LLM Evaluation

LLM: Large Language Model

Ask about this topic

Overview

The field studies how to measure language-model capabilities, reliability, and risks, and whether those measurements support sound comparisons and deployment decisions. Approaches include curated and synthetic benchmarks, psychometric and adaptive testing, execution-based checks, robustness tests, and human or LLM-based judges. Researchers also examine calibration, uncertainty, contamination, prompt sensitivity, and whether scores reflect the intended capability across languages and domains. A central tension is that efficient aggregate scores can hide important errors and measurement biases, while more realistic, human-grounded evaluations are costly and difficult to standardize.

Latest papers

CardsList

2,446 papers231 papers in the last four weeks, down 17% on the four weeks before. 1.5% of all new papers.

May 4, 2026

Evaluating Prompting and Execution-Based Methods for Deterministic Computation in LLMs

Large Language Models (LLMs) have demonstrated strong capabilities in natural language understanding and reasoning. However, their ability to perform exact, deterministic computation remains unclear. In this work, we systematically evaluate multiple prompting strategies, including Chain-of-Thought (CoT), Least-to-Most decomposition, Program-of-Thought (PoT), and Self-Consistency (SC), on tasks requiring precise and error-free outputs, including binary counting, longest substring detection, and arithmetic evaluation. To support this study, we introduce a synthetic dataset with diverse natural language instructions, enabling controlled evaluation of exact computation across multiple task types. Our results show that standard prompting methods achieve only moderate accuracy on sequence-based tasks. CoT provides limited improvement, while Least-to-Most suffers from error accumulation. In contrast, PoT achieves perfect accuracy by generating executable code and delegating computation to an external interpreter. Self-Consistency improves robustness through majority voting, but incurs substantial computational overhead. We further train a small domain-specific model (CodeT5-small) to generate executable programs, which achieves perfect accuracy on held-out synthetic test data across all tasks with minimal training cost. Overall, our findings suggest that LLMs may simulate reasoning patterns rather than reliably perform exact symbolic computation. For deterministic tasks, combining LLMs with external tools or using specialized models provides a more reliable and efficient solution.

Hongkun Yu

Virginia Tech Blacksburg, VA, USA

LLM EvaluationLLM PromptingLLM Code GenerationPrompt Engineering
AskPDF
May 4, 2026

Benchmarking Local Language Models for Social Robots using Edge Devices

Social-educational robots designed for socially interactive pedagogical support, such as the Robot Study Companion (RSC), rely on responsive, privacy-preserving interaction despite severely limited compute. However, there is a gap in systematic benchmarking of language models for edge computing in pedagogical applications. This paper benchmarks 25 open-source language models for local deployment on edge hardware. We evaluate each model across three dimensions: inference efficiency (tokens per second, energy consumption), general knowledge (a six-category MMLU subset), and teaching effectiveness (LLM-rated pedagogical quality), validated against five independent human raters using the Raspberry Pi(RPi)4 as the primary platform, with additional comparisons on the RPi5 and a laptop GPU. Results reveal pronounced trade-offs: throughput and energy efficiency vary by over an order of magnitude across models, MMLU accuracy ranges from near-random to 57.2%, and teaching effectiveness does not correlate monotonically with either metric. Among the evaluated models, Granite4 Tiny Hybrid (7B) achieves a strong overall balance, reaching 2.5 tokens per second, 0.90 tokens per joule, and 54.6% MMLU accuracy; high MMLU accuracy does not appear necessary for strong teaching scores. Human validation on four representative models preserved the automated rank ordering (Pearson r = 0.967, n = 4). Based on these findings, we propose a three-tier local inference architecture for the RSC that balances responsiveness and accuracy on resource-constrained hardware.

Dorian Lamouille, Matevž B. Zorec, Farnaz Baksh +1

University of Tartu Tartu, Estonia · Institute of Computer Science University of Tartu Tartu, Estonia · Institute of Technology University of Tartu Tartu, Estonia

LLM EvaluationOn-Device Language Model InferenceAI in EducationSmall Language ModelsRoboticsSocial Robotics
AskPDF
May 4, 2026

Evaluating Reasoning Models for Queries with Presuppositions

Millions of users turn to AI models for their information needs. It is conceivable that a large number of user queries contain assumptions that may be factually inaccurate. Prior work notes that large language models (LLMs) often fail to challenge such erroneous assumptions, and can reinforce users' misinformed opinions. However, given the recent advances, especially in model's reasoning capabilities, we revisit whether large reasoning models (LRMs) can reason about the underlying assumptions and respond to user queries appropriately. We construct queries with varying degrees of presuppositions spanning health, science, and general knowledge, and use it to evaluate several widely-deployed models When compared to non-reasoning models, we find that reasoning models achieve a slightly higher accuracy (2-11%), but they still fail to challenge a large fraction (26-42%) of false presuppositions. Further, reasoning models remain susceptible to how strongly the presupposition is expressed.

Rose Sathyanathan, Kinshuk Vasisht, Danish Pruthi

Indian Institute of Science Bengaluru, KA, India

LLM EvaluationLanguage Model RobustnessLarge Reasoning ModelsLanguage Model Reasoning Evaluation
AskPDF
May 4, 2026

SemEval-2026 Task 7: Everyday Knowledge Across Diverse Languages and Cultures

We present our shared task on evaluating the adaptability of LLMs and NLP systems across multiple languages and cultures. The task data consist of an extended version of our manually constructed BLEnD benchmark (Myung et al. 2024), covering more than 30 language-culture pairs, predominantly representing low-resource languages spoken across multiple continents. As the task is designed strictly for evaluation, participants were not permitted to use the data for training, fine-tuning, few-shot learning, or any other form of model modification. Our task includes two tracks: (a) Short-Answer Questions (SAQ) and (b) Multiple-Choice Questions (MCQ). Participants were required to predict labels and were allowed to submit any NLP system and adopt diverse modelling strategies, provided that the benchmark was used solely for evaluation. The task attracted more than 140 registered participants, and we received final submissions from 62 teams, along with 19 system description papers. We report the results and present an analysis of the best-performing systems and the most commonly adopted approaches. Furthermore, we discuss shared insights into open questions and challenges related to evaluation, misalignment, and methodological perspectives on model behaviour in low-resource languages and for under-represented cultures.

Nedjma Ousidhoum, Junho Myung, Carla Perez-Almendros +27

Cardiff University · KAIST · MBZUAI +13

Multilingual Language Model EvaluationLLM EvaluationCross-Cultural Language Model EvaluationCultural Knowledge in Language ModelsLow-Resource Language Processing
AskPDF
May 4, 2026

HalluScan: A Systematic Benchmark for Detecting and Mitigating Hallucinations in Instruction-Following LLMs

Large Language Models (LLMs) have demonstrated remarkable capabilities across diverse natural language processing tasks, yet they remain susceptible to hallucinations -- generating content that is factually incorrect, unfaithful to provided context, or misaligned with user instructions. We present HalluScan, a comprehensive benchmark framework that systematically evaluates hallucination detection and mitigation across 72 configurations spanning 6 detection methods, 4 open-weight model families, and 3 diverse domains. We introduce three key contributions: (1) HalluScore, a novel composite metric that achieves a Pearson correlation of r = 0.41 with human expert judgments; (2) Adaptive Detection Routing (ADR), an intelligent routing algorithm achieving 2.0x cost reduction with only 0.1% AUROC degradation; and (3) systematic error cascade decomposition revealing substantial variation in hallucination error types across domains. Our experiments reveal that NLI Verification achieves the highest overall AUROC of 0.88, while RAV achieves the second-highest AUROC of 0.66.

Ahmed Cherif

1Sofrecom Tunisia, Orange Innovation, Tunis, 1053, Tunisia.

LLM EvaluationLLM Hallucination MitigationHallucination in Language ModelsLLM ReliabilityLLM Hallucination Detection
AskPDF
May 4, 2026

When Correct Isn't Usable: Improving Structured Output Reliability in Small Language Models

Deployed language models must produce outputs that are both correct and format-compliant. We study this structured-output reliability gap using two mathematical benchmarks -- GSM8K and MATH -- as a controlled testbed: ground truth is unambiguous and the output contract is strict (JSON with required fields). We evaluate three 7-9B models under five prompting strategies and report output accuracy -- the joint event of mathematical correctness and valid JSON structure -- as the primary metric. A systematic format failure emerges: NAIVE prompting (no system prompt) achieves up to 85% task accuracy on GSM8K but 0% output accuracy across all models and datasets. REFERENCE prompting (a minimal hand-written JSON format prompt) fares little better, yielding 0% output accuracy for two of four models tested. Constrained decoding enforces syntactic validity but incurs 3.6x-8.2x latency overhead and in several settings degrades task performance substantially. To overcome this limitation, we developed AloLab, an iterative system-prompt optimizer (meta-agent: Claude Sonnet 4.5) requiring only black-box API access to the target model; it reaches 84-87% output accuracy on GSM8K and 34-40% on MATH across five independent runs per model, with 29/30 paired McNemar comparisons against the best static prompt significant at p < 0.05, at near-NAIVE inference latency and without model fine-tuning. The same format failure extends to GPT-4o (OpenAI, 2024), a proprietary closed-source model: REFERENCE achieves 0% output accuracy due to systematic markdown-fence wrapping, while AloLab reaches 95.2% [94.8, 95.6]. An ablation replacing the Sonnet 4.5 meta-agent with Claude 3 Haiku reduces mean output accuracy to 61.0% and increases run-to-run standard deviation from <1 pp to 21.8 pp, confirming that meta-agent capability is a primary driver of optimization quality.

Cosimo Galeone, Minsu Park, Giuseppe Ettorre +1

Alomana, Grottaglie, Italy

LLM EvaluationLLM Input OptimizationLLM PromptingSmall Language ModelsStructured Output Generation
AskPDF
May 4, 2026

Complexity Horizons of Compressed Models in Analog Circuit Analysis

The deployment of Large Language Models (LLMs) for specialized engineering domains, such as circuit analysis, often faces a trade-off between reasoning accuracy and computational efficiency. Traditional evaluation methods treat model performance as a flat metric, failing to account for the hierarchical nature of engineering knowledge. We propose a performance-aware model compression strategy that utilizes prerequisite graphs to optimize model selection for circuit analysis tasks. By structuring electronics design concepts as Directed Acyclic Graphs (DAGs), we can identify the specific complexity horizons of an LLM's compressed variants' tiers. Our framework introduces an agentic pipeline for generating prerequisite-based datasets and a strategic evaluation engine that dynamically cascades queries across a spectrum of compressed variants of an LLM. This approach allows to select the smallest compressed model, given its conceptual knowledge boundaries in circuit analysis. Experimental results on analog electronics datasets demonstrate that prerequisite graphs provide a granular map of model compression with respect to the performance given circuit analysis complexity. (Source Code: https://github.com/pacomesimon/LLM_prereq_graphs_circuit_analysis, Demo: https://huggingface.co/spaces/pacomesimon/LLM_prereq_graphs_circuit_analysis)

Pacome Simon Mbonimpa

Carnegie Mellon University-Africa

LLM EvaluationLLM CompressionLLM Routing
AskPDF
May 4, 2026

Submodular Benchmark Selection

Evaluating large language models across many benchmarks is expensive, yet many benchmarks are highly correlated. We formalize the selection of a small, informative subset as submodular maximization under a multivariate Gaussian model. Entropy (log-determinant covariance) and mutual information between selected and remaining benchmarks arise as natural objectives. Both are submodular; entropy selection coincides with pivoted Cholesky and has spectral residual bounds, while mutual information is non-monotone in general but empirically monotone for small subsets, so we optimize it greedily. Experiments on three matrices from ten public leaderboards show that mutual information selection outperforms entropy for imputation at small subsets.

Alexander Smola

Boson AI 4677 Old Ironsides Dr Santa Clara CA 95054

Submodular OptimizationLLM Evaluation
AskPDF
May 4, 2026

Retrieval and Multi-Hop Reasoning in 1M-Token Context Windows: Evaluating LLMs on Classical Chinese Text

We evaluate the long-context retrieval and reasoning capabilities of five frontier large language models with advertised 1M-token context windows on a classical Chinese corpus. Two complementary studies are reported. Test 1 measures single-needle retrieval at 1M tokens of input, with three biographical needles planted at three depths and pairs of real (training-prior-consistent) and altered (training-prior-contradicting) variants to separate genuine in-context retrieval from reliance on memorised training data. Test 2, a follow-up designed to probe whether long-context capability degrades when retrieval requires intermediate reasoning, measures three-hop chain traversal across three context tiers (256K, 512K, and 1M tokens). We find that single-needle retrieval at 1M is essentially solved for the strongest models - Gemini 3.1 Pro, Claude Opus 4.7, and GPT-5.5 each achieve 100% - but that multi-hop performance reveals three distinct decay signatures: a stable regime (Gemini Pro, Claude) maintaining greater than 80% accuracy through 512K with modest degradation at 1M; a late-cliff regime (GPT-5.5, Qwen3.6-plus) collapsing sharply between 512K and 1M; and a smooth-decline regime (DeepSeek V4 Pro) decaying gradually across the entire range. The findings suggest that nominal context-window length is a poor proxy for usable long-context multi-hop capability, and that the sharpest discriminator between current 1M-context flagships is the 512K-to-1M transition.

Eric H. C. Chow

School of Humanities, The University of Hong Kong

Long-Context RetrievalIn-Context RetrievalLLM EvaluationLong-Context Language Model EvaluationLong-Context Inference
AskPDF
May 4, 2026

CLaC at SemEval-2026 Task 6: Response Clarity Detection in Political Discourse

In this paper, we present our system for SemEval-2026 Task 6 (CLARITY) on response clarity and evasion detection in question-answer pairs from U.S. presidential interviews, comparing fine-tuned encoders with prompt-based LLMs. Our LLM ensemble achieves 80 macro-F1 on the 3-class Task 1 (9th/41) and 59 on the 9-class Task 2 (3rd/33). Across 8 transformer encoders optimized through a four-stage pipeline, partial encoder layer unfreezing outperforms full fine-tuning by a wide margin. Combining English and multilingual encoders further improves ensemble performance over either family alone, despite multilingual models being individually weaker. Prompt-based LLMs, without any task-specific parameter updates, outperform fine-tuned encoders, particularly on minority classes; among open-weight LLMs, parameter count does not predict performance. Enriched input, concatenating the full interviewer turn, improves LLM performance but not that of encoders, an effect that persists with Longformer's extended context window, suggesting the divergence is not attributable to sequence-length capacity alone in our settings. The Clear Reply/Ambivalent boundary remains the dominant failure mode, mirroring the disagreement among human annotators. Our code, prompts, model configurations, and results are publicly available.

Nawar Turk, Lucas Miquet-Westphal, Leila Kosseim

Computational Linguistics at Concordia (CLaC) Lab · Dept. of Computer Science and Software Engineering · Concordia University, Montréal, Québec, Canada

LLM EvaluationLanguage Model Ensembles
AskPDF
May 4, 2026

So Many Opinions, So Many LLMs: Comparing Large Language Models to Traditional Machine Learning for Open- Ended Survey Analysis

Open-ended surveys offer valuable insights, but they are notoriously difficult to analyze at scale. Building on previous work that employed traditional machine learning to classify text ("So Many Responses, So Little Time: A Machine-Learning Approach to Analyzing Open-Ended Survey Data") [1], this study investigates how different large language models (LLMs) understand and analyze NSSE open-ended survey responses. We focus on several cutting-edge LLMSs-OpenAI's GPT series, Twitter-roBERTa-base model, and Meta's LLaMA-and compare their performance to the previous machine learning models in tasks like sentiment analysis and thematic classification. Our research analysis assesses model agreement, classification accuracy, and interpretability of reasoning. The findings reveal that current LLMs routinely beat classic machine learning models in classification accuracy, particularly in understanding complex mood and theme patterns in student replies. While LLMs have superior accuracy, they differ greatly in how explicitly and consistently they justify their predictions and apply category boundaries. These distinctions highlight crucial trade-offs when using LLMs for qualitative analysis: increased predictive strength comes with issues in consistency and explainability. Our findings illustrate the benefits and drawbacks of utilizing various LLMs for large-scale qualitative research, and we provide practical advice for researchers looking to balance automation and interpretive rigor.

Abdullah Akinde, Mariam Akinde, Rasheedat Emiola +1

Austin Peay State University, United States

LLM EvaluationLLM InterpretabilitySentiment AnalysisText Classification
AskPDF
May 3, 2026

EditPropBench: Measuring Factual Edit Propagation in Scientific Manuscripts

Local factual edits in scientific manuscripts often create non-local revision obligations. If a dataset changes from 215 to 80 documents, claims such as 'medium-scale' or 'a few hundred items' may also become stale, even though they do not repeat the edited number. In an audit of recent arXiv cs.CL benchmark and dataset papers, we find fact-dependent qualitative claims in 37.2% of papers, suggesting that this dependency pattern is common in the target genre. We introduce EditPropBench, a benchmark for measuring whether LLM editors propagate factual edits through dependent manuscript claims. Each item contains an ML/NLP-style synthetic manuscript, a targeted edit, and a controlled fact graph with sentence-level labels for direct targets, required downstream updates, and unrelated text that should remain unchanged. We summarize cascade success with Edit-Ripple Adherence (ERA), the fraction of required downstream updates correctly revised, and validate the metric with adversarial probes and stress-test variants. On the hardest cases, where dependent claims use implicit or free-form wording rather than repeating the edited value, five LLM editing systems span ERA 0.148-0.705. Even the strongest misses roughly 30% of required cascade updates. This advantage persists in a mixed evaluation that includes easy cases solvable by deterministic substitution. EditPropBench shows that current LLM editors can repair many implicit consequences of factual edits, but reliable scientific revision still requires cascade-aware checking.

Garvin Kruthof

Technical University of Munich

Knowledge EditingLLM EvaluationFactual Consistency EvaluationLanguage Model Editing
AskPDF
May 3, 2026

What Single-Prompt Accuracy Misses: A Multi-Variant Reliability Audit of Language Models

Single-prompt accuracy is the dominant way to benchmark language models, but it can miss reliability failures that matter. We evaluate a 15-model open-weight corpus, with the main reliability analyses focused on 10 instruct models across five classification and reasoning benchmarks under five prompt variants each, measuring accuracy, token-probability calibration, verbal-confidence calibration, verbal parse rate, and prompt-perturbation spread for every (model x dataset x variant) cell. We find three broad results. First, evaluation design can materially change the conclusion. Switching Expected Calibration Error (ECE) token from a raw to a label-set-normalised definition changes per-cell calibration by a mean absolute 0.149. More strikingly, pairing a chain-of-thought prompt with a first-character evaluator on ARC-Challenge reduces apparent accuracy by 72-88% across all five primary models; two independent repair procedures recover 93.8% and 102.7% of the lost performance, indicating an evaluator-side rather than model-side failure. Second, confidence signals are fragile. On MMLU-Pro, every primary model verbally reports confidence substantially above both its accuracy and its token-probability confidence on the same rows, and verbal parse rate can collapse for a single model on a single prompt variant. Third, prompt robustness does not track parameter count reliably. Across 10 instruct models, the correlation between model size and prompt-perturbation spread ranges from -0.244 to 0.474 across benchmarks. Taken together, these results show that reliability conclusions for small language models depend not only on the model being evaluated, but also on the evaluation pipeline used to measure it. We argue that calibration definitions, evaluator logic, verbal parseability, and prompt robustness should be reported explicitly when making reliability claims.

Ranit Karmakar, Jayita Chatterjee

Harvard University, Boston, MA 02115 · Boston, MA 02134

LLM EvaluationLanguage Model CalibrationLLM ReliabilityBenchmark ValidityLanguage Model RobustnessPrompt Robustness
AskPDF
May 3, 2026

Conventional Commit Classification using Large Language Models and Prompt Engineering

Conventional commits provide a structured format for writing commit messages, which improves readability, software maintenance, and enables automation tools such as changelog generators and semantic versioning systems. Existing approaches to conventional commit classification typically rely on ML/DL models trained on large labeled datasets. In this paper, we investigated a training-free alternative by leveraging large language models (LLMs) through prompt engineering. Rather than building a task-specific classifier, we evaluate three prompting strategies, such as zero-shot, few-shot, and chain-of-thought, across three open-source LLMs of varying scale: Mistral-7B-Instruct, LLaMA-3-8B, and DeepSeek-R1-32B. Classification is performed directly on code diffs extracted from a balanced dataset of 3,200 commits mined from the InfluxDB repository, without any model fine-tuning. Our results show that few-shot prompting consistently achieves the highest accuracy, while chain-of-thought prompting does not yield additional gains for this classification task. Among the evaluated models, DeepSeek-R1-32B achieves the strongest overall performance, suggesting that model scale plays a meaningful role in conventional commit classification. These findings provide practical guidance for researchers and practitioners seeking to automate commit classification without the overhead of curating and maintaining labeled training data.

H. M. Sazzad Quadir, Sakib Al Hasan, Md. Nurul Ahad Tawhid

Institute of Information Technology University of Dhaka Dhaka, Bangladesh

Software EngineeringLLM EvaluationLLM Prompting
AskPDF
May 3, 2026

Language models fail at extended rule following

Large language models are highly capable of answering difficult questions by retrieving, recombining, and attending to information in long contexts. For agentic tasks, an additional capability is required: the preservation of an exact state while repeatedly applying rules. We find that this reliability is absent across language models. To demonstrate, we query 126 leading model variants with the task of counting a long string of repeated characters, and we find they all cannot accurately count above a model-dependent, syntax-sensitive counting capacity threshold. Failures are abrupt and persist even with increasing model size, inference time computation, and external tool. Mechanistic probing indicates that models use a finite number of internal states to mimic counting as a rule and fail once these states are exhausted. Furthermore, such states are the basis for performing complex tasks beyond counting. These results indicate that fundamentally new model architectures are required for autonomous agents to achieve truly reliable rule following capabilities.

Tianxiang Dai, Jonathan Fan

Department of Electrical Engineering, Stanford University, Stanford, CA 94305.

LLM EvaluationLLM ReliabilityState TrackingAlgorithmic Reasoning
AskPDF
May 3, 2026

StressEval: Failure-Driven Dynamic Benchmarking for Knowledge-Intensive Reasoning in Large Language Models

Static benchmarks for LLMs are increasingly compromised by contamination and overfitting especially on knowledge intensive reasoning tasks While recent dynamic benchmarks can alleviate staleness they often increase difficulty at the expense of answerability and controllability In this paper we propose StressEval a failure driven data synthesis framework that turns observed model failures into dynamic challenging and controllable test instances StressEval consists of three stages first it constructs a semi structured difficulty card that identifies the failed reasoning step and its root cause second it applies a dual perspective instance synthesis method that targets both knowledge gaps and reasoning breakdowns while preserving the underlying difficulty factors and third it applies a gating mechanism to retain only grounded unambiguous instances Seeding from multiple knowledge intensive reasoning datasets we employ StressEval to build Dynamic OneEval a focused suite of challenging dynamic benchmark Across several state of the art LLMs Dynamic OneEval yields substantially larger performance drops than the original benchmarks while retaining explicit difficulty factors enabling more actionable iteration

Yongrui Chen, Yangyang Ma, Xiaoying Huang +4

Southeast University · Hangzhou Dianzi University · Zhejiang University +1

LLM EvaluationBenchmark DesignSynthetic Benchmark Generation
AskPDF
May 3, 2026

Spatiotemporal Hidden-State Dynamics as a Signature of Internal Reasoning in Large Language Models

Large reasoning models (LRMs) generate extended solutions, yet it remains unclear whether these traces reflect substantive internal computation or merely verbosity and overthinking. Although recent hidden-state analyses suggest that internal representations carry correctness-related signals, their coarse aggregations may obscure the token and layer structure underlying reasoning computation. We investigate hidden-state transitions across decoding steps and layers, and identify a distinct spatiotemporal pattern in LRMs: successful trajectories exhibit broad temporal dynamics with localized layer-wise concentration, while this structure is weaker in non-reasoning models and knowledge-heavy domains. We formalize this characteristic as Spatiotemporal Amplitude of Latent Transition (StALT), a training-free trajectory statistic that summarizes temporal changes between adjacent tokens weighted by within-token layer saliency. Across diverse models and benchmarks, StALT reliably separates correct from incorrect trajectories in reasoning-intensive regimes, providing a competitive label-free correctness signal alongside strong output-space and length-based baselines. Intervention analyses further show that this spatiotemporal amplitude responds systematically to manipulations that increase or reduce the demand for internal reasoning, supporting its association with latent reasoning dynamics in LRMs. These findings provide empirical evidence that LRMs exhibit measurable hidden-state dynamics and offer a practical probe for understanding internal computation beyond output-based evaluation.

Kotaro Furuya, Takahito Tanimura

Research and Development Group, Hitachi, Ltd. 1-280 Higashi-Koigakubo, Kokubunji-shi, Tokyo 185-8601, Japan

Transformer InterpretabilityLLM EvaluationLLM InterpretabilityLarge Reasoning Models
AskPDF
May 3, 2026

Prompt Framing Distorts Count-Based Evaluation of LLM Error Detection: Evidence from Numeric Anchoring

Count-based F1 is widely used as a proxy for LLM error-detection quality, but this paper shows that it can rise dramatically without a corresponding improvement in span localization, a gap termed F1 Inflation. The paper introduces ErrorBench, a controlled stress-test protocol for prompt-induced count distortion. ErrorBench evaluates six contemporary LLMs under five prompt conditions over 4,290 responses from 143 CoNLL-2014 passages. Under CoNLL-2014 M2-style scoring, anchored prompts produce up to 0.79 points of F1 Inflation, and up to 0.96 under strict matching. A 100-passage replication using the official ERRANT 3.0.0 pipeline and multi-reference scoring reproduces the pattern: averaged over six models, the Blind-to-Anchored prompt shift raises Count-F1 by +0.21 while raising multi-reference ERRANT F0.5 by only +0.04. The study finds larger count responses in highly instruction-compliant GPT/Claude systems and smaller responses in the Gemini family under this stress-test protocol. The findings suggest that LLM proofreading and document-review evaluations should avoid pre-populated error counts and should report span-aware metrics alongside count-based metrics.

Dekun Yang

Zhejiang University

Prompt SensitivityLLM EvaluationLanguage Model Error Detection
AskPDF
May 3, 2026

RMGAP: Benchmarking the Generalization of Reward Models across Diverse Preferences

Reinforcement Learning from Human Feedback has become the standard paradigm for language model alignment, where reward models directly determine alignment effectiveness. In this work, we focus on how to evaluate the generalizability of reward models. By "generalizability", we mean the ability of RMs to correctly rank responses to align with diverse user preferences. However, existing reward model benchmarks are typically designed around a universal preference, failing to assess this generalization. To address this critical gap, we introduce RMGAP, a benchmark comprising 1,097 instances across Chat, Writing, Reasoning, and Safety domains. Since different users exhibit diverse preferences for the same task, we first generate four distinct responses with different linguistic profiles for each collected prompt. However, the original prompt set lacks the specificity to convey different preferences. We therefore construct tailored prompts by contrasting these candidates and designing scenarios in which one response becomes the uniquely appropriate choice. Moreover, we observe that users often express the same preference using different phrasings, and thus extend each prompt with two paraphrased variants. Our evaluation of 24 state-of-the-art RMs reveals their substantial limitations: even the best RM achieves only 49.27% Best-of-N accuracy, highlighting considerable room for improvement in reward model generalization. Related data and code are available at https://github.com/nanzhi84/RMGAP.

Yangyang Zhou, Yi-Chen Li

Beijing University of Posts and Telecommunications · National Key Laboratory for Novel Software Technology, Nanjing University, China · School of Artificial Intelligence, Nanjing University, China

Reward ModelingLLM EvaluationPreference ModelingReward Model Evaluation
AskPDF
May 3, 2026

Molecular Representations for Large Language Models

Large Language Models (LLMs) are increasingly being used to support scientific discovery. In chemistry, tasks such as reaction prediction and structure elucidation require reasoning about the structures of molecules. As such, LLM-based systems for chemistry must interact reliably with molecular structures. Most previous studies of LLMs in chemistry have used SMILES strings or IUPAC names as molecular representations; however, the suitability of these formats has not been systematically assessed. In this work, we introduce MolJSON, a novel molecular representation for LLMs, and systematically compare it with five common chemical formats. We evaluated each representation with GPT-5-nano, GPT-5-mini, GPT-5, and Claude Haiku 4.5 using a set of 78,045 questions spanning translation, shortest path, and constrained generation reasoning tasks. We observed substantial variation across representations in the ability of LLMs to interpret and generate molecular graphs, with MolJSON consistently outperforming existing formats. On translation tasks, GPT-5 achieved 71.0% accuracy when converting IUPAC names to MolJSON, compared with 43.7% when converting the same inputs to SMILES. For constrained generation, GPT-5 reached 95.3% accuracy generating MolJSON, compared with 76.3% for IUPAC and 64.0% for SMILES. As an input format for shortest-path reasoning, GPT-5 successfully answered 98.5% of questions with MolJSON, compared with 92.2% for SMILES and 82.7% for IUPAC, whilst also using fewer reasoning tokens. We observed systematic errors associated with atom count and ring complexity for SMILES strings and IUPAC names, whereas MolJSON was more robust to these failure modes. Our results show that the choice of molecular representation has a material impact on LLM performance, and that explicit molecular graph schemas, such as MolJSON, are a promising direction for LLM-based systems in chemistry.

Nicholas T. Runcie, Fergus Imrie, Charlotte M. Deane

Department of Statistics University of Oxford Oxford, UK

LLM EvaluationLarge Language Model-Guided Scientific DiscoveryMolecular Representation Learning
AskPDF
May 2, 2026

Format Sensitivity Index: Token-Controlled Prompt Wrapper Robustness and Schema Compliance in LLM Benchmarking

Prompt wrappers often differ only in formatting, yet they can change model scores enough to flip leaderboard conclusions. We study this variance under a token-controlled protocol and introduce two complementary metrics: the Format Sensitivity Index (FSI), the accuracy range induced by wrapper choice, and the Parseability Sensitivity Index (PSI), the corresponding range in answer parseability. Across 140,000 OpenRouter generations spanning 7 QA tasks, 5 wrapper families, and 4 instruct models from 7B to 72B parameters, we find that mean FSI varies by over 30x across models and is largely explained by compliance failures. A fixed-effects regression shows that parseability remains a strong predictor of accuracy even after controlling for task, model, and wrapper. We argue that reporting accuracy without wrapper variance and compliance is statistically fragile, and we give practical recommendations for both benchmarking and structured-output deployments.

Deep Pankajbhai Mehta

Adobe Inc.

Prompt SensitivityLLM EvaluationBenchmark ValidityStructured Output GenerationLanguage Model RobustnessSensitivity Analysis
AskPDF
May 2, 2026

Prosa: Rubric-Based Evaluation of LLMs on Real User Chats in Brazilian Portuguese

Rankings produced by holistic LLM-as-a-judge scoring are sensitive to the bias of the chosen judge model. We show that switching to binary rubric scoring with multi-judge filtering removes this sensitivity: decomposing the judgement matters more than the judge model itself. To support this claim, we introduce Prosa, the first real user multi-turn Brazilian Portuguese chat benchmark: 1,000 WildChat conversations scored by three judges from three model families on 16 models. Under filtered rubric scoring the three judges agree on every one of the 16 ranks, whereas under holistic scoring they agree on only 7 of 16. Additionally, the rubric filtering pipeline increases the average score gap between neighbouring models by 47%, thereby improving Prosa's discriminative power. Evaluating a new model on Prosa costs approximately $2.1 when using Gemini 3 Flash as the judge. We release the benchmark and the filtering code to ensure that future models can be assessed under identical conditions. These artifacts also make our rubric-based scoring method reusable beyond Prosa, supporting other open-ended evaluation settings.

Roseval Malaquias Junior, Giovana Kerche Bonás, Thales Sales Almeida +6

Maritaca AI · Jusbrasil

LLM EvaluationLLM-as-a-JudgeLanguage Model Generation EvaluationMulti-Turn Dialogue EvaluationRubric-Based EvaluationBias in LLM-as-a-Judge
AskPDF
May 2, 2026

Evaluating LLMs on Large-Scale Graph Property Estimation via Random Walks

With the rapidly improving reasoning abilities of Large Language Models (LLMs), there is also a rising demand to use them in a wide variety of domains. This brings about the need to carefully evaluate the limits of the capabilities of these models with various tests and benchmarks. Graph structures are ubiquitous in real-world data, and are often used to represent and analyze relationship patterns within data. Many benchmarks have already been proposed in the graph literature to test the reasoning ability of LLMs to follow and execute graph algorithms. However, due to the limited context length of LLMs, these benchmarks consist of very small graphs. In real-world data, the size of graphs can be significantly larger, and in many cases, not fully accessible. In this paper, we examine a class of problems that arises with very large graphs having limited accessibility. We propose a large graph benchmark dataset, EstGraph, and introduce four distinct tasks designed to estimate large graph properties. We evaluate the reasoning abilities of LLMs on these tasks using a wide variety of graph datasets. In addition, we provide task-specific prompt constructions based on random walk sampling of large graphs (up to millions of nodes) that effectively convey sufficient information to LLMs within the limits of context length.

Sunil Kumar Maurya, Xin Liu

University of Tokyo Tokyo, Japan · AIST Tokyo, Japan

LLM EvaluationLLM Reasoning with GraphsGraph ReasoningLLM Reasoning
AskPDF
May 2, 2026

Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks

Evaluating large language models (LLMs) for medical applications remains challenging due to benchmark saturation, limited data accessibility, and insufficient coverage of relevant tasks. Existing suites have either saturated, heavily depend on restricted datasets, or lack comprehensive model coverage. We introduce Medmarks, a fully open-source evaluation suite with 30 benchmarks spanning question answering, information extraction, medical calculations, and open-ended clinical reasoning. We perform a systematic evaluation of 61 models across 71 configurations using verifiable metrics and LLM-as-a-Judge. Our results show that frontier reasoning models (Gemini 3 Pro Preview, GPT-5.1, & GPT-5.2) achieve the highest performance across both benchmarks, most frontier proprietary models are significantly more token efficient than open-weight alternatives, medically fine-tuned models outperform their generalist counterparts, and that models are susceptible to answer-order bias (particularly smaller models and Grok 4). A subset of our evals (Medmarks-T) can be directly used as reinforcement learning environments to post-train LLMs for medical reasoning. Code is available at https://github.com/MedARC-AI/Medmarks

Benjamin Warner, Ratna Sagari Grandhi, Max Kieffer +32

Sophont · MedARC · Prime Intellect

LLM EvaluationHealthcareClinical Language Model EvaluationMedical LLMs
AskPDF
May 2, 2026

Structural Ranking of the Cognitive Plausibility of Computational Models of Analogy and Metaphors with the Minimal Cognitive Grid

In this paper, we employ the Minimal Cognitive Grid (MCG), a framework created to evaluate the cognitive plausibility of artificial systems, to offer a systematic assessment of leading computational models of analogy and metaphor, including the Structure-Mapping Engine (SME), CogSketch, METCL, and Large Language Models (LLMs). We present a formal and quantitative operationalization of the MCG framework and, through the analysis of its three main dimensions (Functional/Structural Ratio, Generality, and Performance Match), examine how well each system aligns with standard cognitive theories of the modeled phenomena, thus allowing for comparison of the models with respect to their cognitive plausibility, according to consistent and generalizable mathematical criteria.

Alessio Donvito, Antonio Lieto

University of Bari, Italy · University of Salerno, CIIT Lab / ICAR-CNR, Italy

LLM EvaluationComputational Cognitive ModelingAnalogical Reasoning
AskPDF
May 2, 2026

On Stable Long-Form Generation: Benchmarking and Mitigating Length Volatility

Large Language Models (LLMs) excel at long-context understanding but exhibit significant limitations in long-form generation. Existing studies primarily focus on single-generation quality, generally overlooking the volatility of the output. This volatility not only leads to significant computational costs but also severely impacts the models' reliable application. To address this gap, our work unfolds in three stages: benchmarking, probing, and mitigation. We first propose the VOlatility in Long-form Text Benchmark (VOLTBench), a novel heterogeneous-task benchmark designed to systematically quantify the length volatility of long-form generation. Subsequently, by analyzing attention traces, we conduct an in-depth probe to identify several common internal patterns that cause this volatility. Finally, to mitigate long-form output volatility, we propose Stable Generation via Logits Boosting (GLoBo), a lightweight decoding-stage optimization strategy, designed to significantly enhance both the length accuracy and stability of long-form generation without additional training. Extensive experiments on VOLTBench provide the first systematic confirmation of severe long-form output instability in mainstream models and validate that our proposed method successfully improves the mean output length of the base model by 148% and reduces the length volatility by 69%, while maintaining high generation quality.

Zhitao He, Haolin Yang, Rui Min +2

Hong Kong University of Science and Technology

LLM EvaluationLong-Form Document GenerationInference-Time OptimizationText Generation
AskPDF
May 2, 2026

The Partial Testimony of Logs: Evaluation of Language Model Generation under Confounded Model Choice

Offline evaluation of language models from usage logs is biased when model choice is confounded: the same user-side factors that influence which model is used can also influence how its output is judged, so raw comparisons of logged scores mix self-selected populations rather than estimating a common quantity of interest. A small randomized experiment can break this bias by overriding model choice, but in practice such experiments are scarce and costly. We study a three-source design that combines a large confounded observational log (OBS) for scale, a small randomized experiment (EXP) for unconfounded scoring, and an offline simulator (SIM) that replays candidate models on cached contexts. Our main result is an identification theorem showing that the randomized experiment and the simulator are together enough to recover causal model values; the observational log enters only afterward, to reduce estimation error rather than to make the causal comparison valid. Six estimator families are evaluated in a controlled semi-synthetic validation and in two real-task cached benchmarks for summarization and coding. No family dominates every regime; relative performance depends on the amount of unbiased EXP supervision and on how closely the target reward aligns with OBS-derived structure.

Jikai Jin, Vasilis Syrgkanis

LLM EvaluationOff-Policy EvaluationLanguage Model Generation Evaluation
AskPDF
May 1, 2026

Psychologically Potent, Computationally Invisible: LLMs Generate Social-Comparison-Eliciting Posts They Fail to Detect

We introduce Xiaohongshu Social Comparison Reader Elicitation (XHS-SCoRE), a reader-grounded benchmark for detecting whether text-only Xiaohongshu (RedNote) posts elicit Upward, Downward, or Neutral/no clear social comparison from a first-person reader perspective. The task targets a socially meaningful relational, behaviorally real signal not reducible to sentiment. Across prompted LLM classifiers and supervised Chinese encoders, we find a consistent generation--detection mismatch: the signal is textually learnable in-domain, but not robustly accessible to prompt-based classification. Prompted LLM classifiers show stable failures, especially neutralization of comparison-eliciting posts and model-specific directional skew. A controlled pilot shows that LLM-generated Xiaohongshu-style posts can shift perceived standing and comparison-related affect even when prompt-based detection of the same construct remains fragile. XHS-SCoRE contributes a benchmark for reader-grounded comparison detection and a diagnostic framework for studying when socially meaningful relational cues remain only partially visible to prompt-based inference.

Hua Zhao, Jiapei Gu, Michelle Mingyue Gu

Department of English Language Education, The Education University of Hong Kong · Analytics/Assessment Research Centre, The Education University of Hong Kong

Social Media AnalysisLLM EvaluationText Classification
AskPDF
May 1, 2026

Can AI Debias the News? LLM Interventions Improve Cross-Partisan Receptivity but LLMs Overestimate Their Own Effectiveness

Partisan news media erode cross-partisan trust, but large language models (LLMs) offer a potential means of debiasing such content at scale. Across two pre-registered experiments, we tested whether LLM-generated debiasing of liberal news headlines could improve conservative readers' trust-relevant judgments. Study 1 found that subtle lexical debiasing (replacing emotive words with more moderate synonyms) had no effect on any outcome. Study 2 found that a more substantive reframing intervention significantly increased conservatives' perceived trustworthiness, completeness, and willingness to engage with liberal news headlines, without producing a backfire effect among a sample of liberals. In Study 1, the intervention produced robust effects among LLM-simulated silicon participants, whereas it had no impact on human readers. In Study 2, the intervention's effects among silicon participants aligned directionally with human responses but were significantly larger in magnitude for some outcomes. Moderation analyses revealed that the model's implicit theory of who responds to debiasing diverged from the psychological profile that actually predicted human responsiveness. These findings demonstrate that LLM-based debiasing can improve cross-partisan receptivity when targeting ideological framing rather than surface-level language, but that current models lack both the quantitative accuracy and qualitative psychological fidelity to evaluate their own interventions without human oversight.

Faisal Feroz, Jonas R. Kunst

Department of Experimental Psychology, University of Oxford · Uehiro Oxford Institute, University of Oxford · Psychology Programme, School of Social Sciences, Nanyang Technological University +1

LLM EvaluationHuman Behavior SimulationPolitical Bias in Language Models
AskPDF
May 1, 2026

When LLMs Stop Following Steps: A Diagnostic Study of Procedural Execution in Language Models

Large language models (LLMs) often achieve strong performance on reasoning benchmarks, but final-answer accuracy alone does not show whether they faithfully execute the procedure specified in a prompt. We introduce a controlled diagnostic benchmark for procedural execution, where models are given a step-wise arithmetic procedure and two numeric inputs, and must return the final computed value. Complexity is varied through procedure length and look-back dependencies over intermediate variables. Average first-answer accuracy drops from 63% on 5-step procedures to 20% on 95-step procedures. Generation-level analysis shows that failures often involve missing answers, premature answers, self-correction after an initial error and under-executed traces. These findings suggest that apparent reasoning ability can mask substantial weaknesses in faithful long-horizon procedural execution.

Sailesh Panda, Pritam Kadasi, Abhishek Upperwal +1

Indian Institute of Technology Gandhinagar · Soket AI

LLM EvaluationInstruction FollowingLLM Reasoning
AskPDF