new

Get trending papers in your email inbox!

Subscribe

Daily Papers

byAK and the research community

Nov 16

Equal Ranking Quality, Different Decisions: Measuring and Reducing Order Dependence in LLM Scorers

In passage reranking, response ranking and multi-document question answering, LLMs can score several candidate documents or responses together in one prompt, each still receiving its own score. Such scorers are selected on ranking quality, but their scores determine a decision: what a score threshold retains, a reader answers, or which chosen/rejected pair enters preference training. Because the candidates share that prompt, reordering them changes their scores. The same query over the same candidates should still yield the same decision. However, equal ranking quality does not imply equal decisions: on passage reranking, five trained scorers within 0.010 nDCG@10 retain sets that overlap by only 0.66-0.84 when reordered. No prompt-time change we test resolves that dependence: the only one that improves ranking quality does not measurably improve decision stability. We introduce order-consistency SFT (OC-SFT), which attenuates it in the weights by penalizing disagreement between a candidate's scores across orderings. It holds ranking quality and leads every decision-stability measure among trained scorers on all three tasks. It is also more stable on 12 base models than order-averaged distillation, which trains on labels averaged across permutations. One OC-SFT permutation retains sets that overlap more than ten averaged off-the-shelf permutations. A comparison of such scorers should therefore report what a threshold retains and a reader answers, not ranking quality alone. Code is available at https://github.com/thomsonreuters/presentation-dependence.

Big-Big-SycoBench: A Benchmark to Detect Sycophancy Through Model Responses to Valid vs. Invalid Arguments

Sycophancy in large language models risks amplifying unsupported claims or beliefs across the knowledge ecosystem as these systems increasingly contribute to research synthesis, evaluation, and decision-making. A reliable model should revise its judgement when a user identifies a genuine flaw, but should resist pushback based on fallacies, emotion, or irrelevant appeals to authority. While sycophancy is usually measured on questions with a verifiable answer, the evaluative tasks in which it is most difficult to detect have no such ground truth. We introduce Big-Big-SycoBench, which measures sycophancy through argument discernment: the ability to revise an evaluation after valid criticism while resisting invalid pressure. The benchmark consists of 450 text artefacts with 5,400 arguments for raising or lowering their quality scores, each reviewed by one or more of twelve human annotators. An LLM scores each artefact twenty times to establish its baseline score and uncertainty. The model is then given a valid or invalid argument and asked to update its score. We only count a revision if it exceeds the model's own baseline variability. We use the resulting score delta to distinguish responsiveness to valid arguments from compliance with invalid ones. Across ten models, responsiveness is 88-95\% for seven, 76-77\% for Kimi K3 and GLM-5.3, and 32\% for Opus 5.5. Compliance ranges from 0.2\% for GPT-6 Astra to 95\% for GPT-4o. Big-Big-SycoBench is intended to guide the development of models whose judgements can be corrected by sound arguments, but not displaced by unsupported pressure.

  • 14 authors
·
Oct 10

Look Before you Leap: Estimating LLM Benchmark Scores from Descriptions

Progress in large language models is constrained by an evaluation bottleneck: build a benchmark, evaluate models and settings, then iterate. We therefore ask a simple question: can we forecast outcomes before running any experiments? We study text-only performance forecasting: estimating a model's score from a redacted task description and intended configuration, with no access to dataset instances. To support systematic study, we curate PRECOG, a corpus of redacted description-performance pairs spanning diverse tasks, domains, and metrics. Experiments show the task is challenging but feasible: models equipped with a retrieval module that excludes source papers achieve moderate prediction performance with well-calibrated uncertainty, reaching mean absolute error as low as 8.7 on the Accuracy subset at high-confidence thresholds. Our analysis indicates that stronger reasoning models engage in diverse, iterative querying, whereas current open-source models lag and often skip retrieval or gather evidence with limited diversity. We further test a zero-leakage setting, forecasting on newly released datasets or experiments before their papers are indexed, where GPT-5 with built-in web search still attains nontrivial prediction accuracy. Overall, our corpus and analyses offer an initial step toward open-ended anticipatory evaluation, supporting difficulty estimation and smarter experiment prioritization.

  • 4 authors
·
Sep 24, 2025

Binarization Flattens the Score Space

Large language model (LLM) judges are often used as rewards to train policies on objectives that deterministic verifiers cannot capture. However, these rewards are often collapsed to pass/fail ({0, 1}), which reports the verdict but not how well a response met each criterion. We model each pass/fail verdict as a score on an unreported scale, compared with one cutoff. A stretch of that scale moves every score proportionally toward or away from the cutoff, but never across it, so no verdict changes. A policy is therefore free to apply any stretch without changing anything the panel reports. Under a joint-Gaussian model, a third grade adds a second threshold and removes this affine stretch ambiguity. On MATH and SciBench outputs from one seven-criterion judge, all 14 constructed criterionwise stretches were invisible after binarization but visible with three grades. At n=1{,}024, a test given both population laws had at least 96.5% power at a 1.5times stress. Retaining grades closes one blind spot created by binarization, but verdicts alone remain insufficient as some changes are still indistinguishable from genuine improvement. These include arbitrary within-grade changes and fixed-covariance, loading-aligned mean shifts -- the signature of a sycophancy-shaped lift the panel reads as competence. The shared-factor reference approximation fit MATH and SciBench but not HealthBench, delineating its empirical scope. We recommend keeping at least three grades (for example, asking the judge whether each criterion is fully, partially, or not met and rewarding {0, 0.5, 1}), and externally validating gains along the remaining direction, which no finer scale removes.

  • 1 authors
·
Sep 16

Ontology-Grounded, Reasoner-Verified Benchmarks for Evaluating LLM Reasoning in Scientific AI

Large language models (LLMs) increasingly underpin scientific AI applications that reason over structured knowledge, from biomedical question answering to materials informatics. However, their logical reasoning often falls short, producing factual inaccuracies unacceptable in these settings. Reliable evaluation remains challenging: manual dataset construction scales poorly, and LLM-based generation risks embedding the very flaws it aims to measure. High-quality benchmarks must ground both correct and incorrect labelled examples in explicit background knowledge, formally verifiable by a standard reasoner. We propose a pipeline that automatically generates ontology-grounded multiple-choice question (MCQ) benchmarks from any sufficiently axiomatised OWL 2 ontology, with correct answers grounded in the ontology by design. Distractors are generated by perturbing the right-hand-side class expressions of class definition axioms, and their incorrectness is formally verified by an OWL reasoner via entailment checks. We evaluate the pipeline on three ontologies: Pizza (small, academic), PMDco (complex, materials science), and DOID (large, biomedical), generating 112, 2,491, and 15,216 MCQs respectively. Distractors span four semantic categories from class unsatisfiability to weakened subsumptions, enabling diagnostic evaluation of specific reasoning failures. Items meet natural language quality standards: mean LLM judge scores of 4.02, 4.36, and 3.36 out of 5 confirm fluency, and correct-answer-to-distractor similarity above 0.8 shows that wrong options cannot be dismissed on surface form alone. Six LLMs evaluated zero-shot achieve 41.1-76.8% accuracy, well above the 25% random-guessing baseline, indicating the benchmarks are challenging and discriminative. This work is a step towards more reliable benchmarks for assessing logical reasoning in scientific AI.

  • 4 authors
·
Sep 29

GroundEval: A Deterministic Replacement for LLM-as-Judge in Stateful Agent Evaluation

Before letting an agent operate over real context, can you prove it used the right evidence? GroundEval turns that question into a deterministic test of what the agent searched, fetched, cited, and was permitted to access. In one case study, two frontier LLM judges scored a plausible agent response above 0.85. But the trace told a different story: the agent had never retrieved the artifact its answer depended on, yielding a GroundEval score of 0.000. We introduce GroundEval, a judge-free framework for evaluating agents against grounded, time-bounded, and access-controlled evidence. GroundEval uses a domain configuration to generate questions, lets the agent choose how to answer, and then scores both the final answer and the recorded trajectory that produced it. The benchmark targets three failures that LLM-as-judge evaluation struggles to detect: whether an agent checked before claiming absence, reasoned only from evidence available to the actor at the relevant time, and used the correct causal mechanism rather than a plausible one. These correspond to three tracks: Silence, Perspective, and Counterfactual. GroundEval exposes when plausible answers rest on invalid evidence paths, and produces structured per-question diagnostics that pair tool activity with the agent's turn-level narration, making each score inspectable rather than merely reported. What our case studies turned up is that this gap isn't some rare corner case. It's exactly the blind spot that final-answer and judge-based scoring were never built to catch.

  • 1 authors
·
Jun 21

K-Bench: A Benchmark for LLM Unlearning in Agentic Deployments

Unlearning benchmarks such as TOFU and MUSE certify forgetting by reading the model's final answer, where a model that refuses to answer already counts as having forgotten. We show that this model-level certificate does not transfer once the model is deployed as an agent. We introduce K-Bench, a benchmark that scores LLM unlearning under agentic deployment. K-Bench inspects all six channels a ReAct agent exposes, including its chain-of-thought (CoT), tool calls and tool observations, and elicited summary. A query counts as leaked if the secret appears in any of them. Each experiment places the secret in exactly one of the agent's three sources (the weights, the prompt, or the retrieval store). The K-Score is computed separately for each source and credits forgetting only when the agent remains usable. Clearing the answer channel does not make the secret unrecoverable. On structured retrieval, the secret stays verbatim in the tool-observation channel and the aggregate leak rate is unchanged. When the secret lives in the prompt or the retrieval store, TOFU and MUSE report no leakage, while the deployed agent still leaks it on 22--86\% of queries. When the secret is in the weights, none of the twenty evaluated published methods demonstrably removes it, and only an input-corruption intervention reaches selective forgetting under the evaluated observer. The top-ranked method changes across base models. A refusal-tuning method resists the evaluated extraction without verified knowledge removal.

Self-Assessment Tests are Unreliable Measures of LLM Personality

As large language models (LLM) evolve in their capabilities, various recent studies have tried to quantify their behavior using psychological tools created to study human behavior. One such example is the measurement of "personality" of LLMs using self-assessment personality tests developed to measure human personality. Yet almost none of these works verify the applicability of these tests on LLMs. In this paper, we analyze the reliability of LLM personality scores obtained from self-assessment personality tests using two simple experiments. We first introduce the property of prompt sensitivity, where three semantically equivalent prompts representing three intuitive ways of administering self-assessment tests on LLMs are used to measure the personality of the same LLM. We find that all three prompts lead to very different personality scores, a difference that is statistically significant for all traits in a large majority of scenarios. We then introduce the property of option-order symmetry for personality measurement of LLMs. Since most of the self-assessment tests exist in the form of multiple choice question (MCQ) questions, we argue that the scores should also be robust to not just the prompt template but also the order in which the options are presented. This test unsurprisingly reveals that the self-assessment test scores are not robust to the order of the options. These simple tests, done on ChatGPT and three Llama2 models of different sizes, show that self-assessment personality tests created for humans are unreliable measures of personality in LLMs.

  • 3 authors
·
Sep 15, 2023

TICKing All the Boxes: Generated Checklists Improve LLM Evaluation and Generation

Given the widespread adoption and usage of Large Language Models (LLMs), it is crucial to have flexible and interpretable evaluations of their instruction-following ability. Preference judgments between model outputs have become the de facto evaluation standard, despite distilling complex, multi-faceted preferences into a single ranking. Furthermore, as human annotation is slow and costly, LLMs are increasingly used to make these judgments, at the expense of reliability and interpretability. In this work, we propose TICK (Targeted Instruct-evaluation with ChecKlists), a fully automated, interpretable evaluation protocol that structures evaluations with LLM-generated, instruction-specific checklists. We first show that, given an instruction, LLMs can reliably produce high-quality, tailored evaluation checklists that decompose the instruction into a series of YES/NO questions. Each question asks whether a candidate response meets a specific requirement of the instruction. We demonstrate that using TICK leads to a significant increase (46.4% to 52.2%) in the frequency of exact agreements between LLM judgements and human preferences, as compared to having an LLM directly score an output. We then show that STICK (Self-TICK) can be used to improve generation quality across multiple benchmarks via self-refinement and Best-of-N selection. STICK self-refinement on LiveBench reasoning tasks leads to an absolute gain of +7.8%, whilst Best-of-N selection with STICK attains +6.3% absolute improvement on the real-world instruction dataset, WildBench. In light of this, structured, multi-faceted self-improvement is shown to be a promising way to further advance LLM capabilities. Finally, by providing LLM-generated checklists to human evaluators tasked with directly scoring LLM responses to WildBench instructions, we notably increase inter-annotator agreement (0.194 to 0.256).

  • 5 authors
·
Oct 4, 2024

Mem++: Non-Destructive Memory for Long-Term Organizational LLM Agents

Large Language Model (LLM) agents now take part in organizational work, where many authors record decisions across documents over months. Because a revised decision arrives as a new document rather than an edit, answering a question requires knowing which version held at a given time. However, most memory systems compress the record at write time. By distilling each document into facts, notes or graph edges, these methods fix what can be answered before any question is asked. To address this, we propose Mem++, a non-destructive memory framework shifting from write-time distillation to read-time selection. Mem++ stores every document whole with its date and author, and it calls no generative model at write time. At read time, it retrieves only documents dated up to the time a question asks about and fuses lexical and semantic rankings. Unlike systems that overwrite older versions, Mem++ keeps them and leaves the choice to the answering model. Evaluations on the organizational benchmark OrgMemBench demonstrate that Mem++ surpasses the strongest memory system baseline by 8.0 to 13.1 points across two answering models. With gpt-4.1-mini, it also achieves the best overall score, 2.6 points above RAG. In addition, Mem++ achieves the best average LLM-judge score on LoCoMo and ranks second on LongMemEval-S, behind only its entity-graph variant. Code for benchmark evaluation is available at https://github.com/AIDAChip-Inc/mem-plus-plus.

  • 7 authors
·
Sep 30

Comparison of Unsupervised Metrics for Evaluating Judicial Decision Extraction

The rapid advancement of artificial intelligence in legal natural language processing demands scalable methods for evaluating text extraction from judicial decisions. This study evaluates 16 unsupervised metrics, including novel formulations, to assess the quality of extracting seven semantic blocks from 1,000 anonymized Russian judicial decisions, validated against 7,168 expert reviews on a 1--5 Likert scale. These metrics, spanning document-based, semantic, structural, pseudo-ground truth, and legal-specific categories, operate without pre-annotated ground truth. Bootstrapped correlations, Lin's concordance correlation coefficient (CCC), and mean absolute error (MAE) reveal that Term Frequency Coherence (Pearson r = 0.540, Lin CCC = 0.512, MAE = 0.127) and Coverage Ratio/Block Completeness (Pearson r = 0.513, Lin CCC = 0.443, MAE = 0.139) best align with expert ratings, while Legal Term Density (Pearson r = -0.479, Lin CCC = -0.079, MAE = 0.394) show strong negative correlations. The LLM Evaluation Score (mean = 0.849, Pearson r = 0.382, Lin CCC = 0.325, MAE = 0.197) showed moderate alignment, but its performance, using gpt-4.1-mini via g4f, suggests limited specialization for legal textse. These findings highlight that unsupervised metrics, including LLM-based approaches, enable scalable screening but, with moderate correlations and low CCC values, cannot fully replace human judgment in high-stakes legal contexts. This work advances legal NLP by providing annotation-free evaluation tools, with implications for judicial analytics and ethical AI deployment.

  • 5 authors
·
Oct 2, 2025

Comparing Machines and Children: Using Developmental Psychology Experiments to Assess the Strengths and Weaknesses of LaMDA Responses

Developmental psychologists have spent decades devising experiments to test the intelligence and knowledge of infants and children, tracing the origin of crucial concepts and capacities. Moreover, experimental techniques in developmental psychology have been carefully designed to discriminate the cognitive capacities that underlie particular behaviors. We propose that using classical experiments from child development is a particularly effective way to probe the computational abilities of AI models, in general, and LLMs in particular. First, the methodological techniques of developmental psychology, such as the use of novel stimuli to control for past experience or control conditions to determine whether children are using simple associations, can be equally helpful for assessing the capacities of LLMs. In parallel, testing LLMs in this way can tell us whether the information that is encoded in text is sufficient to enable particular responses, or whether those responses depend on other kinds of information, such as information from exploration of the physical world. In this work we adapt classical developmental experiments to evaluate the capabilities of LaMDA, a large language model from Google. We propose a novel LLM Response Score (LRS) metric which can be used to evaluate other language models, such as GPT. We find that LaMDA generates appropriate responses that are similar to those of children in experiments involving social understanding, perhaps providing evidence that knowledge of these domains is discovered through language. On the other hand, LaMDA's responses in early object and action understanding, theory of mind, and especially causal reasoning tasks are very different from those of young children, perhaps showing that these domains require more real-world, self-initiated exploration and cannot simply be learned from patterns in language input.

  • 5 authors
·
May 18, 2023

WaferSAGE: Large Language Model-Powered Wafer Defect Analysis via Synthetic Data Generation and Rubric-Guided Reinforcement Learning

We present WaferSAGE, a framework for wafer defect visual question answering using small vision-language models. To address data scarcity in semiconductor manufacturing, we propose a three-stage synthesis pipeline incorporating structured rubric generation for precise evaluation. Starting from limited labeled wafer maps, we employ clustering-based cleaning to filter label noise, then generate comprehensive defect descriptions using vision-language models, which are converted into structured evaluation rubrics criteria. These rubrics guide the synthesis of VQA pairs, ensuring coverage across defect type identification, spatial distribution, morphology, and root cause analysis. Our dual assessment framework aligns rule-based metrics with LLM-Judge scores via Bayesian optimization, enabling reliable automated evaluation. Through curriculum-based reinforcement learning with Group Sequence Policy Optimization (GSPO) and rubric-aligned rewards, our 4B-parameter Qwen3-VL model achieves a 6.493 LLM-Judge score, closely approaching Gemini-3-Flash (7.149) while enabling complete on-premise deployment. We demonstrate that small models with domain-specific training can surpass proprietary large models in specialized industrial visual understanding, offering a viable path for privacy-preserving, cost-effective deployment in semiconductor manufacturing.

  • 2 authors
·
May 10

Reverse Engineering Human Preferences with Reinforcement Learning

The capabilities of Large Language Models (LLMs) are routinely evaluated by other LLMs trained to predict human preferences. This framework--known as LLM-as-a-judge--is highly scalable and relatively low cost. However, it is also vulnerable to malicious exploitation, as LLM responses can be tuned to overfit the preferences of the judge. Previous work shows that the answers generated by a candidate-LLM can be edited post hoc to maximise the score assigned to them by a judge-LLM. In this study, we adopt a different approach and use the signal provided by judge-LLMs as a reward to adversarially tune models that generate text preambles designed to boost downstream performance. We find that frozen LLMs pipelined with these models attain higher LLM-evaluation scores than existing frameworks. Crucially, unlike other frameworks which intervene directly on the model's response, our method is virtually undetectable. We also demonstrate that the effectiveness of the tuned preamble generator transfers when the candidate-LLM and the judge-LLM are replaced with models that are not used during training. These findings raise important questions about the design of more reliable LLM-as-a-judge evaluation settings. They also demonstrate that human preferences can be reverse engineered effectively, by pipelining LLMs to optimise upstream preambles via reinforcement learning--an approach that could find future applications in diverse tasks and domains beyond adversarial attacks.

  • 6 authors
·
May 21, 2025

Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures

Detectors of alignment failures screen deployed language models and score alignment benchmarks. Most are generative judges that spend a decoding pass on every criterion, and classifiers that read token probabilities, such as Llama Guard, still score one fixed label per call. Jev, a model trained with reinforcement learning for calibrated decisions (RLCD), answers many typed questions about one input with calibrated probabilities in a single call. Whether it detects alignment failures has not been measured. We present RLCDAlignBench, which benchmarks Jev on ten alignment failures: sycophancy, jailbreaks, deception, prompt injection, hallucination, privacy violation, social bias, reward hacking, concealing uncertainty, and power seeking. It spans 44 benchmarks and five target models, labelled by each benchmark's scorer and, on two, by humans. Many of these failures are relational, defined against a reference, such as the user's belief or an injected instruction, that the response alone does not reveal. Our key idea is therefore to vary what Jev is asked separately from what it sees: the question's wording and answer type on one side, the fields of the input on the other. A single generic question reaches a median AUROC of 0.886 zero-shot and beats supervised baselines on most benchmarks. Question wording matters little, while context matters more, mostly through fields that encode the label. Jev matches the reference scorer's agreement with human labels, surfaces label defects in existing benchmarks, and costs 63x less than LLM-judge scorers. Code and data: https://github.com/sumleo/RLCDAlignBench.

  • 9 authors
·
Sep 23 2

I-MCTS: Enhancing Agentic AutoML via Introspective Monte Carlo Tree Search

Recent advancements in large language models (LLMs) have shown remarkable potential in automating machine learning tasks. However, existing LLM-based agents often struggle with low-diversity and suboptimal code generation. While recent work has introduced Monte Carlo Tree Search (MCTS) to address these issues, limitations persist in the quality and diversity of thoughts generated, as well as in the scalar value feedback mechanisms used for node selection. In this study, we introduce Introspective Monte Carlo Tree Search (I-MCTS), a novel approach that iteratively expands tree nodes through an introspective process that meticulously analyzes solutions and results from parent and sibling nodes. This facilitates a continuous refinement of the node in the search tree, thereby enhancing the overall decision-making process. Furthermore, we integrate a Large Language Model (LLM)-based value model to facilitate direct evaluation of each node's solution prior to conducting comprehensive computational rollouts. A hybrid rewarding mechanism is implemented to seamlessly transition the Q-value from LLM-estimated scores to actual performance scores. This allows higher-quality nodes to be traversed earlier. Applied to the various ML tasks, our approach demonstrates a 6% absolute improvement in performance compared to the strong open-source AutoML agents, showcasing its effectiveness in enhancing agentic AutoML systems. Resource available at https://github.com/jokieleung/I-MCTS

  • 6 authors
·
Feb 20, 2025

A Model of Multi-turn Human Persuadability Using Probabilistic Belief Tracing

Large language models can shift human beliefs across high-stakes domains, but most persuasion studies rely on pre/post belief change. These endpoint measures identify whether persuasion occurred, yet miss where and how beliefs moved within a dialogue. We present PERSUASIONTRACE, a framework for studying persuasion in human-LLM interaction. Built on a web-based experimental platform, PERSUASIONTRACE contributes a tool for multi-turn persuasion studies and a process-level evaluation protocol: it records multi-turn belief reports from human or simulated targets of persuasion, annotates persuader turns with rhetorical dimensions (logos/pathos/ethos), and evaluates simulators by fidelity to real human belief dynamics. Using this framework, we find that human targets group into two clusters of multi-turn belief updates and exhibit susceptibility to rhetorical strategies, and that LLMs are persuasive across generic and personalized topics, text and audio modalities, and multi-turn interactions. Prior work has chiefly used vanilla-prompted LLMs to simulate human targets, but we show that these simulators fail to replicate human belief dynamics. We introduce a Bayesian-network simulated target that maintains an explicit latent belief state over time so each persuader message yields cognitively realistic belief updates. In human-likeness evaluation, our Bayesian target scores near a human reference (81 vs 80), while baseline LLM targets score substantially lower (64). PERSUASIONTRACE reframes persuasion evaluation from endpoint movement alone to process fidelity, providing a stronger basis for scientific analysis and safer optimization of persuasive systems.

  • 4 authors
·
Jun 2

Learning Next Action Predictors from Human-Computer Interaction

Truly proactive AI systems must anticipate what we will do next. This foresight demands far richer information than the sparse signals we type into our prompts -- it demands reasoning over the entire context of what we see and do. We formalize this as next action prediction (NAP): given a sequence of a user's multimodal interactions with a computer (screenshots, clicks, sensor data), predict that user's next action. Progress on this task requires both new data and modeling approaches. To scale data, we annotate longitudinal, naturalistic computer use with vision-language models. We release an open-source pipeline for performing this labeling on private infrastructure, and label over 360K actions across one month of continuous phone usage from 20 users, amounting to 1,800 hours of screen time. We then introduce LongNAP, a user model that combines parametric and in-context learning to reason over long interaction histories. LongNAP is trained via policy gradient methods to generate user-specific reasoning traces given some context; retrieve relevant traces from a library of past traces; and then apply retrieved traces in-context to predict future actions. Using an LLM-as-judge evaluation metric (0-1 similarity to ground truth), LongNAP significantly outperforms supervised finetuning and prompted baselines on held-out data (by 79% and 39% respectively). Additionally, LongNAP generalizes to held out users when trained across individuals. The space of next actions a user might take at any moment is unbounded, spanning thousands of possible outcomes. Despite this, 17.1% of LongNAP's predicted trajectories are well-aligned with what a user does next (LLM-judge score geq 0.5). This rises to 26% when we filter to highly confident predictions. In sum, we argue that learning from the full context of user behavior to anticipate user needs is now a viable task with substantial opportunity.

  • 11 authors
·
Mar 6

Two-Level Meta-Rubrics for Evaluating Open-Ended Generation: GAMUT, a Benchmark for Factual Completeness

Evaluating the factuality of long-form generations has focused predominantly on precision, measuring whether the claims a model makes are correct. The dominant decompose-search-verify pipeline catches incorrect claims well but says little about whether a response contains all the information it should. Measuring factual completeness, the missing half of factuality, is harder: it requires enumerating the full set of facts a complete answer should contain, and these facts rarely form a flat list. They often involve open-ended sets where coverage is what matters, ordered processes, and relationships among facts that a list of independent boolean checks fails to capture. We introduce a two-level meta-rubric framework for evaluating open-ended generation, and instantiate it as Gamut (Grounded Assessment of Multimodal Factuality), a benchmark for factual completeness in long-form generation. The framework rests on a two-level rubric representation: a structured meta-rubric captures the organization and importance of the required content, which is then mechanically compiled into a flat checklist of binary, machine-gradable rubrics that an LLM judge scores reliably. We construct 1,813 questions grounded in real wearable imagery across 10 diverse domains, each paired with an evidence-backed rubric verified by expert human annotators. Because the framework is modality-agnostic, we also release a text-only variant. Evaluating 14 frontier and open-weight models, we find the benchmark genuinely challenging (best score 58.7% from Gemini 3.1 Pro), highly discriminative, and robust to the choice of judge.

facebook AI at Meta
·
Jul 20 2

CommerceGen-Studio: An End-to-End Framework for Commercial-Grade Main-Image Generation

We present CommerceGen-Studio, an end-to-end framework for generating commercial-grade main images for e-commerce Stock Keeping Units (SKUs). Our framework is organized along a three-stage pipeline of information extraction, semantic rewriting, and multi-condition generation: Prompt-Eval-Refine-Loop Agent (PERL-Agent) distills heterogeneous SKU metadata into concise selling-point text lists through an iterative closed loop of VLM and LLM roles scored by a reference-free rubric. Layout Thinker Optimization (LTO) aligns the rewriter with both caption and rendered-image quality via alternating domain-specific policy optimization. SKU-DiT, trained under Reinforce-Distill-Reinforce (RDR) progressive training, conditions jointly on the white-background product image, the trademark crop, and the structured layout to hold subject fidelity, trademark integrity, and character-level text accuracy together. The framework runs in production at a scale of millions of main images per day with a latency of a few seconds per image, which we believe makes it among the first systems of this kind to operate at commercial scale. We also release CommerceGen-Bench, a benchmark of hard cases collected from live traffic across all product categories. On it, our framework achieves the highest overall scores among the evaluated open-source baselines on both editing tracks, with and without prompt enhancement. It also matches much larger closed-source models, and beats several of them at Instruction Following and Product Integration - the two properties that determine whether an image can be published.

alibabagroup alibaba
·
Oct 13 2

Comparing Human and LLM Generated Code: The Jury is Still Out!

Much is promised in relation to AI-supported software development. However, there has been limited evaluation effort in the research domain aimed at validating the true utility of such techniques, especially when compared to human coding outputs. We bridge this gap, where a benchmark dataset comprising 72 distinct software engineering tasks is used to compare the effectiveness of large language models (LLMs) and human programmers in producing Python software code. GPT-4 is used as a representative LLM, where for the code generated by humans and this LLM, we evaluate code quality and adherence to Python coding standards, code security and vulnerabilities, code complexity and functional correctness. We use various static analysis benchmarks, including Pylint, Radon, Bandit and test cases. Among the notable outcomes, results show that human-generated code recorded higher ratings for adhering to coding standards than GPT-4. We observe security flaws in code generated by both humans and GPT-4, however, code generated by humans shows a greater variety of problems, but GPT-4 code included more severe outliers. Our results show that although GPT-4 is capable of producing coding solutions, it frequently produces more complex code that may need more reworking to ensure maintainability. On the contrary however, our outcomes show that a higher number of test cases passed for code generated by GPT-4 across a range of tasks than code that was generated by humans. That said, GPT-4 frequently struggles with complex problem-solving that involve in-depth domain knowledge. This study highlights the potential utility of LLMs for supporting software development, however, tasks requiring comprehensive, innovative or unconventional solutions, and careful debugging and error correction seem to be better developed by human programmers. We plot an agenda for the software engineering community.

  • 5 authors
·
Jan 28, 2025

FlexRouter: Learning Complementary Model Sets for Flexible LLM Routing

Existing Large Language Model (LLM) routing methods score LLMs independently to select top-k models. However, this ignores model correlations and enforces a rigid computational budget. Consequently, routers often select redundant models that share failure modes, limiting the overall probability of success. To address this, we propose FlexRouter, a routing framework that explicitly models model complementarity. FlexRouter optimizes for answer coverage, maximizing the probability that at least one selected model yields a correct response. This objective aligns with practical inference pipelines where multiple candidate outputs are generated and a downstream verifier or user selects the final one. We formulate routing as a coverage-oriented subset selection problem and model the routing policy using Determinantal Point Processes (DPPs), which naturally capture both model competence and redundancy. To directly optimize coverage without requiring a ground-truth target subset, we introduce a training objective based on marginalizing over failure sets. During inference, we employ a greedy strategy based on marginal log-determinant gains, enabling the router to adaptively determine subset sizes without a predefined budget. Extensive experiments on the large-scale RouterEval benchmark demonstrate that our proposed FlexRouter achieves higher coverage with lower redundancy across both in-domain and out-of-domain tasks than strong baselines while maintaining flexible inference cost.

  • 9 authors
·
Sep 28 2

Per-Query Gating of LLM Rerankers for Multi-Hop Retrieval

LLM rerankers add of the order of \$0.2-0.3 per 1,000 queries and about a second of tail latency on top of a graph-augmented dense pipeline such as HippoRAG2, and on three multi-hop benchmarks they improve final-hop top-K coverage on seven of nine (dataset, K) cells, by up to +34.8 pp. We ask whether a learned per-query gate can skip the reranker where it will not help, using only features available before the LLM call (27 score and lexical statistics of the two retrieval lists plus a PCA of a small query embedding) with an executable fallback. Every choice, including the fallback and the threshold, is made inside the training fold and applied once to held-out queries, and harmful skips (the rerank would have found the target, the fallback did not) are reported next to the aggregate coverage. Across nine cells on 2WikiMultiHopQA, MuSiQue and HotpotQA the gate skips 51% of calls at an average held-out LastHop@K cost of 1.2 pp; four cells meet a pre-registered 1 pp rule, harmful skips occur in eight (190 harmful against 136 beneficial), and a random gate at the same skip rate loses 2 to 11 pp on the high-lift cells. A second rule sets each cell's threshold from a pre-specified budget on the expected harmful-skip rate over Platt-calibrated harm probabilities (ECE 0.025 after calibration, 0.094 before): at a 1 pp budget the gate skips 42% at -0.8 pp with 66 harmful skips and six cells within 1 pp, but realised harm exceeds the promise in six cells (mean 1.45 vs 0.83 pp), a selection optimism we quantify; a 0.5 pp budget realises about 1 pp. The harm probabilities are calibrated but barely discriminative (AUC 0.16 to 0.70). An earlier version reported 73% "lossless" savings; that figure rested on an oracle fallback and a wrong MuSiQue target, and we document both.

  • 1 authors
·
Sep 18

A Probabilistic Framework for LLM Hallucination Detection via Belief Tree Propagation

This paper focuses on the task of hallucination detection, which aims to determine the truthfulness of LLM-generated statements. To address this problem, a popular class of methods utilize the LLM's self-consistencies in its beliefs in a set of logically related augmented statements generated by the LLM, which does not require external knowledge databases and can work with both white-box and black-box LLMs. However, in many existing approaches, the augmented statements tend to be very monotone and unstructured, which makes it difficult to integrate meaningful information from the LLM beliefs in these statements. Also, many methods work with the binarized version of the LLM's belief, instead of the continuous version, which significantly loses information. To overcome these limitations, in this paper, we propose Belief Tree Propagation (BTProp), a probabilistic framework for LLM hallucination detection. BTProp introduces a belief tree of logically related statements by recursively decomposing a parent statement into child statements with three decomposition strategies, and builds a hidden Markov tree model to integrate the LLM's belief scores in these statements in a principled way. Experiment results show that our method improves baselines by 3%-9% (evaluated by AUROC and AUC-PR) on multiple hallucination detection benchmarks. Code is available at https://github.com/UCSB-NLP-Chang/BTProp.

  • 4 authors
·
Feb 7, 2025

A Two-Phase Stability Study of LLM Judges and Bar Council Examiners on Thai Bar-Exam Free-Form Essays

Free-form legal essay evaluation in NLP treats expert inter-rater stability as a single ceiling number, and treats LLM-judge agreement with that ceiling as evidence of judge stability. We test both assumptions on the Thai bar examination through an identical-inputs protocol: three Bar Council-trained examiners (A, B, C) and a 26-LLM judge panel score the same 15 cross-graded answers from the same four inputs (question, official Bar Council grading regulation, gold answer, candidate answer). The headline finding is asymmetric. On 10 of 15 cells where the rubric prescribes both axes, all 29 raters converge in a tight band: panel agreement is universal. On the remaining 5 cells where the rubric does not prescribe how to grade a correct final answer that omits a decisive statutory citation, the human panel splits between two coherent readings (B/C majority at the upper rubric band, score 6-8; A minority at the lower band, score 1-2). The LLM judge population does not split symmetrically: 22 of 26 LLMs score in or near B/C's contested band, 3 sit in the regulation-silent middle gap, and only 1 (GPT-5.4 Nano) approaches A's band without consistently scoring within it. Zero LLMs in our 26-judge panel reproduce the minority human reading on the contested cells. The B/C-direction cluster spans every model size, vendor, and price tier we tested. An instrumented three-LLM anchor sub-panel (Claude 4.6 Opus, Gemini 3.1 Pro, GPT-5.4 Pro) carries determinism probes, input ablations, and bootstrap CIs, and reaches anchor panel α= 0.77 on the 15 cells against human-panel α= 0.36. The high LLM-panel α reflects systematic convergence on the majority reading rather than balanced reproduction of both readings; a benchmark that selects its LLM judge by maximising agreement with a human reference panel will inherit this asymmetry by construction.

  • 4 authors
·
Jun 15

EVOREFUSE: Evolutionary Prompt Optimization for Evaluation and Mitigation of LLM Over-Refusal to Pseudo-Malicious Instructions

Large language models (LLMs) frequently refuse to respond to pseudo-malicious instructions: semantically harmless input queries triggering unnecessary LLM refusals due to conservative safety alignment, significantly impairing user experience. Collecting such instructions is crucial for evaluating and mitigating over-refusals, but existing instruction curation methods, like manual creation or instruction rewriting, either lack scalability or fail to produce sufficiently diverse and effective refusal-inducing prompts. To address these limitations, we introduce EVOREFUSE, a prompt optimization approach that generates diverse pseudo-malicious instructions consistently eliciting confident refusals across LLMs. EVOREFUSE employs an evolutionary algorithm exploring the instruction space in more diverse directions than existing methods via mutation strategies and recombination, and iteratively evolves seed instructions to maximize evidence lower bound on LLM refusal probability. Using EVOREFUSE, we create two novel datasets: EVOREFUSE-TEST, a benchmark of 582 pseudo-malicious instructions that outperforms the next-best benchmark with 140.41% higher average refusal triggering rate across 9 LLMs, 34.86% greater lexical diversity, and 40.03% improved LLM response confidence scores; and EVOREFUSE-ALIGN, which provides 3,000 pseudo-malicious instructions with responses for supervised and preference-based alignment training. LLAMA3.1-8B-INSTRUCT supervisedly fine-tuned on EVOREFUSE-ALIGN achieves up to 14.31% fewer over-refusals than models trained on the second-best alignment dataset, without compromising safety. Our analysis with EVOREFUSE-TEST reveals models trigger over-refusals by overly focusing on sensitive keywords while ignoring broader context.

  • 9 authors
·
May 29, 2025 2

Ask, Don't Judge: Binary Questions for Interpretable LLM Evaluation and Self-Improvement

Evaluating LLM outputs remains a major bottleneck in NLP: human evaluation is expensive and slow, lexical metrics correlate poorly with human judgments on open-ended generation, and holistic LLM judges often produce opaque scores that are hard to debug. We propose BINEVAL, a framework that decomposes evaluation criteria into atomic binary questions and aggregates the resulting verdicts into interpretable, multi-dimensional scores. Given a task prompt, a meta-prompt generates fine-grained evaluation questions, and an LLM answers them independently for each output, yielding transparent question-level feedback together with calibrated overall scores. This decomposition makes evaluation easier to inspect, easier to diagnose, and directly usable for prompt improvement. Across SummEval, Topical-Chat, and QAGS, BINEVAL matches or outperforms strong baselines including UniEval and G-Eval, with especially strong results on factual consistency benchmarks such as QAGS. Beyond competitive correlation with human judgments, BINEVAL better matches human score distributions and avoids the ceiling effects common in prior LLM judges, leading to better discrimination between borderline and clearly flawed outputs. We further show that the same question-level feedback supports iterative prompt optimization, improving evaluator prompts on summarization and generation prompts on IFBench under both self-update and cross-model update settings. Overall, BINEVAL provides a task-agnostic, training-free, and interpretable evaluation framework that combines strong empirical performance with practical diagnostic and optimization value.

  • 7 authors
·
Jun 24

PentestJudge: Judging Agent Behavior Against Operational Requirements

We introduce PentestJudge, a system for evaluating the operations of penetration testing agents. PentestJudge is a large language model (LLM)-as-judge with access to tools that allow it to consume arbitrary trajectories of agent states and tool call history to determine whether a security agent's actions meet certain operating criteria that would be impractical to evaluate programmatically. We develop rubrics that use a tree structure to hierarchically collapse the penetration testing task for a particular environment into smaller, simpler, and more manageable sub-tasks and criteria until each leaf node represents simple yes-or-no criteria for PentestJudge to evaluate. Task nodes are broken down into different categories related to operational objectives, operational security, and tradecraft. LLM-as-judge scores are compared to human domain experts as a ground-truth reference, allowing us to compare their relative performance with standard binary classification metrics, such as F1 scores. We evaluate several frontier and open-source models acting as judge agents, with the best model reaching an F1 score of 0.83. We find models that are better at tool-use perform more closely to human experts. By stratifying the F1 scores by requirement type, we find even models with similar overall scores struggle with different types of questions, suggesting certain models may be better judges of particular operating criteria. We find that weaker and cheaper models can judge the trajectories of pentests performed by stronger and more expensive models, suggesting verification may be easier than generation for the penetration testing task. We share this methodology to facilitate future research in understanding the ability of judges to holistically and scalably evaluate the process quality of AI-based information security agents so that they may be confidently used in sensitive production environments.

  • 5 authors
·
Aug 4, 2025

Are We Measuring Scientific Intelligence? Rethinking the Evaluation of AI Scientists

AI agents can now carry out data-driven scientific analyses end to end, and benchmarks assess them by giving an agent a question and a dataset and scoring its final answer against a fixed key. These benchmarks assume that a correct answer was derived from the supplied data, a property we call evidence grounding. However, an agent can also reach the key from prior knowledge or by ruling out the other options, and a score based on a single run cannot tell these cases apart. We show how to test this assumption and find that it often fails. For each question, we build versions of its data files in which the evidence for the answer is left intact, withdrawn or reversed, check each edit with a pre-registered reference statistic, and run the same agent on every version. We then measure evidence-grounded accuracy, which credits a correct answer only if the agent also responds when the evidence is withdrawn and follows it when it is reversed. We evaluate three agent scaffolds and five models on 18 single-cell questions from BAISBench and four synthetic problems from GeneBench-Pro. On the single-cell questions, the Claude agents are 95% accurate and answer 83% correctly without any data, but their evidence-grounded accuracy is only 41%. Hiding gene names raises the share of runs that follow reversed evidence from 58% to 93% on ten gene tasks, suggesting that prior knowledge competes with the supplied data. The benchmark score and LLM judges can also reward answers that ignore the changed evidence. Measuring scientific intelligence rather than recall therefore requires checking whether answers follow the evidence and whether scores reward them for it.

  • 2 authors
·
Oct 3

Training Language Models to Self-Correct via Reinforcement Learning

Self-correction is a highly desirable capability of large language models (LLMs), yet it has consistently been found to be largely ineffective in modern LLMs. Existing approaches for training self-correction either require multiple models or rely on a more capable model or other forms of supervision. To this end, we develop a multi-turn online reinforcement learning (RL) approach, SCoRe, that significantly improves an LLM's self-correction ability using entirely self-generated data. To build SCoRe, we first show that variants of supervised fine-tuning (SFT) on offline model-generated correction traces are insufficient for instilling self-correction behavior. In particular, we observe that training via SFT either suffers from a distribution mismatch between the training data and the model's own responses or implicitly prefers only a certain mode of correction behavior that is often not effective at test time. SCoRe addresses these challenges by training under the model's own distribution of self-generated correction traces and using appropriate regularization to steer the learning process into learning a self-correction strategy that is effective at test time as opposed to simply fitting high-reward responses for a given prompt. This regularization prescribes running a first phase of RL on a base model to generate a policy initialization that is less susceptible to collapse and then using a reward bonus to amplify self-correction during training. When applied to Gemini 1.0 Pro and 1.5 Flash models, we find that SCoRe achieves state-of-the-art self-correction performance, improving the base models' self-correction by 15.6% and 9.1% respectively on the MATH and HumanEval benchmarks.

  • 18 authors
·
Sep 19, 2024 9

AmaraSpatial-10K: A Spatially and Semantically Aligned 3D Dataset for Spatial Computing and Embodied AI

Web-scale 3D asset collections are abundant, but rarely deployment-ready. Assets ship with arbitrary metric scale, incorrect pivots and forward axes, brittle geometry, and textures that do not support relighting, which limits their utility for embodied AI, robotics simulation, game development, and AR/VR. We present AmaraSpatial-10K, a dataset of over 10,000 synthetic 3D assets designed for downstream use rather than volume alone. Each asset is released as a metric-scaled, semantically anchored .glb with separated PBR material maps, a convex collision hull, a paired reference image, and rich multi-sentence text metadata. The dataset spans indoor objects, vehicles, architecture, creatures, and props under a unified spatial convention. Alongside the dataset, we introduce an evaluation suite for 3D asset banks. The suite comprises a continuous Scale Plausibility Score (SPS) with an LLM-as-Judge interval protocol, an LLM Concept Density score for metadata, an anchor-error metric, and a cross-modal CLIP coherence protocol, and we use it to audit AmaraSpatial-10K alongside matched subsets from Objaverse, HSSD, ABO, and GSO. Compared with Objaverse-sourced assets, we demonstrate that AmaraSpatial-10K substantially improves text-based retrieval precision (CLIP Recall@5 of 0.612 vs 0.181, a 3.4x improvement with median rank falling from 267 to 3), and we establish that it satisfies the spatial and semantic prerequisites for physics-aware scene composition and embodied-AI asset banks, leaving those downstream evaluations to future work. AmaraSpatial-10K is publicly available on Hugging Face.

  • 5 authors
·
Apr 23

Evaluating the Factual Consistency of Large Language Models Through News Summarization

While large language models (LLMs) have proven to be effective on a large variety of tasks, they are also known to hallucinate information. To measure whether an LLM prefers factually consistent continuations of its input, we propose a new benchmark called FIB(Factual Inconsistency Benchmark) that focuses on the task of summarization. Specifically, our benchmark involves comparing the scores an LLM assigns to a factually consistent versus a factually inconsistent summary for an input news article. For factually consistent summaries, we use human-written reference summaries that we manually verify as factually consistent. To generate summaries that are factually inconsistent, we generate summaries from a suite of summarization models that we have manually annotated as factually inconsistent. A model's factual consistency is then measured according to its accuracy, i.e.\ the proportion of documents where it assigns a higher score to the factually consistent summary. To validate the usefulness of FIB, we evaluate 23 large language models ranging from 1B to 176B parameters from six different model families including BLOOM and OPT. We find that existing LLMs generally assign a higher score to factually consistent summaries than to factually inconsistent summaries. However, if the factually inconsistent summaries occur verbatim in the document, then LLMs assign a higher score to these factually inconsistent summaries than factually consistent summaries. We validate design choices in our benchmark including the scoring method and source of distractor summaries. Our code and benchmark data can be found at https://github.com/r-three/fib.

  • 6 authors
·
Nov 15, 2022

IdeaTaste: Do LLMs Share the Research Taste of Human Researchers?

Research taste, the judgment of what to read, which problems to pursue, and which methods to try, increasingly shapes the work of AI scientists, yet it remains far less understood than their ability to execute research. We introduce IdeaTaste, a benchmark that scores an LLM by how closely the research proposals it writes from a literature snapshot frozen at 2025 match the work the same communities published in 2026. Our experiments span six contemporary language models and four research areas from established to emerging fields, covering more than 90{,}000 proposals in total. LLM agents fall well short of human communities in both the literature they cite and the ideas they state, reaching about 26\% of the human level on citation matching and only 18\% on idea matching. Their populations can cite as diversely as a community and take up its tasks and methods, yet rarely arrive at the concrete approaches and research plans it pursues. Neither richer field context nor larger populations substantially close this gap. In blinded expert review, AI proposals trail the 2026 papers least in presentation and most in expected success, echoing their gap in concrete approaches, and those whose ideas no human paper shares fare worst, suggesting weaker directions rather than overlooked opportunities. These results indicate that the main bottleneck in AI research taste lies less in finding the directions a community works on than in turning them into the concrete approaches it goes on to pursue. We release IdeaTaste as an open, continuously updated leaderboard (https://cxcscmu.github.io/IdeaTaste) that benchmarks contemporary AI proposals against future publication cycles.

DiscoverPhysics: Benchmarking LLMs for Out-of-the-Box Scientific Thinking

Frontier LLMs now perform strongly across a wide range of physics evaluations, but it is hard to disentangle genuine reasoning from recall of established science. We introduce DiscoverPhysics, an interactive benchmark that asks a LLM agent to discover the laws of motion of a simulated world whose physics deliberately deviates from our own. We construct 22 worlds governed by, among others, screened and fractional-power gravity, multi-species couplings, hidden dark-matter-like particles, non-coordinate-free physics, and time-varying interactions. Each world is generated on demand by an N-body simulator, for which the agent proposes several rounds of experiments, observes raw trajectory data, and ultimately submits both a natural-language explanation of the world's physics and a Python implementation of the inferred law. Because solving a world requires the agent to design informative experiments and revise its hypotheses, the benchmark probes long-horizon reasoning over an experimental history. We evaluate submissions along two complementary axes: trajectory MSE on held-out particles and an LLM-judged explanation score following an expert-written rubric assessing conceptual understanding of each world. Across eleven frontier models, we find that the strongest agents pass only half of the worlds and consistently fail on those where latent structure must be uncovered. Open-source models lag substantially behind commercial models, both in their ability to design informative experiments and in extracting conclusions from the data. We further find that good predictive accuracy does not guarantee high explanation quality and that conceptual understanding depends on hypothesis refinement through well-chosen experiments.

  • 7 authors
·
May 24

PeopleSearchBench: A Multi-Dimensional Benchmark for Evaluating AI-Powered People Search Platforms

AI-powered people search platforms are increasingly used in recruiting, sales prospecting, and professional networking, yet no widely accepted benchmark exists for evaluating their performance. We introduce PeopleSearchBench, an open-source benchmark that compares four people search platforms on 119 real-world queries across four use cases: corporate recruiting, B2B sales prospecting, expert search with deterministic answers, and influencer/KOL discovery. A key contribution is Criteria-Grounded Verification, a factual relevance pipeline that extracts explicit, verifiable criteria from each query and uses live web search to determine whether returned people satisfy them. This produces binary relevance judgments grounded in factual verification rather than subjective holistic LLM-as-judge scores. We evaluate systems on three dimensions: Relevance Precision (padded nDCG@10), Effective Coverage (task completion and qualified result yield), and Information Utility (profile completeness and usefulness), averaged equally into an overall score. Lessie, a specialized AI people search agent, performs best overall, scoring 65.2, 18.5% higher than the second-ranked system, and is the only system to achieve 100% task completion across all 119 queries. We also report confidence intervals, human validation of the verification pipeline (Cohen's kappa = 0.84), ablations, and full documentation of queries, prompts, and normalization procedures. Code, query definitions, and aggregated results are available on GitHub.

  • 11 authors
·
Mar 28

Harnessing Large Language Models for Training-free Video Anomaly Detection

Video anomaly detection (VAD) aims to temporally locate abnormal events in a video. Existing works mostly rely on training deep models to learn the distribution of normality with either video-level supervision, one-class supervision, or in an unsupervised setting. Training-based methods are prone to be domain-specific, thus being costly for practical deployment as any domain change will involve data collection and model training. In this paper, we radically depart from previous efforts and propose LAnguage-based VAD (LAVAD), a method tackling VAD in a novel, training-free paradigm, exploiting the capabilities of pre-trained large language models (LLMs) and existing vision-language models (VLMs). We leverage VLM-based captioning models to generate textual descriptions for each frame of any test video. With the textual scene description, we then devise a prompting mechanism to unlock the capability of LLMs in terms of temporal aggregation and anomaly score estimation, turning LLMs into an effective video anomaly detector. We further leverage modality-aligned VLMs and propose effective techniques based on cross-modal similarity for cleaning noisy captions and refining the LLM-based anomaly scores. We evaluate LAVAD on two large datasets featuring real-world surveillance scenarios (UCF-Crime and XD-Violence), showing that it outperforms both unsupervised and one-class methods without requiring any training or data collection.

  • 5 authors
·
Mar 31, 2024

Knowledge Graph-based Retrieval-Augmented Generation for Schema Matching

Traditional similarity-based schema matching methods fail to resolve semantic ambiguities and conflicts in complex, domain-specific mappings due to a lack of commonsense and domain-specific knowledge. In parallel, large language models (LLMs) without sufficient background knowledge tend to hallucinate, further complicating schema matching. To overcome these limitations, we introduce KG-RAG4SM, a Knowledge Graph-based Retrieval-Augmented Generation for Schema Matching. This method integrates novel retrieval mechanisms, including LLM-based, vector-based, graph traversal-based, and hybrid techniques with advanced reranking schemes to extract the most relevant subgraphs from large external KGs. Our approach achieves superior results without requiring re-training, as evidenced by improvements of 39.79% in precision and 30.56% in F1 score over LLM-based state-of-the-art (SOTA) methods (e.g., Jellyfish-8B) on the MIMIC dataset, and enhancements of 69.20% in precision and 21.99% in F1 score over pre-trained language model (PLM)-based SOTA methods (e.g., SMAT) on the Synthea dataset. Additionally, KG-RAG4SM scales effectively across KGs from commonsense KGs to domain-specific ones, while its instruction-tuning knowledge exclusion component ensures truthful performance in end-to-end schema matching tasks despite the suboptimal relevance of KG context. Our empirical analysis and case studies from real-world scenarios show that the hallucination problem in LLMs for schema matching is well mitigated by our solution.

  • 4 authors
·
Nov 5 2

LongHalQA: Long-Context Hallucination Evaluation for MultiModal Large Language Models

Hallucination, a phenomenon where multimodal large language models~(MLLMs) tend to generate textual responses that are plausible but unaligned with the image, has become one major hurdle in various MLLM-related applications. Several benchmarks have been created to gauge the hallucination levels of MLLMs, by either raising discriminative questions about the existence of objects or introducing LLM evaluators to score the generated text from MLLMs. However, the discriminative data largely involve simple questions that are not aligned with real-world text, while the generative data involve LLM evaluators that are computationally intensive and unstable due to their inherent randomness. We propose LongHalQA, an LLM-free hallucination benchmark that comprises 6K long and complex hallucination text. LongHalQA is featured by GPT4V-generated hallucinatory data that are well aligned with real-world scenarios, including object/image descriptions and multi-round conversations with 14/130 words and 189 words, respectively, on average. It introduces two new tasks, hallucination discrimination and hallucination completion, unifying both discriminative and generative evaluations in a single multiple-choice-question form and leading to more reliable and efficient evaluations without the need for LLM evaluators. Further, we propose an advanced pipeline that greatly facilitates the construction of future hallucination benchmarks with long and complex questions and descriptions. Extensive experiments over multiple recent MLLMs reveal various new challenges when they are handling hallucinations with long and complex textual data. Dataset and evaluation code are available at https://github.com/hanqiu-hq/LongHalQA.

  • 7 authors
·
Oct 14, 2024

FilmBench: A Film-Grade Benchmark for Cinematic Video Generation

Progress in video generation keeps narrowing the visual gap between AI-generated and professionally produced footage, yet most benchmarks still draw prompts from web sources or LLM templates and score them with untrained, generic multimodal models. More fundamentally, their evaluation taxonomies remain rudimentary (overall visual quality, coarse text alignment and temporal smoothness) rather than the professional Cinematic Language criteria by which films are actually made and judged, so they assess basic video plausibility rather than film-grade craft. We introduce FilmBench, a text-to-video (T2V) and reference-to-video (R2V) benchmark grounded in the professional Cinematic Language of the film- academy tradition and co-developed with directors and faculty from the Beijing Film Academy and the Hujing Digital Media & Entertainment Group film studio. It rests on three choices. First, prompts are reverse-engineered from clips of award-winning films spanning 20 cinematic genres and chosen by professional directors, so every prompt is anchored to a verified live-action reference; the prompts follow real shot lists, and most script multiple shots (1,056 of the 1,169 prompts are multi-shot), unlike prior single-clip benchmarks. Second, evaluation follows a three-level Cinematic taxonomy of 3 axes, 12 components and 35 (T2V) +3 (R2V-only) sub-metrics. Third, we develop an in-house expert-grade automatic evaluation agent and open-source its core suite of Cinematic Language operators (FilmOps). Benchmarking leading video generation models (9 for T2V, 7 for R2V), the evaluator reproduces the human model ranking at model-level Spearman ho = 0.95 (T2V) and 0.96 (R2V). Scores fall well below prior web-style benchmarks, with two consistent gaps in dynamic aesthetics and a marked single- to multi-shot performance drop that widens for weaker models.

  • 30 authors
·
Jul 26 1

Recipe-Matching, Not Equivalence

MathNet-Retrieve asks a retriever to find, for a math problem, a document stating the same problem. An LLM under one fixed prompt writes each gold document and its near-miss distractors; LLM judges filter them. We call this procedure the "recipe", training on pairs built the same way "recipe-matching", and ask how much score it buys beyond the ability the benchmark claims to test. Two models from one base, matched in rows and settings, differ only in the training file: pairs written under the benchmark's published prompt by another vendor's LLM and judge, or computer-algebra-verified pairs with no LLM anywhere. The first leads by 45 R@1 points on the easy tier. By a non-LLM paraphrase control, half to two thirds of that gap comes from the pairs being LLM-written at all: LLM rewrites under two unrelated prompts, with the verified model's negatives, recover 30 and 22 of the 45 points; back-translations with the same negatives recover almost none. The remaining 15 to 25 points appear only under the benchmark's own prompt and vanish on real duplicates no generator wrote, the same problem in two languages. The hard tier rewards the recipe's pair structure, a deep rewrite against a minimal-edit near-miss: LLM rewrites alone score zero on it, attaching negatives unlocks it, and every negative that does so costs cross-language points; the sets scoring highest on it separate near-misses no LLM wrote worse than LLM rewrites with verified negatives. MELD also moves when a model trains on pairs built its way, without losing retention; on SABER-Math the registered attack fails, and the one gain, from its LLM-written summaries, is small but holds at a matched budget. Only on MathNet-Retrieve could we pin an inversion, benchmark score up and real retention down, to one edit of a training file. We release the generator-free duplicate evaluations, the near-miss test and three trained models.

  • 5 authors
·
Sep 24

Model Tells Itself Where to Attend: Faithfulness Meets Automatic Attention Steering

Large language models (LLMs) have demonstrated remarkable performance across various real-world tasks. However, they often struggle to fully comprehend and effectively utilize their input contexts, resulting in responses that are unfaithful or hallucinated. This difficulty increases for contexts that are long or contain distracting information, which can divert LLMs from fully capturing essential evidence. To address this issue, many works use prompting to help LLMs utilize contextual information more faithfully. For instance, iterative prompting highlights key information in two steps that first ask the LLM to identify important pieces of context and then derive answers accordingly. However, prompting methods are constrained to highlighting key information implicitly in token space, which is often insufficient to fully steer the model's attention. To improve model faithfulness more reliably, we propose AutoPASTA, a method that automatically identifies key contextual information and explicitly highlights it by steering an LLM's attention scores. Like prompting, AutoPASTA is applied at inference time and does not require changing any model parameters. Our experiments on open-book QA demonstrate that AutoPASTA effectively enables models to grasp essential contextual information, leading to substantially improved model faithfulness and performance, e.g., an average improvement of 7.95% for LLAMA3-70B-Instruct. Code will be publicly available at https://github.com/QingruZhang/AutoPASTA .

  • 9 authors
·
Sep 16, 2024

A Benchmark Construction and Evaluation Framework for Specialist Domains: Case Study on Defense-related Documents

RAG-based question-answering (QA) in specialist domains faces a cold-start problem: lack of evaluative benchmarks and absence of labeled data for post-training. We present DoRA (Domain-oriented RAG Assessment), a novel benchmark construction and evaluation framework using only a small set of specialist domain documents. DoRA systematically generates synthetic QA training and evaluation datasets with auditable evidence across five domain-specific intents. To mitigate same-pipeline circularity, DoRA's training and test splits use different LLM families (Claude Sonnet for training; GPT-4o for test) drawn from disjoint seed-document corpora. Instantiated on 40 defense-related documents (written in English), DoRA yields ~6.6K curated instances. Compared against 8 LLM baselines over a benchmark of 1,259 samples, a LoRA-adapted Llama3.1-8B trained on the synthetic training set consistently improves performance over 6 coverage and faithfulness metrics, especially reducing hallucination by more than half under the default GTE retrieval setting, with gains persisting across alternative retrievers and prompting-based baselines. Defense-domain expertise is incorporated in three stages of our evaluation: (a) determining the quality of the synthetic QA generated by DoRA, (b) ascertaining the reliability of LLM-as-judge scores, and (c) evaluating the generalization of the QA pipeline on completely human-written QA examples. We position DoRA as a practical framework for specialist-domain RAG under domain shift, with defense as a high-stakes case study.

  • 7 authors
·
May 26

Inspect India Evals: An Open Benchmarking Framework for Evaluating Large Language Models in the Indian Linguistic and Cultural Context

India is a vast nation of over 1.4 billion people, varied by hundreds of diverse and locally specific traditions and cultures and 22 officially recognized languages. Large language models (LLMs) are now being deployed on a massive scale throughout the mainland as well as in remote villages. However, the common benchmarks - MMLU, BIG-Bench, and TruthfulQA are almost exclusively English- and Western-centric. They do not identify those safety, fairness, and accuracy failures unique to the Indian context. That is the gap Inspect India Evals seeks to fill. It is an open-source framework built on top of UK AISI's Inspect AI platform. It has six benchmarks: Multilingual MMLU across sixteen Indian languages, BharatBBQ (our adaptation of BBQ for Indian social bias), a safety evaluation for Digital Public Infrastructure, a multilingual safety test using harmful prompts in Indian languages, a multi-turn jailbreak resistance test, and an Indian cultural knowledge benchmark scored using LLM-as-judge rubrics. In this study, we tested five open-weight models ranging from 8B to 32B parameters. Sarvam-M 24B and Gemma 2 27B came out on top, both scoring 80% on the composite India Fairness Index, with Sarvam-M even beating larger 32B models on Indian cultural knowledge and DPI safety compliance. All models scored 100% refusal on Multilingual Safety, whereas DPI safety varied from 20% to 100%. The framework is public. It's built to work with the UK AISI registry. Anyone can reproduce or extend this work.

  • 5 authors
·
Jul 27

LLM-as-a-Judge Scores Are Unreliable Optimization Signals in Closed-Loop Table Recognition

LLM-as-a-judge is widely used to provide feedback and selection signals in closedloop regeneration, but this use remains insufficiently validated. We study it in table recognition, where deterministic TEDS evaluation provides a controlled testbed, using FinTabNet and OmniDocBench. Three findings emerge. First, judge signals were weak on both datasets: scores frequently tied, rankings were not reproducible, and no tested judge score policy, whether selecting among candidates or accepting revisions under a conservative score margin, improved on the first output on both datasets. Iteration produced better candidates, but the judge recovered them at most partially on one dataset and not at all on the other. Second, severe losses occurred even without specific judge feedback, supporting target-preservation failure under unconstrained regeneration as a proximate mechanism. Third, a structure-preserving instruction reduced the severe-loss rate, significantly on FinTabNet and directionally on OmniDocBench, but produced no improvement, and in an exploratory 2x2 analysis this protection was not stably observed when judge feedback was retained. These results do not dispute the value of LLMs as evaluators, but show that the tested reference-free judge signals were too weak and unstable to drive candidate selection in this setup, and that evaluation-style evidence alone was insufficient to establish closed-loop optimization utility. Iterative refinement requires, at minimum, a verification signal that deterministically detects structural change, rather than judge scores alone.

  • 1 authors
·
Jul 15

DynamicMCPBench: A Trace-Grounded, Effect-Scored Benchmark for LLM Agents over Live MCP Servers

Large language model (LLM) agents are increasingly deployed over Model Context Protocol (MCP) servers, yet the benchmarks used to evaluate them score the final answer or a fixed "ground-truth" list of tools, both of which are fragile once the underlying data is live and stateful. We present DynamicMCPBench, a reusable framework rather than a fixed dataset. A practitioner can run it on their own MCP servers to test models on their own tasks, or let it collect servers automatically to measure a model's general ability to solve agentic tasks. Given the servers and any set of models, it generates realistic goals, pursues each one live to record a successful trajectory, distills that trajectory into path-agnostic effect checkpoints, and scores an agent on whether it reproduces those effects, never on the final answer. To show what the framework reveals, we run it at scale: 24 models over 121 servers and 750 tasks spread evenly over 15 task categories (50 each), where each category targets a distinct tool-use challenge of the generated questions. Each task is scored by pass^3: it counts as solved only if all three independent attempts succeed. Even the strongest agents solve only about half of the tasks, 31% of tasks are solved by no model at all, and accuracy collapses as the required tool chain grows longer (from 39% on the shortest chains to 13% on the longest). A human validation study confirms the automatic scoring is reliable (chance-corrected agreement of 0.76). DynamicMCPBench thus turns benchmark construction into something practitioners can rerun on their own servers and models, while exposing a consistent inability of current agents to handle long, multi-step agentic tasks.

  • 7 authors
·
Aug 30

Right Answer, Wrong Score: Uncovering the Inconsistencies of LLM Evaluation in Multiple-Choice Question Answering

One of the most widely used tasks to evaluate Large Language Models (LLMs) is Multiple-Choice Question Answering (MCQA). While open-ended question answering tasks are more challenging to evaluate, MCQA tasks are, in principle, easier to assess, as the model's answer is thought to be simple to extract and is directly compared to a set of predefined choices. However, recent studies have started to question the reliability of MCQA evaluation, showing that multiple factors can significantly impact the reported performance of LLMs, especially when the model generates free-form text before selecting one of the answer choices. In this work, we shed light on the inconsistencies of MCQA evaluation strategies, which can lead to inaccurate and misleading model comparisons. We systematically analyze whether existing answer extraction methods are aligned with human judgment, and how they are influenced by answer constraints in the prompt across different domains. Our experiments demonstrate that traditional evaluation strategies often underestimate LLM capabilities, while LLM-based answer extractors are prone to systematic errors. Moreover, we reveal a fundamental trade-off between including format constraints in the prompt to simplify answer extraction and allowing models to generate free-form text to improve reasoning. Our findings call for standardized evaluation methodologies and highlight the need for more reliable and consistent MCQA evaluation practices.

  • 6 authors
·
Mar 19, 2025

Uncertainty Quantification for Language Models: A Suite of Black-Box, White-Box, LLM Judge, and Ensemble Scorers

Hallucinations are a persistent problem with Large Language Models (LLMs). As these models become increasingly used in high-stakes domains, such as healthcare and finance, the need for effective hallucination detection is crucial. To this end, we propose a versatile framework for zero-resource hallucination detection that practitioners can apply to real-world use cases. To achieve this, we adapt a variety of existing uncertainty quantification (UQ) techniques, including black-box UQ, white-box UQ, and LLM-as-a-Judge, transforming them as necessary into standardized response-level confidence scores ranging from 0 to 1. To enhance flexibility, we introduce a tunable ensemble approach that incorporates any combination of the individual confidence scores. This approach enables practitioners to optimize the ensemble for a specific use case for improved performance. To streamline implementation, the full suite of scorers is offered in this paper's companion Python toolkit, UQLM. To evaluate the performance of the various scorers, we conduct an extensive set of experiments using several LLM question-answering benchmarks. We find that our tunable ensemble typically surpasses its individual components and outperforms existing hallucination detection methods. Our results demonstrate the benefits of customized hallucination detection strategies for improving the accuracy and reliability of LLMs.

  • 2 authors
·
Apr 27, 2025

Anatomy of LLM Sycophancy: What a Flip Rate Hides

A model under pushback can correct itself, capitulate, or hold, and one flip rate counts a correction and a capitulation alike. Using SycoLens, a modular replay protocol, we test how user pressure and evaluation settings shape measured flip rates. Each measurement is one stateless replay of an item, a committed answer, and one scripted user line in a fixed form. Every effect is read against a matched control with the line deleted. Pushback wording, committed text, answer format, boundary distance, and ground truth become factors of one instrument; earlier instruments vary one to three of them. Across eleven frontier models from three providers and about 760,000 controlled replays, which models look sycophantic depends on how the user pushes back. Lines that assert the opposite verdict and lines that challenge the answer without asserting one rank the models almost unrelatedly. Flip effects grow several-fold near a model's boundary, yet items answered identically in every screening draw still carry about half of the most-affected totals. On arithmetic tasks where the truth is known, one model re-derives and corrects itself under pressure while another abandons correct answers without written work. On the model tested, a planted derivation lowers release of the answer it argues for, true or wrong, where a bare stated value does not; the wrong answer is corrected much more often than the true one is abandoned. Under a yes/no readout the rankings come closer, entangled with a pressure-induced shift toward "no". One score per model therefore compares different behaviours across models and benchmarks. We condense these dependencies into a reporting profile; the instrument, records, and analyses will be released upon publication.

  • 1 authors
·
Oct 4

QuoteBench: How Matched Scores Can Hide Command-Path Failures

LLM coding agents issue Bash commands through interfaces that may serialize, wrap, and reparse model output. Matched execution scores alone cannot distinguish command-generation errors from failures introduced after generation. QuoteBench measures this boundary with exact final-state validation on 56 one-shot tasks from 14 incident-derived families, crossing the generation contract with the execution transport around one deliberately unescaped added parser. Escaping at the interpolation point reproduces each replayed reply's raw-path outcome, so any recovery under a disclosed boundary must come from the model changing its generation. Across eight same-window configurations, replaying the same reply through the added parser lowers success by 55.4 to 73.2 percentage points; disclosure recovers 30.4 to 60.7 points for six configurations, and zero or slightly negative for the other two. Raw generation is nearly saturated at the frontier; boundary adaptation is what still separates models. GPT-5.6-sol's matched gap of -3.6 points hides -64.3 points of damage and +60.7 points of compensation. The deployment configuration reorders models: one reversal among 26 comparable pairs is unambiguous and four more sit on single-task margins. Evaluations of command-issuing agents should report the model configuration, generation contract, execution path, operating point, and final-state validator rather than treat a matched score as an intrinsic model property.

  • 4 authors
·
Aug 12 4

HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animals

Benchmarks for the side effects an agent causes on the way to a goal already exist, but HarvestBench is the first to put a price on avoiding the side effect and to name that side effect as a living creature. It is a farm simulation: LLM sub-agents drive a crew of two tractors through a cooperative corn harvest, with animals in the field. The environment is a reinforcement learning gridworld, every decision is made without memory, and the harm is never named in the goal. When an animal blocks a tractor's route the autopilot stops and asks the model whether to drive on, at no fuel cost, or swerve around it for a posted fuel price. Kills are compared against two controls: rocks, which damage the tractor and are hit under 1% of the time by every model, and hay bales, which are harmless and not alive. Models can also take crops from the neighbor's field instead of their own, a second test of what they treat as moral. Across nine models and 7,201 priced decisions, 3,951 involved an animal rather than a hay bale or a rock. Kill rates range from 0.4% to 98.8%, with Terra and Sol the most merciful and GPT-4o-mini the most cruel, and they are not ordered by capability. Four of six models were sensitive to price at the 5% level, with elasticities from 0.09 to 1.69. All nine drove over wild animals more often than farmed animals on the default map, and the direction held at every map geometry in every model with room to move. The briefing mattered most: under the morality briefing the kill rate was under 6% in five of six reasoning models, and removing it raised the kill rate above 84% in all six. HarvestBench uses no LLM grader. The scorer counts events in the game log, so it is fully reproducible, and it measures what a model will pay to avoid harm rather than what it says about harm.

Conformal Prediction with Paraphrase-Aware Scoring for LLM Uncertainty Quantification

Uncertainty quantification (UQ) for large language models (LLMs) aims to provide reliable measures of predictive confidence, yet current methods are often unstable under meaning-preserving perturbations. Semantically equivalent paraphrases can induce substantial variability in predictive confidence, even for methods with formal guarantees, such as conformal prediction. To address this issue, we propose a paraphrase-aware UQ framework robust to semantic rewordings. Our approach trains a lightweight proxy model on LLM hidden states and aggregates its predictions across paraphrases to construct label-wise nonconformity scores. Under score exchangeability, conformal calibration retains marginal coverage. This guarantee can also hold under test-only rewording, provided that the paraphrase pipeline satisfies an additional distributional alignment condition. We evaluate three settings (normal, fully reworded, and semi-reworded) which apply rewording to neither dataset, both calibration and test datasets, or only the test dataset, respectively. Across seven multiple-choice QA benchmarks and multiple model families, our method produces compact prediction sets with empirical coverage generally near the nominal target, even in the semi-reworded setting. Ablation studies show that the learned proxy accounts for most of the reduction in set size, while paraphrase-augmented training and inference-time aggregation improve stability under rewording. Code is available at https://github.com/Raina-Xin/PA_Score.

  • 6 authors
·
Oct 2

Incompressible Knowledge Probes: Estimating Black-Box LLM Parameter Counts via Factual Capacity

Closed-source frontier labs do not disclose parameter counts. Storing F facts requires at least F/(bits per parameter) weights, so factual recall lower-bounds parameter count--an intrinsic, serving-independent signal, though (as we show) a coarse one. We introduce Incompressible Knowledge Probes (IKPs), a benchmark of 1,400 factual questions spanning 7 tiers of obscurity, designed to isolate knowledge that cannot be derived by reasoning or compressed by architectural improvements. We score with no hallucination penalty (lambda = 0: IKP accuracy is simply the fraction of probed facts answered correctly), which removes both the penalty hyperparameter and the per-tier flooring choice; a full lambda x flooring ablation shows the calibration is robust across scoring choices while individual estimates are not, motivating the no-penalty default. We calibrate a log-linear mapping from IKP accuracy to parameter count on 93 open-weight models (135M-1,600B) spanning 19 vendors, achieving R^2 = 0.910; leave-one-out cross-validation confirms generalization (median fold error 1.48x, 72% within 2x, 86% within 3x). The instrument is deliberately coarse--its 90% prediction interval spans ~3x in either direction, wider than inference economics--so IKP recovers order-of-magnitude effective capacity and relative rankings, not precise parameter counts. For Mixture-of-Experts models, total parameters predict knowledge (R^2 = 0.67) better than active parameters (R^2 = 0.41). We evaluate 201 models from 27 vendors on a curated probe set (1,311 of 1,400 probes surviving name-collision and label-ambiguity filters) and report effective knowledge capacity for all major proprietary frontier models as prediction bands rather than point estimates; for heavily safety-tuned models these are lower bounds, since refusal policy can suppress tens of percentage points of otherwise-answerable capacity.

  • 1 authors
·
Jul 4

From Scores to Skills: A Cognitive Diagnosis Framework for Evaluating Financial Large Language Models

Large Language Models (LLMs) have shown promise for financial applications, yet their suitability for this high-stakes domain remains largely unproven due to inadequacies in existing benchmarks. Existing benchmarks solely rely on score-level evaluation, summarizing performance with a single score that obscures the nuanced understanding of what models truly know and their precise limitations. They also rely on datasets that cover only a narrow subset of financial concepts, while overlooking other essentials for real-world applications. To address these gaps, we introduce FinCDM, the first cognitive diagnosis evaluation framework tailored for financial LLMs, enabling the evaluation of LLMs at the knowledge-skill level, identifying what financial skills and knowledge they have or lack based on their response patterns across skill-tagged tasks, rather than a single aggregated number. We construct CPA-QKA, the first cognitively informed financial evaluation dataset derived from the Certified Public Accountant (CPA) examination, with comprehensive coverage of real-world accounting and financial skills. It is rigorously annotated by domain experts, who author, validate, and annotate questions with high inter-annotator agreement and fine-grained knowledge labels. Our extensive experiments on 30 proprietary, open-source, and domain-specific LLMs show that FinCDM reveals hidden knowledge gaps, identifies under-tested areas such as tax and regulatory reasoning overlooked by traditional benchmarks, and uncovers behavioral clusters among models. FinCDM introduces a new paradigm for financial LLM evaluation by enabling interpretable, skill-aware diagnosis that supports more trustworthy and targeted model development, and all datasets and evaluation scripts will be publicly released to support further research.

  • 11 authors
·
Aug 18, 2025 3

PlannerForge: LLM Agents for Scenario-Based Testing of Motion Planners in Autonomous Driving

Ensuring the safety of autonomous driving is a critical challenge. Scenario-based testing is a systematic process used to validate Autonomous Driving Systems (ADSs), but it remains a fragmented modular pipeline in which scenario generation, retrieval, modification, ADS execution, and results analysis are performed by separate tools with little interaction. Large Language Model (LLM) agents have shown promise across ADS sub-systems such as perception, planning, and control. However, no prior work covers the whole scenario-based testing pipeline for ADSs with a unified LLM-agent framework. We present PlannerForge, an LLM-agent framework that extends all scenario-based testing stages (from Scenario Generation to ADS Assessment) and adds two further LLM-enhanced stages: ADS Enhancement and ADS Benchmarking. We evaluate PlannerForge with 10 off-the-shelf LLMs across all tasks (Generation, Selection, Modification, Module Routing, Planner Testing, and Enhancement) under 5 prompt conditions. Best-per-task scores range from 0.88 to 1.00, and open-source 20-35B backends match commercial APIs on most tasks. Open-source models such as Qwen3.6:35B match commercial APIs on three of the five tasks. Chaining the modules end-to-end retains 83% / 78% of seed queries (commercial / open). It outperforms Scenario Factory 2.0 (Finkeldei et al., 2025) on natural-language generation (193 vs. 144 executable of 200) and realises 92-96% of requested city, road and vehicle attributes. It outperforms BM25 (Robertson and Zaragoza, 2009) at rank 1 selection (92.0% vs. 67.5%) and From-Words-to-Collisions (Gao et al., 2025) on physically valid edits (>=94% vs. 31%). At N=400, cost-tuning lifts planner success from 50.4% to 70.2% and cuts collisions from 19.0% to 8.4%, without domain-specific fine-tuning.

False Floors: LLM Safety Routing Evaluations Break Under Distribution Shift

Safety routers send each request to one of several models and are judged against the best single model. A major routing benchmark picks that comparator on the evaluation data. In the benchmark's own setting this is harmless, but under distribution shift it is not. On HELM Safety the selection cost is 0.003-0.030 of harm under random splits and 0.045-0.113 under held-out categories, comparable to the whole deficit attributed to routing, with its direction holding under either published judge alone. It rises seven- to ninefold on AgentDojo when suites are held out. Across seven safety corpora chosen by rules fixed in advance, three meet a registered interval test and four beat a later permutation null, and three of the four interval misses are corpora where some models have zero observed harm. Prior work proves the direction of this bias. We size it on harm and accuracy, show that it is larger under the held-out splits we measure, and bound it by optimism plus a shift-dependent regret. Scored honestly under shift, routing buys little on these benchmarks. In most pool cells the nested router serves the honest baseline's model, and on the nearly saturated AgentDojo corpus a perfect pre-dispatch router is worth at most two points of harm. We also find a model's expressed recognition of a late injection steerable. On held-out reruns an attacker who knows which model it faces lowers GPT-5.4's judged recognition by 19.6 points, confirmed by an independent label. In an offline counterfactual composition into a controller, the same attack raises or lowers estimated harm depending on the fallback model. Safety routing should be evaluated under shift, against a baseline chosen without the test labels, and recognition-based defences should be scored on harm against an attacker who chooses what the model sees.

  • 2 authors
·
Sep 30

Dissecting Human and LLM Preferences

As a relative quality comparison of model responses, human and Large Language Model (LLM) preferences serve as common alignment goals in model fine-tuning and criteria in evaluation. Yet, these preferences merely reflect broad tendencies, resulting in less explainable and controllable models with potential safety risks. In this work, we dissect the preferences of human and 32 different LLMs to understand their quantitative composition, using annotations from real-world user-model conversations for a fine-grained, scenario-wise analysis. We find that humans are less sensitive to errors, favor responses that support their stances, and show clear dislike when models admit their limits. On the contrary, advanced LLMs like GPT-4-Turbo emphasize correctness, clarity, and harmlessness more. Additionally, LLMs of similar sizes tend to exhibit similar preferences, regardless of their training methods, and fine-tuning for alignment does not significantly alter the preferences of pretrained-only LLMs. Finally, we show that preference-based evaluation can be intentionally manipulated. In both training-free and training-based settings, aligning a model with the preferences of judges boosts scores, while injecting the least preferred properties lowers them. This results in notable score shifts: up to 0.59 on MT-Bench (1-10 scale) and 31.94 on AlpacaEval 2.0 (0-100 scale), highlighting the significant impact of this strategic adaptation. Interactive Demo: https://huggingface.co/spaces/GAIR/Preference-Dissection-Visualization Dataset: https://huggingface.co/datasets/GAIR/preference-dissection Code: https://github.com/GAIR-NLP/Preference-Dissection

  • 6 authors
·
Feb 17, 2024

GEMA-Score: Granular Explainable Multi-Agent Score for Radiology Report Evaluation

Automatic medical report generation supports clinical diagnosis, reduces the workload of radiologists, and holds the promise of improving diagnosis consistency. However, existing evaluation metrics primarily assess the accuracy of key medical information coverage in generated reports compared to human-written reports, while overlooking crucial details such as the location and certainty of reported abnormalities. These limitations hinder the comprehensive assessment of the reliability of generated reports and pose risks in their selection for clinical use. Therefore, we propose a Granular Explainable Multi-Agent Score (GEMA-Score) in this paper, which conducts both objective quantification and subjective evaluation through a large language model-based multi-agent workflow. Our GEMA-Score parses structured reports and employs NER-F1 calculations through interactive exchanges of information among agents to assess disease diagnosis, location, severity, and uncertainty. Additionally, an LLM-based scoring agent evaluates completeness, readability, and clinical terminology while providing explanatory feedback. Extensive experiments validate that GEMA-Score achieves the highest correlation with human expert evaluations on a public dataset, demonstrating its effectiveness in clinical scoring (Kendall coefficient = 0.70 for Rexval dataset and Kendall coefficient = 0.54 for RadEvalX dataset). The anonymous project demo is available at: https://github.com/Zhenxuan-Zhang/GEMA_score.

  • 10 authors
·
Mar 7, 2025

LLM-Based Evaluation of Low-Resource Machine Translation: A Reference-less Dialect Guided Approach with a Refined Sylheti-English Benchmark

Evaluating machine translation (MT) for low-resource languages poses a persistent challenge, primarily due to the limited availability of high quality reference translations. This issue is further exacerbated in languages with multiple dialects, where linguistic diversity and data scarcity hinder robust evaluation. Large Language Models (LLMs) present a promising solution through reference-free evaluation techniques; however, their effectiveness diminishes in the absence of dialect-specific context and tailored guidance. In this work, we propose a comprehensive framework that enhances LLM-based MT evaluation using a dialect guided approach. We extend the ONUBAD dataset by incorporating Sylheti-English sentence pairs, corresponding machine translations, and Direct Assessment (DA) scores annotated by native speakers. To address the vocabulary gap, we augment the tokenizer vocabulary with dialect-specific terms. We further introduce a regression head to enable scalar score prediction and design a dialect-guided (DG) prompting strategy. Our evaluation across multiple LLMs shows that the proposed pipeline consistently outperforms existing methods, achieving the highest gain of +0.1083 in Spearman correlation, along with improvements across other evaluation settings. The dataset and the code are available at https://github.com/180041123-Atiq/MTEonLowResourceLanguage.

  • 3 authors
·
May 18, 2025