrative labels and identify new structures. We also intend to use retrieval strategies to perform narrative retrieval within large news corpora.
+
+# 7 Ethics Policy
+
+Intended Use and Misuse Potential The main drive behind the creation of our dataset was to advance research on automated narrative classification and the detection of deceptive content across multiple languages and domains. However, given that possible risk of exploiting the dataset to boost the production of biased manipulative disinformation attempts, we advise responsible use. In this
+
+context, whoever develops a Narrative detection system is also responsible for deciding which Narratives to detect, in an ethical manner.
+
+Furthermore, this research in direction can contribute to the development of independent tools such as browser plugins or independent publicly available services that could help users contextualize the information they consume, contributing to their awareness and improving citizens' media literacy.
+
+Environmental Impact The deployment of LLMs might have a large carbon footprint, especially when training new models. In the context of the reported experiments, we did not train any new LLMs, but only used existing trained models in an in-context zero-shot scenarios, which is relatively cheap in terms of computing.
+
+Fairness The majority of the annotators, primarily researchers with linguistic background and prior annotation experience, come from the institutions of the co-authors of this manuscript. They were fairly remunerated as part of their job.
+
+The remaining part of the annotator pool consisted of (a) some students from the respective academic organizations, (b) few external experienced analysts paid at rates set by their contracting institutions, and (c) experts from a contracted a professional annotation company, who were compensated according to rates based on their country of residence.
+
+# 8 Limitations
+
+Taxonomies and Dataset Representativeness Our taxonomies were edited by experienced media analysts, active in the study of misinformation and fact-checking. As such, the taxonomies over-represent Narratives of interest of media analysts from Western institutions. The selection should not be perceived as covering the complete discourse of the two domains, but rather what such analysts encounter in practice. The original taxonomy for URW used before the heavy revision from the analysts, was not from a peer-reviewed publication but from a technical report.
+
+The dataset presented covers two widely discussed domains around the world and a wide range of media sources. However, it should not be considered as representative of the media in any specific country, nor should it be considered as balanced in any way. Also, it is important to note that while this
+
+work focuses on narratives potentially containing mis/disinformation, the operationalized definition of Narrative can also be used to detect neutral or desirable viewpoints.
+
+Biases Although the annotators were trained and made acquainted with the specifics of the two domains of interest for our task and cross-language quality control mechanisms have been put in place in the annotation process, we are aware that some degree of intrinsic subjectivity will inevitably be present in the dataset. Consequently, models trained on this dataset might exhibit certain biases.
+
+# Acknowledgements
+
+This research is partially funded by the EU NextGenerationEU, through the National Recovery and Resilience Plan of the Republic of Bulgaria, project SUMMIT, No BG-RRP-2.004-0008.
+
+Also, this work has been partially supported by project MIS 5154714 of the National Recovery and Resilience Plan Greece 2.0 funded by the European Union under the NextGenerationEU Program.
+
+Giovanni Da San Martino would like to thank the Qatar National Research Fund, part of Qatar Research Development and Innovation Council (QRDI), for funding this work by grant NPRP14C0916-210015. He also would like to thank the European Union under the National Recovery and Resilience Plan (NRRP), Mission 4 Component 2 Investment 1.3 - Call for tender No. 341 of March 15, 2022 of Italian Ministry of University and Research - NextGenerationEU; Code PE00000014, Concession Decree No. 1556 of October 11, 2022 CUP D43C22003050001, Progett to "SEcurity and Rights in the CyberSpace (SERICS) - Spoke 2 Misinformation and Fakes - DEcision support IoT systEm foR cybeR intelligENCE (Deterrence) for also funding this work.
+
+# 9 References
+
+# References
+
+Samy Amanatullah, Serena Balani, Angela Fraioli, Stephanie LeMasters, and Mike Gordon. 2023. Tell us how you really feel: Analyzing pro-kremlin propaganda devices & narratives to identify sentiment implications.  
+Travis G. Coan, Constantine Boussalis, John Cook, and Mirjam O. Nanko. 2021. Computer-assisted classification of contrarian claims about climate change. Scientific Reports, 11(1):22320.
+
+Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettle-moyer, and Veselin Stoyanov. 2020. Unsupervised cross-lingual representation learning at scale. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8440-8451, Online. Association for Computational Linguistics.  
+Keith Cortis and Brian Davis. 2021. Over a decade of social opinion mining: a systematic review. Artificial Intelligence Review, 54(7):4873-4965.  
+James Dennison. 2021. Narratives: a review of concepts, determinants, effects, and uses in migration research. Comparative Migration Studies, 9(1):50.  
+Robert M. Entman. 2007. Framing bias: Media in the distribution of power. Journal of Communication, 57(1):163-173.  
+Goran Glavaš, Jan Šnajder, Parisa Kordjamshidi, and Marie-Francine Moens. 2014. HiEve: A Corpus for Extracting Event Hierarchies from News Stories. In Proceedings of 9th language resources and evaluation conference, pages 3678-3683. ELRA.  
+Pranav Goel, Jon Green, David Lazer, and Philip Resnik. 2023. Mainstream news articles co-shared with fake news buttress misinformation narratives. *Preprint*, arXiv:2308.06459.  
+Aaron Grattaftori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, Arun Rao, Aston Zhang, Aurelien Rodriguez, Austen Gregerson, Ava Spataru, Baptiste Roziere, Bethany Biron, Binh Tang, Bobbie Chern, Charlotte Caucheteux, Chaya Nayak, Chloe Bi, Chris Marra, Chris McConnell, Christian Keller, Christophe Touret, Chunyang Wu, Corinne Wong, Cristian Canton Ferrer, Cyrus Nikolaidis, Damien Al-lonsius, Daniel Song, Danielle Pintz, Danny Livshits, Danny Wyatt, David Esiobu, Dhruv Choudhary, Dhruv Mahajan, Diego Garcia-Olano, Diego Perino, Dieuwke Hupkes, Egor Lakomkin, Ehab AlBadawy, Elina Lobanova, Emily Dinan, Eric Michael Smith, Filip Radenovic, Francisco Guzmán, Frank Zhang, Gabriel Synnaeve, Gabrielle Lee, Georgia Lewis Anderson, Govind Thattai, Graeme Nail, Gregoire Mialon, Guan Pang, Guillem Cucurell, Hailey Nguyen, Hannah Korevaar, Hu Xu, Hugo Touvron, Iliyan Zarov, Imanol Arrieta Ibarra, Isabel Kloumann, Ishan Misra, Ivan Evtimov, Jack Zhang, Jade Copet, Jaewon Lee, Jan Geffert, Jana Vranes, Jason Park, Jay Mahadeokar, Jeet Shah, Jelmer van der Linde, Jennifer Billock, Jenny Hong, Jenya Lee, Jeremy Fu, Jianfeng Chi, Jianyu Huang, Jiawen Liu, Jie Wang, Jiecao Yu, Joanna Bitton, Joe Spisak, Jongsoo Park, Joseph Rocca, Joshua Johnstun, Joshua Saxe, Junteng Jia, Kalyan Vasuden Alwala, Karthik Prasad
+
+Kartikeya Upasani, Kate Plawiak, Ke Li, Kenneth Heafield, Kevin Stone, Khalid El-Arini, Krithika Iyer, Kshitiz Malik, Kuenley Chiu, Kunal Bhalla, Kushal Lakhotia, Lauren Rantala-Yeary, Laurens van der Maaten, Lawrence Chen, Liang Tan, Liz Jenkins, Louis Martin, Lovish Madaan, Lubo Malo, Lukas Blecher, Lukas Landzaat, Luke de Oliveira, Madeline Muzzi, Mahesh Pasupuleti, Mannat Singh, Manohar Paluri, Marcin Kardas, Maria Tsimpoukelli, Mathew Oldham, Mathieu Rita, Maya Pavlova, Melanie Kambadur, Mike Lewis, Min Si, Mitesh Kumar Singh, Mona Hassan, Naman Goyal, Narjes Torabi, Nikolay Bashlykov, Nikolay Bogoychev, Niladri Chatterji, Ning Zhang, Olivier Duchenne, Onur Celebi, Patrick Alrassy, Pengchuan Zhang, Pengwei Li, Petar Vasic, Peter Weng, Prajjwal Bhargava, Pratik Dubal, Praveen Krishnan, Punit Singh Koura, Puxin Xu, Qing He, Qingxiao Dong, Ragavan Srinivasan, Raj Ganapathy, Ramon Calderer, Ricardo Silveira Cabral, Robert Stojnic, Roberta Raileanu, Rohan Maheswari, Rohit Girdhar, Rohit Patel, Romain Sauvestre, Ronnie Polidoro, Roshan Sumbaly, Ross Taylor, Ruan Silva, Rui Hou, Rui Wang, Saghar Hosseini, Sachana Chennabasappa, Sanjay Singh, Sean Bell, Seohyun Sonia Kim, Sergey Edunov, Shaoliang Nie, Sharan Narang, Sharath Raparthy, Sheng Shen, Shengye Wan, Shruti Bhosale, Shun Zhang, Simon Vandenhende, Soumya Batra, Spencer Whitman, Sten Sootla, Stephanie Collot, Suchin Gururangan, Sydney Borodinsky, Tamar Herman, Tara Fowler, Tarek Sheasha, Thomas Georgiou, Thomas Scialom, Tobias Speckbacher, Todor Mihaylov, Tong Xiao, Ujjwal Karn, Vedanuj Goswami, Vibhor Gupta, Vignesh Ramanathan, Viktor Kerkez, Vincent Gonguet, Virginie Do, Vish Vogeti, Vitor Albiero, Vladan Petrovic, Weiwei Chu, Wenhan Xiong, Wenyin Fu, Whitney Meers, Xavier Martinet, Xiaodong Wang, Xiaofang Wang, Xiaqing Ellen Tan, Xide Xia, Xinfeng Xie, Xuchao Jia, Xuewei Wang, Yaelle Goldschlag, Yashesh Gaur, Yasmine Babaei, Yi Wen, Yiwen Song, Yuchen Zhang, Yue Li, Yuning Mao. Zacharie Delpierre Coudert. Zheng Yan. Zhengxing Chen. Zoe Papakipos. Aaditya Singh. Aayushi Srivastava. Abha Jain. Adam Kelsey. Adam Shajnfeld. Adithya Gangidi. Adolfo Victoria. Ahuva Goldstand. Ajay Menon. Ajay Sharma. Alex Boesenberg. Alexei Baevski. Allie Feinstein. Amanda Kallet. Amit Sangani. Amos Teo. Anam Yunus. Andrei Lupu. Andres Alvarado. Andrew Caples. Andrew Gu. Andrew Ho. Andrew Poulton. Andrew Ryan. Ankit Ramchandani. Annie Dong. Annie Franco. Anuj Goyal. Aparajita Saraf. Arkabandhu Chowdhury. Ashley Gabriel. Ashwin Bharambe. Assaf Eisenman. Azadeh Yazdan. Beau James. Ben Maurer. Benjamin Leonhardi. Bernie Huang. Beth Loyd. Beto De Paola. Bhargavi Paranjape. Bing Liu. Bo Wu. Boyu Ni. Braden Hancock. Bram Wasti. Brandon Spence. Brani Stojkovic. Brian Gamido. Britt Montalvo. Carl Parker. Carly Burton. Catalina Mejia. Ce Liu. Changhan Wang. Changkyu Kim. Chao Zhou. Chester Hu. ChingHsiang Chu. Chris Cai. Chris Tindal. Christoph Feichtenhofer. Cynthia Gao. Damon Civin. Dana Beaty. Daniel Kreymer. Daniel Li. David Adkins. David Xu. Davide Testuggine. Delia David. Devi Parikh
+
+Diana Liskovich, Didem Foss, Dingkang Wang, Duc Le, Dustin Holland, Edward Dowling, Eissa Jamil, Elaine Montgomery, Eleonora Presani, Emily Hahn, Emily Wood, Eric-Tuan Le, Erik Brinkman, Esteban Arcaute, Evan Dunbar, Evan Smothers, Fei Sun, Felix Kreuk, Feng Tian, Filippos Kokkinos, First Ozgenel, Francesco Caggioni, Frank Kanayet, Frank Seide, Gabriela Medina Florez, Gabriella Schwarz, Gada Badeer, Georgia Swee, Gil Halpern, Grant Herman, Grigory Sizov, Guangyi, Zhang, Guna Lakshminarayanan, Hakan Inan, Hamid Shojanazeri, Han Zou, Hannah Wang, Hanwen Zha, Haroun Habeeb, Harrison Rudolph, Helen Suk, Henry Aspegren, Hunter Goldman, Hongyuan Zhan, Ibrahim Damlaj, Igor Molybog, Igor Tufanov, Ilias Leontiadis, Irina-Elena Veliche, Itai Gat, Jake Weissman, James Geboski, James Kohli, Janice Lam, Japhet Asher, Jean-Baptiste Gaya, Jeff Marcus, Jeff Tang, Jennifer Chan, Jenny Zhen, Jeremy Reizenstein, Jeremy Teboul, Jessica Zhong, Jian Jin, Jingyi Yang, Joe Cummings, Jon Carvill, Jon Shepard, Jonathan McPhie, Jonathan Torres, Josh Ginsburg, Junjie Wang, Kai Wu, Kam Hou U, Karan Saxena, Kartikay Khandelwal, Katayoun Zand, Kathy Matosich, Kaushik Veeraraghavan, Kelly Michelena, Keqian Li, Kiran Jagadeesh, Kun Huang, Kunal Chawla, Kyle Huang, Lailin Chen, Lakshya Garg, Lavender A. Leandro Silva, Lee Bell, Lei Zhang, Liangpeng Guo, Licheng Yu, Liron Moshkovich, Luca Wehrstedt, Madian Khabsa, Manav Avalani, Manish Bhatt Martynas Mankus, Matan Hasson, Matthew Lennie, Matthias Reso, Maxim Groshev, Maxim Naumov Maya Lathi, Meghan Keneally, Miao Liu, Michael L. Seltzer, Michal Valko, Michelle Restrepo, Mihir Patel Mik Vyatskov,Mikayel Samvelyan Mike Clark Mike Macey, Mike Wang,Miquel Jubert Hermoso Mo Metanat Mohammad Rastegari Munish Bansal Nandhini Santhanam,Natascha Parks,Natasha White,Navyata Bawa,Nayan Singhal,Nick Egebo Nicolas Usunier,Nikhil Mehta,Nikolay Pavlovich Laptev,Ning Dong,Norman Cheng,Oleg Chernoguz Olivia Hart Omkar Salpekar,Ozlem Kalinli Parkin Kent Parth Parekh Paul Saab,Pavan Balaji Pedro Rittner Philip Bontrager Pierre Roux Piotr Dollar Polina Zvyagina Prashant Ratanchandani. British Yuvraj Qian Liang,Rachad AlaoRachel RodriguezRafi AyubRagtoham MurthyRaghu Nayani,Rahul Mitra,Rangaprabhu Parthasarathy Raymond Li Rebekkah Hogan Robin Battey Rocky Wang Russ Howes Rudy Rinott Sachin Mehta Sachin Siby,Sai Jayesh Bondu,Samyak Datta Sara ChughSara Hunt Sargun Dhillon Sasha Sidorov Satadru PanSaurabh MahajanSaurabh Verma Seiji Yamamoto Sharadh Ramaswamy Shaun Lindsay Shaun Lindsay Sheng Feng Shenghao Lin Shengxin Cindy ZhaShishir Patil Shiva Shankar Shuqiang Zhang Shuqiang Zhang Sinong Wang Sneha Agarwal Soji Sajuyigbe Soumith Chintala Stephanie Max Stephen Chen Steve Kehoe Steve Satterfield Sudarshan Govindaprasad Sumit Gupta Summer Deng,Sungmin Cho Sunny Virk Suraj Subramanian Sy Choudhury Sydney Goldman Tal Remez,Tamar Glaser Tamara Best Thilo Koehler Thomas RobinsonTianhe Li Tianjun Zhang Tim
+
+Matthews, Timothy Chou, Tzook Shaked, Varun Vontimitta, Victoria Ajayi, Victoria Montanez, Vijai Mohan, Vinay Satish Kumar, Vishal Mangla, Vlad Ionescu, Vlad Poenaru, Vlad Tiberiu Mihailescu, Vladimir Ivanov, Wei Li, Wenchen Wang, Wenwen Jiang, Wes Bouaziz, Will Constable, Xiaocheng Tang, Xiaojian Wu, Xiaolan Wang, Xilun Wu, Xinbo Gao, Yaniv Kleinman, Yanjun Chen, Ye Hu, Ye Jia, Ye Qi, Yenda Li, Yilin Zhang, Ying Zhang, Yossi Adi, Youngjin Nam, Yu, Wang, Yu Zhao, Yuchen Hao, Yundi Qian, Yunlu Li, Yuzi He, Zach Rait, Zachary DeVito, Zef Rosnbrick, Zhaoduo Wen, Zhenyu Yang, Zhiwei Zhao, and Zhiyu Ma. 2024. The Llama 3 herd of models. Preprint, arXiv:2407.21783.
+
+Brian Hughes, Cynthia Miller-Idriss, Rachael Pitch-Loeb, Beth Goldberg, Kesa White, Meili Criezis, and Elena Savoia. 2021. Development of a codebook of online anti-vaccination rhetoric to manage COVID-19 vaccine misinformation. International Journal of Environmental Research and Public Health, 18(14).
+
+Jan-Christoph Klie, Michael Bugert, Beto Boullosa, Richard Eckart de Castilho, and Iryna Gurevych. 2018. The inception platform: Machine-assisted and knowledge-oriented interactive annotation. In Proceedings of the 27th International Conference on Computational Linguistics: System Demonstrations, pages 5-9. Association for Computational Linguistics. Event Title: The 27th International Conference on Computational Linguistics (COLING 2018).
+
+Bonka Kotseva, Irene Vianini, Nikolaos Nikolaidis, Nicolò Faggiani, Kristina Potapova, Caroline Gasparro, Yaniv Steiner, Jessica Scornavacche, Guillaume Jacquet, Vlad Dragu, Leonida Della Rocca, Stefano Bucci, Aldo Podavini, Marco Verile, Charles Macmillan, and Jens P. Linge. 2023. Trend analysis of COVID-19 mis/disinformation narratives-A 3-year study. PLOS ONE, 18(11):e0291423.
+
+Yue Li, Carolina Scarton, Xingyi Song, and Kalina Bontcheva. 2023. Classifying COVID-19 vaccine narratives. In Proceedings of the 14th International Conference on Recent Advances in Natural Language Processing, pages 648-657, Varna, Bulgaria. INCOMA Ltd., Shoumen, Bulgaria.
+
+Bing Liu and Lei Zhang. 2012. A Survey of Opinion Mining and Sentiment Analysis, pages 415-463. Springer US, Boston, MA.
+
+Nikolaos Nikolaidis, Jakub Piskorski, and Nicolas Stefanovitch. 2024. Exploring the usability of persuasion techniques for downstream misinformation-related classification tasks. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 6992-7006, Torino, Italia. ELRA and ICCL.
+
+Keith Norambuena, Brian Felipe, Tanushree Mitra, and Chris North. 2023. A survey on event-based news narrative extraction. ACM Comput. Surv., 55(14s).
+
+Sérgio Nunes, Alípio Mario Jorge, Evelin Amorim, Hugo Sousa, António Leal, Purificação Moura Silvano, Inês Cantante, and Ricardo Campos. 2024. Text2Story lusa: A dataset for narrative analysis in European Portuguese news articles. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 15773-15782, Torino, Italia. ELRA and ICCL.
+
+Jakub Piskorski, Alípio Jorge, Maria da Purificação Silvano, Nuno Guimarães, Ana Filipa Pacheco, and Nana Yu. 2024. Overview of the CLEF-2024 CheckThat! Lab task 3 on persuasion techniques.
+
+Jakub Piskorski, Nicolas Stefanovitch, Giovanni Da San Martino, and Preslav Nakov. 2023a. SemEval-2023 task 3: Detecting the category, the framing, and the persuasion techniques in online news in a multi-lingual setup. In Proceedings of the 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 2343-2361, Toronto, Canada. Association for Computational Linguistics.
+
+Jakub Piskorski, Nicolas Stefanovitch, Nikolaos Nikolaidis, Giovanni Da San Martino, and Preslav Nakov. 2023b. Multilingual multifaceted understanding of online news in terms of genre, framing, and persuasion techniques. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3001-3022, Toronto, Canada. Association for Computational Linguistics.
+
+Brenda Santana, Ricardo Campos, Evelin Amorim, Alpio Jorge, Purificacao Silvano, and Sérgio Nunes. 2023. A survey on narrative extraction from textual data. Artif. Intell. Rev., 56(8):8393-8435.
+
+P. Silvano, E. Amorim, A. Leal, I. Cantante, A. Jorge, R. Campos, and N. Yu. 2024. Untangling a web of temporal relations in news articles. In Proceedings of the 7th International Workshop on Narrative Extraction from Texts (Text2Story'24). Glasgow. Scotland. March 24. pp. 77-92.
+
+Ivan Vykopal, Matúš Pikuliak, Ivan Srba, Robert Moro, Dominik Macko, and Maria Bielikova. 2024. Disinformation Capabilities of Large Language Models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 14830-14847, Bangkok, Thailand. Association for Computational Linguistics.
+
+Manuel Weigand, Maximilian Weber, and Johannes Gruber. 2022. *Conspiracy Narratives in the Protest Movement Against COVID-19 Restrictions in Germany: A Long-term Content Analysis of Telegram Chat Groups*. In *Proceedings of the Fifth Workshop on Natural Language Processing and Computational Social Science (NLP+CSS)*, pages 52–58, Abu Dhabi, UAE. Association for Computational Linguistics.
+
+Rongwu Xu, Brian Lin, Shujian Yang, Tianqi Zhang, Weiyan Shi, Tianwei Zhang, Zhixuan Fang, Wei Xu,
+
+and Han Qiu. 2024. The earth is flat because...: Investigating LLMs' belief towards misinformation via persuasive conversation. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 16259-16303, Bangkok, Thailand. Association for Computational Linguistics.
+
+Wajdi Zaghouani, Mustafa Jarrar, Nizar Habash, Houda Bouamor, Imed Zitouni, Mona Diab, Samhaa El-Beltagy, and Muhammed AbuOdeh. 2024. The FIGNEWS Shared Task on News Media Narratives. In Proceedings of The Second Arabic Natural Language Processing Conference, pages 530-547, Bangkok, Thailand. Association for Computational Linguistics.
+
+# A Taxonomies
+
+This section introduces the taxonomies of narratives and subnarratives for the two domains.
+
+# A.1 Ukraine-Russia War (URW)
+
+1. Blaming the war on others rather than the invader: statements attributing responsibility or fault to entities other than Russia in the context of Russia's invasion of Ukraine.
+
+Example: "The economic crisis is due to Western sanctions."
+
+Example: "Ukraine's actions provoked this conflict."
+
+# Sub-Narratives:
+
+(a) Ukraine is the aggressor: Statements that shift the responsibility of the aggression to Ukraine instead of Russia and portray Ukraine as the attacker.
+
+Example: "Ukraine secretly provoked the war because it was harassing Don-bass province citizens."
+
+(b) The West are the aggressors: Statements that shift the responsibility for the conflict and escalation to the Western block.
+
+Example: "The real perpetrators were US/EU. They sabotaged Minsk II agreement only to force Russia to invade."
+
+2. Discrediting Ukraine: Statements that undermine the legitimacy, actions, or intentions of Ukraine or Ukrainians as a nation.
+
+Example: “Ukraine’s government is corrupt and ineffective.”
+
+Example: "Ukrainian soldiers are committing atrocities."
+
+Example: "Ukrainian identity does not exist"
+
+# Sub-Narratives:
+
+(a) Rewriting Ukraine's history: Statements that aim to reestablish history of Ukrainian nation in a way that discredits its reputation.
+
+Example: "Ukraine is not a real nation, it was a fabrication to split Russia and ally with Hitler."
+
+(b) Discrediting Ukrainian nation and society: Statements that aggressively undermine the legitimacy and reputability of Ukrainian ethnicity and people
+
+(c) Discrediting Ukrainian military: Statements that aim to undermine the capabilities, professionalism or effectiveness of the Ukrainian armed forces.  
+(d) Discrediting Ukrainian government and officials and policies: Statements that seek to delegitimize the Ukrainian government, its leaders, and its policies, portraying them as corrupt or incompetent.  
+(e) Ukraine is a puppet of the West: Claims that Ukraine is controlled or heavily influenced by Western powers, particularly the United States and European Union.  
+(f) Ukraine is a hub for criminal activities: Allegations that Ukraine is a center for illegal activities such as human trafficking, drug smuggling, or organized crime  
+(g) Ukraine is associated with Nazism: Accusations that Ukrainian society or government has ties to or sympathies with Nazi ideology, often referencing historical events or extremist groups.  
+(h) Situation in Ukraine is hopeless: Statements that portray Ukraine as having no viable perspectives or no potential positive future. Example: "Ukraine should just give up, it is all over debt and will be exploited by the West anyway."
+
+3. Russia is the Victim: Statements that portray Russia as being unfairly targeted or victimized.
+
+Example: "Russia is being unfairly sanctioned."
+
+Example: "The West is ganging up on Russia without justification."
+
+Example: "Russia is doing what every country would do (e.g. protect its interests/honour etc.)"
+
+# Sub-Narratives:
+
+(a) The West is russophobic: Statements that claim that the negative reaction to Russia's actions are because of the negative perspective of western countries instead of Russia's own actions.
+
+Example: "Politicians in the West blame Russia for everything, instead of looking
+
+at their mistakes."
+
+Example: "In Country X, they banned Tchaikovsky ballets and Chechov's plays because they cannot stand Russia and its culture."
+
+(b) Russia actions in Ukraine are only self-defence: Statements that justify Russia's action solely as legitimate self-defence and not a deliberate action.
+
+Example: “There was no other way than war to defend the Russian-speaking people in Donbass.”
+
+(c) UA is anti-RU extremists: Statements claiming that Ukraine is comprised of extremist elements that are vehemently opposed to Russia.
+
+4. Praise of Russia: Statements that positively highlight Russia's actions, policies, or character
+
+Example: "Russia is leading the way in international diplomacy."
+
+Example: "The Russian economy is resilient and strong."
+
+Example: “Glorifying mentions of Russia’s weapon systems and military might.”
+
+# Sub-Narratives:
+
+(a) Praise of Russian military might: Statements that positively highlight Russia's military institutions, equipment and scale.
+
+Example: “Russia has far more tanks and powerful artillery that US/EU would only dream of.”
+
+(b) Praise of Russian President Vladimir Putin: Statements that present Vladimir Putin positively, including his personal and leadership qualities.
+
+Example: "Any country would want such a strong leader as Putin to lead the way."
+
+(c) Russia is a guarantor of peace and prosperity: Statements that portray Russia solely in a positive manner, emphasising their potential to provide peace and prosperity to those that cooperate.
+
+Example: "Take a look at Africa, Russia supports countries and turns them into independent nations guided by their people's interests where Western countries colonised brutally."
+
+(d) Russia has international support from a number of countries and people: Statements that emphasise the popularity and acceptance of Russia in the international stage.
+
+Example: "The majority of the countries population sides with Russia as per last UN General Assembly vote."
+
+(e) Russian invasion has strong national support: Statements that emphasise the popularity and acceptance of the invasion inside Russia and on Russian-speaking populations.
+
+Example: "The majority of the countries population sides with Russia as per last UN General Assembly vote."
+
+5. Overpraising the West: Statements that excessively and unduly laud or extol the virtues, accomplishments, and moral superiority of Western countries, particularly in the context of international relations and military.
+
+# Sub-Narratives:
+
+(a) NATO will destroy Russia: Statements that suggest or claim that the North Atlantic Treaty Organization (NATO) and its allies are capable or already in the process of eradicating Russia.  
+(b) The West belongs in the right side of history: Statements that portray Western nations and their actions as morally superior and aligned with progress and justice and possess moral superiority.  
+(c) The West has the strongest international support: Statements that emphasize or claim widespread backing for Western policies and actions from the international community, potentially downplaying opposition or criticism.
+
+6. Speculating war outcomes: Statements that predict or make assumptions about the potential results or consequences of a conflict
+
+# Sub-Narratives:
+
+(a) Russian army is collapsing: Statements that suggest or claim that the Russian military is experiencing a significant decline in its effectiveness, strength, or morale.
+
+(b) Russian army will lose all the occupied territories: Speculative statements that predict or assume the potential outcomes of the conflict, specifically regarding the possibility of the Russian military losing control of all the territories it currently occupies.  
+(c) Ukrainian army is collapsing: Statements that suggest or claim that the Ukrainian military is experiencing a significant decline in its effectiveness, strength, or morale.
+
+7. Discrediting the West, Diplomacy: Statements that criticize the Western countries, or international diplomatic efforts.
+
+Example: "The West is hypocritical in its foreign policy."
+
+Example: “Western diplomacy has failed in resolving conflicts.”
+
+Example: "International organizations will not solve anything because..."
+
+# Sub-Narratives:
+
+(a) The EU is divided: Statements that present the EU as a set of divided entities and interests, usually unable to take actions.  
+Example: "The European Council will never vote on sanctions for Russia, since they cannot agree on even the simplest of the issues."  
+(b) The West is weak: Statements presenting the West overall as a non-potent group of countries (that is not as powerful as it used to be).  
+Example: "The weakened West is once again impotent to act in front of the will Russia."  
+(c) The West is overreacting: Statements that claim that the West and its institutions are reacting to Russia's actions in a disproportionate manner.  
+Example: “Putin did not invade the EU but Ukraine. Imposing harsh sanctions is not the way to deal with it, dialogue and debate is.”  
+(d) The West does not care about Ukraine, only about its interests: Statements that claim that the West is only interested in Ukraine for its own benefits, disregard
+
+ing the country's fate.
+
+Example: "The West has indebted Ukraine more than XX bn of dollars, a lucrative deal for western companies to exploit."
+
+Example: "NATO's actions are endangering global security."
+
+(e) Diplomacy does/will not work: Statements discrediting the potential of ongoing or potential diplomatic efforts.
+
+Example: "Diplomats are desperately trying to figure out solutions but now it's too late, they have failed and Russia is free to do whatever."
+
+(f) West is tired of Ukraine: Claims that Western countries, particularly the United States and European nations, are becoming fatigued or disinterested in supporting Ukraine and its efforts.
+
+8. Negative Consequences for the West: Statements that highlight or predict adverse outcomes for Western countries and their interests.
+
+Example: "Sanctions against Russia will backfire on Europe."
+
+Example: "The West is headed for an economic downturn."
+
+# Sub-Narratives:
+
+(a) Sanctions imposed by Western countries will backfire: Statements that catastrophize on the possible negative effects for Western sanctions of Russia.
+
+Example: "The winter is going to be cold and with current gas prices, we are talking of societal unrest."
+
+(b) The conflict will increase the Ukrainian refugee flows to Europe: Statements that catastrophize on the possible refugee outflows due to the conflict.
+
+Example: "Like we did not have refugees from the Middle East, now we will have Ukrainians stressing our housing and healthcare problems."
+
+9. Distrust towards Media: Statements that question the reliability or integrity of media organizations.
+
+Example: "Western media is spreading propaganda."
+
+Example: "You can't trust what the news says about Russia."
+
+# Sub-Narratives:
+
+(a) Western media is an instrument of propaganda: Statements that discredit the media institutions of the West and claim that they are instruments of propaganda. Example: "... but you wouldn't hear this on a western channel, only the party line from State Department."  
+(b) Ukrainian media cannot be trusted: Statements that discredit the media institutions of the Ukraine and claim that they should not be trusted for reporting on the war.
+
+Example: "Ukraine is conducting its own propaganda using their TV channels, news and social media."
+
+10. Amplifying war-related fears: Statements that evoke fear or anxiety about potential threats, dangers or reactions.
+
+Example: "The West is pushing us towards World War III."
+
+Example: "It is a matter of time before war speeds on the West"
+
+Example: "Nuclear war is imminent"
+
+# Sub-Narratives:
+
+(a) By continuing the war we risk WWII: Statements that warn against upsetting Russia's and its leadership, evoking fear of causing WW3.
+
+Example: "The Western elites with their fixation on Russia are sleepwalking towards WW3"
+
+(b) Russia will also attack other countries: Statements that claim that it is imminent that Russia will attack other countries.
+
+Example: "... and be sure, Ukraine is the first not the last country to be invaded. Others will follow."
+
+(c) There is a real possibility that nuclear weapons will be employed: Statements that evoke fear or anxiety about the use of nuclear weapons.
+
+Example: "... and if Western hypocrisy continues to provoke, Putin might be forced to press the red button... for good"
+
+(d) NATO should/will directly intervene: Statements that suggest or claim that the North Atlantic Treaty Organization (NATO) ought to or will take direct military action in a conflict, potentially implying a shift in policy or strategy.
+
+11. Hidden plots by secret schemes of powerful groups: Statements that suggest hidden plots or secretive actions by powerful groups related to the war.
+
+Example: "There's a secret plan by the elites to control global resources."
+
+Example: "The war is just a cover for something much bigger."
+
+# A.2 Climate Change (CC)
+
+1. Criticism of climate policies: Statements that question the effectiveness, economic impact, or motives behind climate policies.
+
+# Sub-Narratives:
+
+(a) Climate policies are ineffective: Statements suggesting that climate policies fail to achieve their intended environmental goals.  
+(b) Climate policies have negative impact on the economy: Statements claiming that climate policies lead to negative economic outcomes.  
+(c) Climate policies are only for profit: Statements that argue climate policies are driven by financial or corporate gain rather than genuine environmental concerns.
+
+2. Criticism of institutions and authorities: Statements that challenge the competence, integrity, or intentions of various institutions and authorities in relation to climate change.
+
+# Sub-Narratives:
+
+(a) Criticism of the EU: Statements that express disapproval or distrust of the EU's role or approach to climate change or the EU in general.  
+(b) Criticism of international entities: Statements that criticize the role and influence of international entities on climate policy.  
+(c) Criticism of national governments: Statements that disapprove of the ways
+
+national governments handle climate change.
+
+(d) Criticism of political organizations and figures: Statements that discredit political organizations and figures in the context of climate change debate.
+
+3. Climate change is beneficial: Statements that present arguments that support that changes in climate can have positive effects as well.
+
+# Sub-Narratives:
+
+(a) CO2 is beneficial: Statements suggesting that increased CO2 levels have positive impacts on the environment.  
+(b) Temperature increase is beneficial: Statements claiming that rising global temperatures can have positive effects.
+
+4. Downplaying climate change: Statements that minimize the significance or impact of climate change.
+
+# Sub-Narratives:
+
+(a) Climate cycles are natural: Statements suggesting that climate change is a natural and cyclical occurrence.  
+(b) Weather suggests the trend is global cooling: Statements using local or short-term weather patterns to argue against global warming.  
+(c) Temperature increase does not have significant impact: Statements claiming that the increase in temperature is not going to have any noticeable effect in nature.  
+(d) CO2 concentrations are too small to have an impact: Statements claiming that the concentrations of CO2 will have a negligible effect.  
+(e) Human activities do not impact climate change: Statements that support that climate change is not caused by human activity.  
+(f) Ice is not melting: Statements claiming that there is not melting of ice.  
+(g) Sea levels are not rising: Statements denying that sea levels have risen (or will rise).
+
+(h) Humans and nature will adapt to the changes: Statements claiming that whatever the changes in climate humans or nature will manage to find solutions to adapt.
+
+# 5. Questioning the measurements and science:
+
+Statements that raise doubts about the scientific methods, data, and consensus on climate change.
+
+# Sub-Narratives:
+
+(a) Methodologies/metrics used are unreliable/faulty: Statements claiming that the scientific methodologies and metrics used to measure climate change are flawed or unreliable.  
+(b) Data shows no temperature increase: Statements asserting that available data does not support the claim of global temperature increase.  
+(c) Greenhouse effect/carbon dioxide do not drive climate change: Statements asserting that available data does not support the claim of global temperature increase.  
+(d) Scientific community is unreliable: Statements discrediting scientists, the scientific community and their actions.
+
+# 6. Criticism of climate movement: Statements that challenge the motives, integrity, or impact of the climate movement.
+
+# Sub-Narratives:
+
+(a) Climate movement is alarmist: Statements suggesting that the climate movement exaggerates the severity of climate change for dramatic effect.  
+(b) Climate movement is corrupt: Statements alleging that the climate movement is influenced by ulterior motives, by corruption or by unethical practices.  
+(c) Ad hominem attacks on key activists: Statements attacking the reputation of key figures (such as scientists, activists, politicians or public figures).
+
+# 7. Controversy about green technologies:
+
+Statements that express skepticism or criticism of environmentally friendly technologies.
+
+# Sub-Narratives:
+
+(a) Renewable energy is dangerous: Statements claiming that renewable energy sources pose significant risks or dangers.  
+(b) Renewable energy is unreliable: Statements asserting that renewable energy sources are not dependable for widespread adoption.  
+(c) Renewable energy is costly: Statements asserting that renewable energy sources are too expensive, inefficient and worth adopting for widespread use.  
+(d) Nuclear energy is not climate friendly: Statements asserting that nuclear sources are or should not be considered as good for the climate.
+
+# 8. Hidden plots by secret schemes of powerful groups: Statements that propose secret plots or hidden agendas related to climate change initiated by powerful entities or groups.
+
+# Sub-Narratives:
+
+(a) Blaming global elites: Statements attributing climate change agendas to secretive and powerful global elites.  
+(b) Climate agenda has hidden motives: Claims that the push for climate action is driven by ulterior motives, such as political power or population control.
+
+# 9. Amplifying Climate Fears: Statements that emphasize and amplify fears about the consequences of climate change.
+
+# Sub-Narratives:
+
+(a) Earth will be uninhabitable soon: Statements predicting that the Earth will become uninhabitable in the near future due to climate change.  
+(b) Amplifying existing fears of global warming: Statements that are using fears related to possible climate worries to spread panic.  
+(c) Doomsday scenarios for humans: Statements presenting intense catastrophic scenarios as results of climate change.  
+(d) Whatever we do it is already too late: Statements that minimize the urgency of addressing climate change by suggesting that any action taken at this point is futile or too late to make a meaningful impact.
+
+![](images/0d05f2aab1cd240cd80f8471e8cb672d8e23a21861cc10db8886d2d699dabedf.jpg)  
+Figure 4: URW: Normalized count of coarse labels per language
+
+10. Green policies are geopolitical instruments: Statements claiming that that environmental policies and initiatives are used as tools for geopolitical power and influence rather than genuine environmental concern.
+
+# Sub-Narratives:
+
+(a) Climate-related international relations are abusive/exploitative: Statements criticizing international relations related to climate change as exploitative or economically abusive.  
+(b) Green activities are a form of neocolonialism: Statements suggesting that green initiatives are a way for developed countries to exert control and influence over developing nations, a modern form of colonial practices.
+
+# B Dataset statistics
+
+Figures 4 and 5 report the normalized count of coarse labels per language. There are substantial differences, notably between RU and other languages for URW and between EN and other languages for CC. These differences could be due either to bias in media of each country, or be due to bias in the sampling of the articles to annotate. The reason RU is absent for CC is that it was not possible to find enough articles to annotate.
+
+In Table 5, we report on the difficulty of annotating labels. We report the proportion of each label to have a disagreement between annotators both with the Other label specifically, or with any other label. Based on the numbers, we categorise each label in a class of difficulty.
+
+# C Prompts used for LLMs
+
+In this Section we provide the different prompts used in the experiments.
+
+![](images/412eac7042cf042a281350478bd5059b2cac23e433776a58feaea143c9991e96.jpg)  
+Figure 5: CC: Normalized count of coarse labels per language
+
+# C.1 0shot (only taxonomy)
+
+You are an experienced analyst making labeling articles with labels specific to the war in Ukraine.
+
+Labels:  
+{codebook(urw}
+
+You will apply a label to each paragraph if it presents at any point a narrative that is given in the above list.
+
+The beginning of each paragraph is marked by a [paragraph_X:] tag.
+
+Answer strictly only with the paragraph tag and the name of the labels detected and not anything else. If you find no label, just output 'Other'. Strip down the markdown artifacts. Join the labels with semicolon.
+
+Do not output anything from the text.
+
+# C.2 Guidelines
+
+You are an experienced analyst labeling articles with labels specific ["to the war in Ukraine"/ "Climate Change"] You will apply a label to each paragraph if it presents at any point a narrative that is given in the annotation guidelines given below.
+
+{guidelines(urw}
+
+The beginning of each paragraph is marked by a [paragraph_X:] tag.
+
+Answer strictly only with the paragraph tag and the name of the labels detected and not anything else. If you find no label, just output 'Other'. Strip down the markdown artifacts. Join the labels with semicolon.
+
+Do not output anything from the text.
+
+# C.3 Taxonomy and guidelines
+
+You are an experienced analyst making labeling articles with labels specific to the war in Ukraine.
+
+Labels:
+
+{codebook_url}
+
+You will apply a label to each paragraph if it presents at any point a narrative that is given in the annotation guidelines given below.
+
+{guidelines(urw}
+
+The beginning of each paragraph is marked by a [paragraph_X:] tag.
+
+Answer strictly only with the paragraph tag and the name of the labels detected and not anything else. If you find no label, just output 'Other'. Strip down the markdown artifacts. Join the labels with semicolon.
+
+Do not output anything from the text.
+
+Answer strictly only with the paragraph tag and the name of the labels detected and not anything else. If you find no label, just output 'Other'. Strip down the markdown artifacts. Join the labels with semicolon.
+
+Do not output anything from the text.
+
+# Second prompt:
+
+You are an experienced analyst making labeling articles with labels specific to the war in Ukraine.
+
+Labels:
+
+{codebook_arw}
+
+You are given the coarse label and you are asked to return the fine label.
+
+You will apply a label if the article presents at any point a narrative that is given in the annotation guidelines given below.
+
+{guidelines(urw}
+
+The beginning of each paragraph is marked by a [paragraph_X:] tag.
+
+Answer strictly only with the paragraph tag and the name of the labels detected and not anything else. If you find no label, just output 'Other'. Strip down the markdown artifacts. Join the labels with semicolon.
+
+Do not output anything from the text.
+
+# C.4 Hierarchical prompting
+
+# First prompt:
+
+You are an experienced analyst making labeling articles with labels specific to the war in Ukraine.
+
+Labels:
+
+{codebook(urw_coarse}
+
+The beginning of each paragraph is marked by a [paragraph_X:] tag.
+
+![](images/754ef1cd5187ed25b468c922821c207a12525bfbfbba2ad5ecf02d169ff839a9.jpg)  
+Figure 6: Performance of XLM-RoBERTa_fine on each sub-Narrative (fine) of the URW domain. Only labels with  $>2$  support are shown
+
+# D Selected model performance per label and language
+
+We provide detailed breakdown of the performance of our top-performing model on fine-grained level (sub-Narratives). We provide both the detailed scores per-label (Figures 6, 7) and per-language (Figures 8, 9, 10, 11) for the two domains (CC and URW).
+
+![](images/109a15a7af15ffd981c4d09b4aafa35c347638ea0d1f99e66984c3adf566ad76.jpg)  
+Figure 7: Performance of XLM-RoBERTa_fine on each sub-Narrative (fine) of the CC domain. Only labels with  $>2$  support are shown
+
+# E Breakdown on Annotator disagreements
+
+We report on the detailed dissatisfaction statistics of the annotators per each fine-grained label, and highlight the difficulty of each label, in Table 5.
+
+![](images/07db5099509e9feea006f54d2c0fb36d908fb9397ea2c36ed112c47b869d9f3f.jpg)
+
+![](images/c9dc3e0ae7e4c3c92501cd9307f7eca6d59aa7b19260e6be65fb175662285bd6.jpg)  
+Figure 8: F1 scores of XLM-RoBERTa_fine on each language of the URW domain.
+
+![](images/c33450a5c6c77fa683c2dec5861c3badeda8860c01001cf3993133fab6aa3c79.jpg)
+
+![](images/fe6a33e8516c4bb4b735dfc7348d41eaedbfc60055839784a5967cc6682e849c.jpg)  
+Figure 10: F1 scores of Llama-3.1-70b on each language of the URW domain.  
+Figure 9: F1 scores of XLM-RoBERTa_fine on each language of the CC domain.  
+Figure 11: F1 scores of Llama-3.1-70b on each language of the CC domain.
+
+<table><tr><td>label</td><td>count</td><td>%conf</td><td>%other</td><td>mcl</td><td>difficulty</td></tr><tr><td>1: URW: Blaming the war on others rather than the invader: Ukraine is the aggressor</td><td>169</td><td>0.309</td><td>0.315</td><td>5</td><td>Medium</td></tr><tr><td>2: URW: Blaming the war on others rather than the invader: The West are the aggressors</td><td>260</td><td>0.377</td><td>0.390</td><td>7</td><td>Medium</td></tr><tr><td>3: URW: Discrediting Ukraine: Rewriting Ukraine&#x27;s history</td><td>23</td><td>0.272</td><td>0.111</td><td>28</td><td>Medium *</td></tr><tr><td>4: URW: Discrediting Ukraine: Discrediting Ukrainian nation and society</td><td>38</td><td>0.380</td><td>0.053</td><td>6</td><td>Medium *</td></tr><tr><td>5: URW: Discrediting Ukraine: Discrediting Ukrainian military</td><td>316</td><td>0.314</td><td>0.217</td><td>1</td><td>Medium *</td></tr><tr><td>6: URW: Discrediting Ukraine: Discrediting Ukrainian government and officials and policies</td><td>502</td><td>0.215</td><td>0.253</td><td>8</td><td>Easy</td></tr><tr><td>7: URW: Discrediting Ukraine: Ukraine is a puppet of the West</td><td>219</td><td>0.235</td><td>0.165</td><td>28</td><td>Easy *</td></tr><tr><td>8: URW: Discrediting Ukraine: Ukraine is a hub for criminal activities</td><td>127</td><td>0.368</td><td>0.243</td><td>6</td><td>Medium *</td></tr><tr><td>9: URW: Discrediting Ukraine: Ukraine is associated with Nazism</td><td>97</td><td>0.189</td><td>0.080</td><td>6</td><td>Easy *</td></tr><tr><td>10: URW: Discrediting Ukraine: Situation in Ukraine is hopeless</td><td>107</td><td>0.303</td><td>0.220</td><td>6</td><td>Medium *</td></tr><tr><td>11: URW: Russia is the Victim: The West is Russophobic</td><td>167</td><td>0.343</td><td>0.427</td><td>2</td><td>Medium</td></tr><tr><td>12: URW: Russia is the Victim: Russia actions in Ukraine are only self-defence</td><td>130</td><td>0.518</td><td>0.273</td><td>1</td><td>Hard *</td></tr><tr><td>13: URW: Russia is the Victim: UA is anti-RU extremists</td><td>34</td><td>0.477</td><td>0.406</td><td>11</td><td>Hard *</td></tr><tr><td>14: URW: Praise of Russia: Praise of Russian military might</td><td>466</td><td>0.140</td><td>0.346</td><td>5</td><td>Easy</td></tr><tr><td>15: URW: Praise of Russia: Praise of Russian President Vladimir Putin</td><td>100</td><td>0.157</td><td>0.348</td><td>17</td><td>Easy</td></tr><tr><td>16: URW: Praise of Russia: Russia is a guarantor of peace and prosperity</td><td>257</td><td>0.352</td><td>0.385</td><td>1</td><td>Medium</td></tr><tr><td>17: URW: Praise of Russia: Russia has international support from a number of countries and people</td><td>228</td><td>0.160</td><td>0.245</td><td>16</td><td>Easy</td></tr><tr><td>18: URW: Praise of Russia: Russian invasion has strong national support</td><td>20</td><td>0.333</td><td>0.250</td><td>17</td><td>Medium *</td></tr><tr><td>19: URW: Overpraising the West: NATO will destroy Russia</td><td>11</td><td>0.263</td><td>0.400</td><td>20</td><td>Medium</td></tr><tr><td>20: URW: Overpraising the West: The West belongs in the right side of history</td><td>36</td><td>0.263</td><td>0.133</td><td>28</td><td>Medium *</td></tr><tr><td>21: URW: Overpraising the West: The West has the strongest international support</td><td>24</td><td>0.254</td><td>0.308</td><td>7</td><td>Medium</td></tr><tr><td>22: URW: Speculating war outcomes: Russian army is collapsing</td><td>51</td><td>0.313</td><td>0.476</td><td>23</td><td>Medium</td></tr><tr><td>23: URW: Speculating war outcomes: Russian army will lose all the occupied territories</td><td>10</td><td>0.312</td><td>0.200</td><td>22</td><td>Medium *</td></tr><tr><td>24: URW: Speculating war outcomes: Ukrainian army is collapsing</td><td>93</td><td>0.242</td><td>0.118</td><td>5</td><td>Easy *</td></tr><tr><td>25: URW: Discrediting the West, Diplomacy: The EU is divided</td><td>105</td><td>0.164</td><td>0.435</td><td>28</td><td>Easy</td></tr><tr><td>26: URW: Discrediting the West, Diplomacy: The West is weak</td><td>145</td><td>0.301</td><td>0.267</td><td>14</td><td>Medium *</td></tr><tr><td>27: URW: Discrediting the West, Diplomacy: The West is overreacting</td><td>34</td><td>0.395</td><td>0.294</td><td>11</td><td>Medium *</td></tr><tr><td>28: URW: Discrediting the West, Diplomacy: The West does not care about Ukraine, only about its interests</td><td>211</td><td>0.381</td><td>0.262</td><td>7</td><td>Medium *</td></tr><tr><td>29: URW: Discrediting the West, Diplomacy: Diplomacy does/will not work</td><td>109</td><td>0.550</td><td>0.434</td><td>28</td><td>Hard *</td></tr><tr><td>30: URW: Discrediting the West, Diplomacy: West is tired of Ukraine</td><td>43</td><td>0.583</td><td>0.171</td><td>28</td><td>Hard *</td></tr><tr><td>31: URW: Negative Consequences for the West: Sanctions imposed by Western countries will backfire</td><td>79</td><td>0.338</td><td>0.723</td><td>12</td><td>Medium</td></tr><tr><td>32: URW: Negative Consequences for the West: The conflict will increase the Ukrainian refugee flows to Europe</td><td>12</td><td>0.000</td><td>NaN</td><td></td><td>Easiest</td></tr><tr><td>33: URW: Distrust towards Media: Western media is an instrument of propaganda</td><td>104</td><td>0.180</td><td>0.452</td><td>11</td><td>Easy</td></tr><tr><td>34: URW: Distrust towards Media: Ukrainian media cannot be trusted</td><td>23</td><td>0.500</td><td>NaN</td><td>6</td><td>Hard</td></tr><tr><td>35: URW: Amplifying war-related fears: By continuing the war we risk WWII</td><td>135</td><td>0.285</td><td>0.339</td><td>37</td><td>Medium</td></tr><tr><td>36: URW: Amplifying war-related fears: Russia will also attack other countries</td><td>133</td><td>0.211</td><td>0.270</td><td>2</td><td>Easy</td></tr><tr><td>37: URW: Amplifying war-related fears: There is a real possibility that nuclear weapons will be employed</td><td>292</td><td>0.259</td><td>0.486</td><td>2</td><td>Medium</td></tr><tr><td>38: URW: Amplifying war-related fears: NATO should/will directly intervene</td><td>63</td><td>0.305</td><td>0.273</td><td>36</td><td>Medium *</td></tr><tr><td>39: CC: Criticism of climate policies: Climate policies are ineffective</td><td>74</td><td>0.407</td><td>0.174</td><td>44</td><td>Hard *</td></tr><tr><td>40: CC: Criticism of climate policies: Climate policies have negative impact on the economy</td><td>96</td><td>0.337</td><td>0.333</td><td>45</td><td>Medium *</td></tr><tr><td>41: CC: Criticism of climate policies: Climate policies are only for profit</td><td>58</td><td>0.508</td><td>0.069</td><td>68</td><td>Hard *</td></tr><tr><td>42: CC: Criticism of institutions and authorities: Criticism of the EU</td><td>54</td><td>0.275</td><td>NaN</td><td>45</td><td>Medium</td></tr><tr><td>43: CC: Criticism of institutions and authorities: Criticism of international entities</td><td>104</td><td>0.434</td><td>0.196</td><td>44</td><td>Hard *</td></tr><tr><td>44: CC: Criticism of institutions and authorities: Criticism of national governments</td><td>225</td><td>0.375</td><td>0.274</td><td>45</td><td>Medium *</td></tr><tr><td>45: CC: Criticism of institutions and authorities: Criticism of political organizations and figures</td><td>190</td><td>0.487</td><td>0.172</td><td>44</td><td>Hard *</td></tr><tr><td>46: CC: Climate change is beneficial: CO2 is beneficial</td><td>19</td><td>0.071</td><td>NaN</td><td>51</td><td>Easiest</td></tr><tr><td>47: CC: Climate change is beneficial: Temperature increase is beneficial</td><td>13</td><td>0.266</td><td>0.500</td><td>54</td><td>Medium</td></tr><tr><td>48: CC: Downplaying climate change: Climate cycles are natural</td><td>36</td><td>0.294</td><td>0.300</td><td>50</td><td>Medium</td></tr><tr><td>49: CC: Downplaying climate change: Weather suggests the trend is global cooling</td><td>13</td><td>0.406</td><td>0.308</td><td>70</td><td>Hard *</td></tr><tr><td>50: CC: Downplaying climate change: Temperature increase does not have significant impact</td><td>5</td><td>0.736</td><td>0.286</td><td>56</td><td>Hard *</td></tr><tr><td>51: CC: Downplaying climate change: CO2 concentrations are too small to have an impact</td><td>18</td><td>0.500</td><td>0.143</td><td>59</td><td>Hard *</td></tr><tr><td>52: CC: Downplaying climate change: Human activities do not impact climate change</td><td>34</td><td>0.387</td><td>0.083</td><td>60</td><td>Medium *</td></tr><tr><td>53: CC: Downplaying climate change: Ice is not melting</td><td>18</td><td>0.117</td><td>NaN</td><td>62</td><td>Easy</td></tr><tr><td>54: CC: Downplaying climate change: Sea levels are not rising</td><td>2</td><td>0.500</td><td>0.500</td><td>47</td><td>Hard</td></tr><tr><td>55: CC: Downplaying climate change: Humans and nature will adapt to the changes</td><td>5</td><td>1.000</td><td>0.400</td><td>50</td><td>Hard *</td></tr><tr><td>56: CC: Questioning the measurements and science: Methodologies/metrics used are unreliable/faulty</td><td>51</td><td>0.333</td><td>0.118</td><td>59</td><td>Medium *</td></tr><tr><td>57: CC: Questioning the measurements and science: Data shows no temperature increase</td><td>10</td><td>0.562</td><td>0.333</td><td>56</td><td>Hard *</td></tr><tr><td>58: CC: Questioning the measurements and science: Greenhouse effect/carbon dioxide do not drive climate change</td><td>5</td><td>0.750</td><td>1.000</td><td></td><td>Hardest</td></tr><tr><td>59: CC: Questioning the measurements and science: Scientific community is unreliable</td><td>45</td><td>0.540</td><td>0.185</td><td>56</td><td>Hard *</td></tr><tr><td>60: CC: Criticism of climate movement: Climate movement is alarmist</td><td>67</td><td>0.619</td><td>0.386</td><td>61</td><td>Hardest *</td></tr><tr><td>61: CC: Criticism of climate movement: Climate movement is corrupt</td><td>35</td><td>0.869</td><td>0.463</td><td>60</td><td>Hardest *</td></tr><tr><td>62: CC: Criticism of climate movement: Ad hominem attacks on key activists</td><td>78</td><td>0.234</td><td>0.217</td><td>60</td><td>Easy *</td></tr><tr><td>63: CC: Controversy about green technologies: Renewable energy is dangerous</td><td>18</td><td>0.421</td><td>0.250</td><td>65</td><td>Hard *</td></tr><tr><td>64: CC: Controversy about green technologies: Renewable energy is unreliable</td><td>42</td><td>0.344</td><td>0.333</td><td>45</td><td>Medium *</td></tr><tr><td>65: CC: Controversy about green technologies: Renewable energy is costly</td><td>27</td><td>0.512</td><td>0.100</td><td>45</td><td>Hard *</td></tr><tr><td>66: CC: Controversy about green technologies: Nuclear energy is not climate friendly</td><td>3</td><td>1.000</td><td>0.750</td><td>45</td><td>Hardest *</td></tr><tr><td>67: CC: Hidden plots by secret schemes of powerful groups: Blaming global elites</td><td>47</td><td>0.438</td><td>0.217</td><td>68</td><td>Hard *</td></tr><tr><td>68: CC: Hidden plots by secret schemes of powerful groups: Climate agenda has hidden motives</td><td>74</td><td>0.548</td><td>0.094</td><td>45</td><td>Hard *</td></tr><tr><td>69: CC: Amplifying Climate Fears: Earth will be uninhabitable soon</td><td>51</td><td>0.432</td><td>0.188</td><td>70</td><td>Hard *</td></tr><tr><td>70: CC: Amplifying Climate Fears: Amplifying existing fears of global warming</td><td>864</td><td>0.124</td><td>0.477</td><td>71</td><td>Easy</td></tr><tr><td>71: CC: Amplifying Climate Fears: Doomsday scenarios for humans</td><td>104</td><td>0.435</td><td>0.239</td><td>70</td><td>Hard *</td></tr><tr><td>72: CC: Amplifying Climate Fears: Whatever we do it is already too late</td><td>18</td><td>0.264</td><td>0.111</td><td>70</td><td>Medium *</td></tr><tr><td>73: CC: Green policies are geopolitical instruments: Climate-related international relations are abusive/exploitative</td><td>7</td><td>0.785</td><td>0.273</td><td>68</td><td>Hardest *</td></tr><tr><td>74: CC: Green policies are geopolitical instruments: Green activities are a form of neo-colonialism</td><td>10</td><td>0.375</td><td>0.333</td><td>68</td><td>Medium *</td></tr></table>
+
+Table 5: Label difficulty, %conf is the percentage of annotation with that label that results in an inconsistency, while %other is the percentage that resulted in an inconsistency with label other. "mcl" is the most common label with which a specific label is confused. "difficulty" make assess the difficulty of the label based on %conf: Easiest (≤ .1), Easy (≤ .25), Medium (≤ .4), and Hard (≤ .4) or Hardest (> 0.6), a star indicate that there is more confusion within the labels of the taxonomy than with the Other class, meaning that it is extra difficult for the annotator

ACL/2025/PolyNarrative_ A Multilingual, Multilabel, Multi-domain Dataset for Narrative Extraction from News Articles/images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac34043a02feddba00ba04c35cc14e53d348af73e14e67504392394c608cd936
+size 920570

ACL/2025/PolyNarrative_ A Multilingual, Multilabel, Multi-domain Dataset for Narrative Extraction from News Articles/layout.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69257e5cba35993f45d5e8516b792a0d645e60dbb92cde6d1db89cc7a9c0e363
+size 706493

ACL/2025/PopAlign_ Diversifying Contrasting Patterns for a More Comprehensive Alignment/41a55097-3fc8-473e-995f-c32fe1d5258f_content_list.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b7e171ea912ddb7391a6c0cc37175155e02ea92742650574eb3b8b42db4775c
+size 224589

ACL/2025/PopAlign_ Diversifying Contrasting Patterns for a More Comprehensive Alignment/41a55097-3fc8-473e-995f-c32fe1d5258f_model.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11bb8e3b1b81c00cae194cd63ce256553daed94322973b486a0175ba0db439c7
+size 250531