Reward Model Evaluation

Follow this topic Ask about this topic

Momentum

2 papers in the last four weeks, with none the four weeks before. 0.0% of all new papers.

Jul 13Week of Sep 28

Latest papers 22

Oct 6, 2026cs.CL

Holdout Best-of-N: Unbiased Evaluation and Its Cost

Reusing the scores that select a Best-of-

N

winner can overstate its expected reward. We study evaluation from a fixed matrix of

K

independent scores per candidate for a policy that selects using

J

fresh scores. A single estimator based only on this matrix is exactly unbiased for expected judge reward under every independent, stable collection of candidate-specific score laws if and only if

J<K

, for every pool size

M\ge N\ge2

. At

J=K-1

, the selector deepens as

K

grows. For independent Gaussian scores with common variance and fixed

M\ge N\ge2

, the unbiased minimax risk in this regime is of order

σ^2/\sqrt K

, attained by Holdout; allowing bias improves the rate to

σ^2/K

. For two candidates, we derive the minimum-variance unbiased estimator at known variance and the sharp asymptotic unbiased minimax constant

1/(π\sqrt2)

, which Holdout attains without knowing the variance. The cyclic average over subsets and ties can be computed in

O(MK\log M)

operations. At fixed selector depth, cyclic evaluation of bounded scores has

O(K^{-1})

risk uniformly in pool size. The impossibility result concerns the fixed matrix: one additional fresh winner score permits unbiased evaluation of the all-

K

policy.

Shrey Shah, Yinheng Li

Best-of-N Sampling Policy Evaluation Minimax Estimation Reward Model Evaluation

Sep 21, 2026cs.CL

When Residualization Helps an Audit: Format Effects, Slice Gains, and Their Limits

Evaluation scores used around LLM systems -- including reward models, rerankers, and LLM judges -- can track surface form instead of the quality they claim to measure. When presented with a terse correct solution and a commented buggy solution for the same MBPP problem, a public preference reward model selects the correct one no better than a coin flip (0.507). Subtracting the predictable surface component from such scores is increasingly common, but removal alone does not yield a more valid measurement: the removed component may carry construct-relevant signal, and residualization cannot tell which is which. Under designed interventions -- unit-test labels with comment-only edits -- residualization attenuates the reward model's format effects by about 0.12 on both correct and buggy code, while the correct-versus-buggy margins move by less than 0.01. In observational NLI and QA settings, we freeze a held-out replication before scoring and re-evaluate it using labels from disjoint annotators; this supports only a narrower conclusion: better agreement with the construct labels on a pre-declared slice where a surface-only predictor errs, not a repaired score. Full-population agreement falls in every observational setting with a reported positive slice gain, and within-question ranking falls in every such QA setting. When construct and surface features are entangled, residualization can decorrelate a score while degrading construct alignment, and, in a controlled model, configurations just as damaging to construct alignment pass every pre-adjustment check, so no committed gate is a guarantee. We assemble these distinctions into a reporting protocol whose outcomes, refusal included, state what an adjusted score may be claimed to show: an audit-time diagnostic reported beside the construct-alignment cost it incurs, never a replacement for the raw score.

Daein Weon, Dong Ho Kang

Kookmin University · UStechlab

Language Model Bias Evaluation Language Model Auditing Reward Model Evaluation

Sep 17, 2026cs.CL

F $^{2}$ DR: A Fine-Grained Full-Pipeline Reward Framework for DeepSearch Workflows

With the widespread industrial deployment of Large Language Models (LLMs), DeepSearch has emerged as the dominant paradigm for resolving complex user queries. It typically operates through an iterative closed-loop workflow consisting of planning and reflection, information retrieval, and answer generation. However, existing reward models (RMs) and evaluation benchmarks are primarily designed for static single-turn tasks, failing to capture the full-pipeline complexity of DeepSearch workflows. To address this limitation, we propose F2DR, a fine-grained full-pipeline DeepSearch reward framework. F2DR evaluates DeepSearch workflows across three dimensions: Content, Trajectory, and Answer, enabling comprehensive process-level assessment. We further construct DeepSearch RM-Bench, a dedicated benchmark for evaluating RMs in DeepSearch scenarios. Extensive experiments demonstrate that F2DR achieves significantly higher evaluation consistency than self-evaluation-based baselines, while DeepSearch RM-Bench exhibits strong discriminative capability across existing open-source RMs. We will publicly release the complete DeepSearch RM-Bench dataset soon.

Bojian Xiong, Wentao Ding, Yujing Lu +11

Tianjin University · TJUNLP Lab, Tianjin University, Tianjin, China · Baidu Inc. +1

Reward Modeling LLM Evaluation AI Agent Benchmarks Reward Model Evaluation

Aug 8, 2026cs.LG

Evaluator Ensembles Under Reward Hacking: Covariance Geometry and Finite-Search Guarantees

Language-model judges and reward models enable scalable supervision, but finite optimization can exploit evaluator errors rather than improve response quality. We characterize this failure through the covariance geometry of evaluator ensembles. For calibrated judges, the ensemble mean retains common-mode error along the all-ones direction, whereas cross-judge disagreement captures only orthogonal error. Consequently, disagreement can be high despite robust aggregation, or low while shared response-dependent errors persist. We prove that common-mode error is not identifiable from internal judge scores alone. Under a joint sub-Gaussian model, we bound best-of-K selection overstatement and target-quality regret, extending the guarantees to predictably adaptive search under conditional calibration. The resulting search terms scale as the square root of log K and are asymptotically tight for Gaussian projected errors. We further show that noisy quality proxies introduce artificial rank-one covariance without changing disagreement, and propose a bounded two-anchor Bernstein certificate for finite-search error and regret. Fixed-seed Gaussian stress tests over 120 (J, rho, K) configurations and real-model audits validate the theory while revealing the limits of disagreement-based diagnostics under increasing search pressure.

Fariya Afrin, Ibne Farabi Shihab

Department of Computer Science, Kalinga Institute of Industrial Technology · Department of Computer Science, Iowa State University

Reward Hacking LLM Evaluation LLM-as-a-Judge Best-of-N Sampling Reward Model Evaluation

Jul 30, 2026cs.AI

OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models

Computer-using agents (CUAs) are advancing rapidly across the digital world. A CUA trajectory records the agent's actions, states, and reasoning. Verifying whether it fulfilled the task instruction is central to CUA evaluation, data curation, and reinforcement learning. Neither human-written verifiers nor human annotators can provide such verification at scale, so the field increasingly turns to vision-language models (VLMs) as judges of CUA trajectories. But a fundamental question has long gone unexamined: are these VLM judges reliable enough? To study it systematically, we introduce OSReward, a realistic, high-quality benchmark that evaluates VLM judges on CUA trajectories. The trajectories come from diverse agent backbones executing human-verified instructions across platforms, and are then rigorously labeled with ground-truth verdicts through multi-stage human annotation. Building on it, we derive OSReward-Hard, a challenge set concentrating genuinely hard cases, and OSReward-Multi for fine-grained efficiency and alignment scoring. The most comprehensive evaluation of VLM judges to date finds even state-of-the-art models fall short of an ideal judge, sharing a systematic leniency bias that mislabels failed runs as successes. The few reliable enough to trust are too expensive to run at scale, while affordable open models trail far behind. To close this gap, we construct and release OS-Shepherd-100K, an open corpus of reasoning-annotated trajectory judgments for the CUA community. On it, we train OS-Shepherd (9B and 35B), open reward models that supply low-cost, stable, and reliable reward signals, matching commercial judges at 30-60x lower cost than the frontier. Extensive analyses further inform the design of reliable CUA reward at scale. Our code, benchmark, dataset, and model checkpoints are available at https://os-copilot.github.io/OSReward-Home/.

Qiushi Sun, Kanzhi Cheng, Yian Wang +20

The University of Hong Kong · Nanjing University · National University of Singapore +4

VLM Evaluation Computer-Use Agent Benchmarks Reward Modeling Computer-Use Agents Multimodal Reward Modeling Reward Model Evaluation

Jul 25, 2026cs.AI

SeekJudge: A Practical Reward Framework for Reinforcement Learning in Computer-Use Agents

Deciding whether a trajectory actually fulfills its instruction governs how we measure computer-use agents on long-horizon graphical-user-interface tasks and how we train them with reinforcement learning. This judgment has long relied on rule-based evaluation, which struggles to align with human intention and goes stale when an app updates or its online content drifts. Existing model-based judges attempt to address these problems but still leave a performance gap to the rule-based evaluation. We propose the \textbf{SeekJudge} framework, in which four role-specialized agents, a Condense, a Ground, a Seek and an Analyze agent, reach a verdict through a Seek--Analyze loop over the trajectory. A seed-calibrated distillation pipeline trains one specialized

9

B model to serve as the shared backbone for all four agents. Measured by downstream success rate on held-out RL test goals, SeekJudge is the first practical model-based reward to match or surpass native rule-based supervision in online RL. Beyond accuracy, SeekJudge provides step-level judgments, runs far cheaper than a closed-source large model, and keeps a small per-call context that scales to much longer trajectories. We further contribute a general architectural improvement to the reward server that speeds up judging in RL. Together these make model-based reward a practical drop-in for rule-based supervision in CUA reinforcement learning.

Yang Wan, Zhenhao Zhang, Jierui Wang +1

1Zhejiang University · University of Rochester

Reward Modeling LLM-as-a-Judge RL for GUI Agents Computer-Use Agents Long-Horizon Agent Evaluation Reward Model Evaluation+1

Jun 19, 2026cs.LG

Discretizing Reward Models

Despite their widespread use, the role of reward models in shaping reinforcement learning is poorly understood. Reward models offer a tempting promise: they automatically estimate response quality in the absence of verifiers or human judges. Unlike "verifiable rewards" which typically produce binary scores, reward models typically produce continuous scores, allowing them to be sensitive to fine-grained differences in responses. However, we show this apparent strength is a serious weakness: many popular reward models are oversensitive, assigning different scores to equally good responses. Theoretically, we show that seemingly perfect reward models can be highly oversensitive; empirically, this oversensitivity can lead to bad policies. In place of existing notions of "reward model accuracy," we propose evaluating reward models using distinct measures of "discriminative ability" and "specificity" (the complement of oversensitivity). As a solution, we describe a training-free algorithm that uses Monte Carlo dropout on any neural reward model to produce discrete reward clusters. Theoretically, we prove there exist discretizations that reduce oversensitivity at minimal expense of discriminative ability; empirically we show, in both controlled and natural RL settings, that discretizing rewards leads to less reward hacking and better policies than training on the original rewards.

Vijay Viswanathan, Shiqi Wang, Devamanyu Hazarika +4

1Carnegie Mellon University · 2Meta Superintelligence Labs

Reward Modeling Reinforcement Learning Reward Model Evaluation

Jun 5, 2026cs.CL

Beyond Rubrics: Exploration-Guided Evaluation Skills for Reward Modeling

Open-ended reward modeling requires judges that can follow subtle, domain-specific preferences when verifiable answers are unavailable. Existing rubric-based methods often address this by generating criteria online for each query, but the extra generation step can add inference overhead and produce rigid or misaligned guidance. We introduce Eval-Skill, an exploration-guided method that synthesizes reusable evaluation skills for reward modeling and reframes reward guidance as context evolution rather than parameter training or per-query rubric generation. Using only 100 cases per domain for skill evolution, Eval-Skill synthesizes reusable domain-level evaluation skills through two progressive stages, workflow generation followed by principle generation, with exploration and selection interleaved across both stages. Once generated, a skill is directly injected into the judge context. Across multiple RM benchmarks, Eval-Skill consistently improves diverse judge backbones; on RewardBench 2, it yields significant gains over vanilla judging for each main backbone (+13.44% for Qwen3-8B, and 18.51% for DeepSeek-V4-Flash). Further analyses of evolution-time scaling, generalizability, and transferability show that compact evaluation skills offer an efficient new paradigm for LLM-based evaluation. Code is available at https://github.com/xing-stellus-yue/Eval-Skill.

Xing Yue, Linjuan Wu, Daoxin Zhang +2

Zhejiang University · Xiaohongshu Inc.

Reward Modeling LLM Evaluation LLM-as-a-Judge Reward Model Evaluation

Jun 2, 2026cs.LG

Skill-RM: Unifying Heterogeneous Evaluation Criteria via Agent Skill

Reward models (RMs) provide critical feedback signals for LLM post-training, notably in reinforced fine-tuning (RFT) and reinforcement learning (RL) pipelines. However, current reward evaluation relies on heterogeneous criteria such as rule-based verifiers, ground-truth references, procedural checklists, and complex rubrics, where a unified mechanism to integrate all types of evidence remains unexplored. To this end, we propose Skill Reward Model (Skill-RM), a unified framework that reformulates reward modeling as the execution of a reusable Reward-Evaluation Skill. By treating reward computation as a structured agentic task, Skill-RM provides a consistent interface to orchestrate heterogeneous resources, dynamically selecting and aggregating evidence tailored to the specific requirements of each input. This approach enables the reward model to move beyond static evaluation, ensuring consistency and transparency across diverse tasks. Extensive experiments on reward benchmarks and downstream applications, including best-of-N selection and reinforcement learning, demonstrate that Skill-RM consistently outperforms traditional judge baselines. Our findings suggest that Skill-RM not only provides a unified solution for reward modeling but also achieves superior performance through the strategic and dynamic orchestration of evidence. The code is at https://github.com/Qwen-Applications/Skill-RM.

Tao Chen, Gangwei Jiang, Pengyu Cheng +10

1Qwen Large Model Application Team, Alibaba · 2Sun Yat-sen University · 3The Chinese University of Hong Kong +3

Reward Modeling Reward Model Evaluation

Jun 2, 2026cs.LG

When RLHF Fails: A Mechanistic Taxonomy of Reward Hacking, Collapse, and Evaluator Gaming

RLHF evaluation should track how failures emerge, where they localize, and which warning signals appear before external quality degrades. We study this problem with a compact RLHF pipeline built for this paper, including PPO, DPO, uncertainty-penalized PPO (UP-PPO), reward-model uncertainty, approximate policy drift, diversity and repetition diagnostics, and two external LLM judges. Rather than treating reward hacking as a single terminal event, we classify matched checkpoint and prompt-level transitions by the directions of learned reward R_phi, judge scores R_dag and R2_dag, and their average R_dag. The main empirical findings are that aggressive PPO produces the clearest localized reward-hacking signal, UP-PPO reduces but does not eliminate that signal, row-level diagnostics reveal failures hidden by checkpoint averages, and pre-transition features partially anticipate future localized reward hacking. The central conclusion is methodological: RLHF failures are training dynamics that can be classified, localized, and partially anticipated, not only final-model pathologies. The repository is available at github.com/zabahana/rlhf-failure-modes-diagnostics. The pipeline is also deployed as a live interactive web demo for model comparison and diagnostic views at rlhf-failures.zelalem.ai.

Zelalem Abahana, David Evans, Satish Mahadevan Srinivasan +1

College of Engineering, The Pennsylvania State University, University Park, PA, USA · Department of Computer Science, University of Virginia, Charlottesville, VA, USA · 3Penn State Great Valley, The Pennsylvania State University, Malvern, PA, USA +1

Reward Hacking LLM Evaluation Reinforcement Learning Reward Model Evaluation RL from Human Feedback

Jun 2, 2026cs.LG

HARVE: Hacking-Aware Reward-Head Vector Editing for Robust Reward Models

Reward models are central to large language model (LLM) alignment, but they remain vulnerable to reward hacking. To evaluate reward-model robustness, we introduce RewardHackBench containing 13 reward-hacking patterns covering real life high-stakes domains and general settings, and we find severe failures on specific subcategories across eight reward models. To mitigate these failures, we propose HARVE, a training-free reward-head editing method for scalar reward models. Instead of fine-tuning the reward model, HARVE identifies a multi-directional hacking subspace from residual stream directions associated with selected hacking subcategories, and removes the component of the reward-head vector aligned with that subspace. This directly reduces the reward head's sensitivity to hacking-related features using only a small set of contrastive gold-hacked examples, without gradient updates or fine-tuning. Comprehensive experiments across eight reward models indicates that \model improves hacking robustness, outperforms fine-tuning baselines, and preserves reward-models' general capability. Further analyses suggest that reward hacking is better captured as a multidimensional residual-space structure than by isolated surface cues.

Shuang Liu, Yuxuan Bo, Qiuyang Zhao +4

Carnegie Mellon University · University of Virginia · Harvard University +4

Reward Modeling Reward Hacking Neural Network Robustness Representation Engineering Reward Model Evaluation

May 31, 2026cs.AI

Expected Value Alignment for Generative Reward Modeling in Formal Mathematics Verification

Large Language Models (LLMs) are increasingly used with formal interactive theorem provers such as Lean 4. Scaling these systems with reinforcement learning or search methods requires process reward models (PRMs) that can evaluate intermediate reasoning steps. Existing reward-model designs expose a practical trade-off. Value-head models provide continuous scores but modify the generative model interface, while generative reward models preserve textual rationales but are poorly matched to continuous floating-point regression because numeric values are split across tokens. We introduce Expected Value Alignment (EVA), a reward-modeling procedure that keeps the surface output discrete while extracting continuous scores from the model's token distribution. The model emits integer scores in a structured JSON format, and EVA computes a continuous score as the expectation over the logits of the corresponding anchor tokens. Training combines the causal language modeling objective with an auxiliary mean squared error loss on these expected values. We instantiate EVA in \textit{Leibniz}, a reward model for Lean 4 formal verification, and evaluate it against zero-shot and reward-modeling baselines. The evaluation demonstrates that continuous logit-based scoring significantly reduces discretization artifacts while retaining the interpretability of generative critiques.

Shihao Ji, Haotao Tan, Zihui Song +1

Reward Modeling Automated Theorem Proving Process Reward Models Reward Model Evaluation

May 31, 2026cs.RO

Position: Good Embodied Reward Models Need Bad Behavior Data

This position paper argues that to obtain reliable embodied reward models, the community must invest in ``bad'' robot data: failed, suboptimal, error-prone, and even hazardous behaviors. While reward models are central to any foundation model's lifecycle, today's embodied reward models are trained primarily on successful behaviors. We analyze three state-of-the-art embodied reward models and find that they systematically over-reward behaviors that real human evaluators would penalize, including unsafe interactions, poor execution, and shortcut strategies that only superficially satisfy tasks. We attribute these failures to a key data gap: the scarcity of negative embodied data which is costly to collect and often filtered out or withheld in existing robotics datasets. Furthermore, we show that even modest exposure to real bad behavior data can improve alignment with human preferences and reduce costly false positives. We therefore call on the embodied AI community to curate and release their bad robot data, build synthetic bad data generation engines, develop more decentralized physical evaluation systems, and design benchmarks for fine-grained embodied reward model evaluations.

Ran Tian, Yilin Wu, Andrea Bajcsy

UC Berkeley · NVIDIA Research · Carnegie Mellon University

Reward Modeling Preference Alignment Reward Model Evaluation

May 30, 2026cs.LG

EST-PRM: Stress-Testing Process Reward Models Before They Become Load-Bearing

Process reward models (PRMs) are widely used in language-model training with dense step-level supervision. They assume PRM scores are stable proxies for step correctness under label-preserving transformations. These transformations change reasoning structure but preserve final answers. We argue this assumption is not well validated. Such transformations can change how PRM scores relate to correctness signals, leading to different failure modes across models.To address this gap, we introduce \textbf{EST-PRM}, a stress-testing framework for dense process rewards. It applies three transformations: (1) step inflation, (2) dependency-aware step reordering, and (3) confidence markers. A vulnerability decomposition is defined that separates reward inflation from loss of correctness sensitivity. Five PRM-style models are evaluated on 4,687 reasoning chains from MATH-500, GSM8K, and PRMBench.The results indicate clear differences in vulnerability patterns across models. Math-Shepherd shows the strongest sensitivity to position perturbations, with a Pearson correlation drop of

0.152 \pm 0.038

and a

32.8 \pm 4.9\%

score inflation rate. Qwen2.5-Math-PRM is most affected by step inflation, reaching a

47.6 \pm 4.3\%

inflation rate. Confidence-based perturbations also distort reward calibration, revealing inconsistencies in correctness estimation. Three mitigation strategies are evaluated, highlighting trade-offs between robustness coverage and false-positive rates.

Ibne Farabi Shihab, Fariya Afrin, Sanjeda Akter +1

Department of Computer Science, Iowa State University · Department of Computer Science, Kalinga Institute of Industrial Technology · Department of Civil, Construction & Environmental Engineering, Iowa State University

Process Reward Models Language Model Calibration Language Model Robustness Reward Model Evaluation

May 11, 2026cs.AI

FormalRewardBench: A Benchmark for Formal Theorem Proving Reward Models

Recent neural theorem provers use reinforcement learning with verifiable rewards (RLVR), where proof assistants provide binary correctness signals. While verifiable rewards are cheap and scalable without reward hacking issues, they suffer from sparse credit assignment: models receive no learning signal from difficult problems where partial progress goes unrewarded. This motivates learned reward models that can evaluate proof quality beyond binary verification. However, comparing reward models is challenging since it typically requires expensive RL training ablations. To address this, we introduce \textbf{FormalRewardBench}, the first benchmark for evaluating reward models in formal theorem proving with Lean 4. Our benchmark consists of 250 preference pairs where correct proofs are paired with incorrect variants generated through five expert curated error injection strategies: forced mistakes, minimal single-point variations, verbose incorrect proofs, natural language justification, and Python code injection. We evaluate frontier LLMs (e.g., Claude Opus 4.5), judge LLMs (e.g., CompassJudger-1-14B), general-purpose LLMs (e.g., Qwen2.5-72B-Instruct), and specialized theorem proving models (e.g., DeepSeek-Prover-V2-7B). Our results reveal that frontier LLMs achieve the highest performance (59.8%) while specialized theorem provers perform the worst (24.4%), suggesting that theorem proving ability does not transfer to proof evaluation. We provide further insights on various error injection mechanisms, highlighting the challenging nature of most injection mechanisms. We release \textbf{FormalRewardBench} publicly to encourage more research on developing reward models in formal mathematics.

Zeynel A. Uluşan, Burak S. Akbudak, Can S. Erer +1

Koç University, Department of Computer Science and Engineering · Codeway Studios · Bo˘gaziçi University, Department of Computer Engineering +1

Reward Modeling Automated Theorem Proving Process Reward Models Lean Theorem Proving Reward Model Evaluation

May 6, 2026cs.CL

Misaligned by Reward: Socially Undesirable Preferences in LLMs

Reward models are a key component of large language model alignment, serving as proxies for human preferences during training. However, existing evaluations focus primarily on broad instruction-following benchmarks, providing limited insight into whether these models capture socially desirable preferences. As a result, important failures in social alignment can remain hidden. We extend reward-model benchmarking to four socially consequential domains: bias, safety, morality, and ethical reasoning. We introduce a framework that converts social evaluation datasets into pairwise preference data, leveraging gold labels where available and directional bias indicators otherwise. This enables us to test whether reward models prefer socially undesirable responses, and whether their preferences produce systematically biased distributions over selected outputs. Across five publicly available reward models and two instruction-tuned models used as reward proxies, we find substantial variation across domains, with no single model performing best overall. The models fall well short of strong social intelligence: they often prefer socially undesirable options, and their preferences produce systematically biased distributions. Moreover, stronger bias avoidance can reduce sensitivity to context, revealing a key alignment trade-off between avoiding biased outcomes and preserving contextual faithfulness. These findings show that standard reward benchmarks are insufficient for assessing social alignment and highlight the need for evaluations that directly measure the social preferences encoded in reward models.

Gayane Ghazaryan, Esra Dönmez

Institute for Natural Language Processing, University of Stuttgart · 2Interchange Forum for Reflecting on Intelligent Systems, University of Stuttgart

Reward Modeling Language Model Safety Evaluation Social Bias in Language Models Preference Alignment Reward Model Evaluation

May 6, 2026cs.CL

StoryAlign: Evaluating and Training Reward Models for Story Generation

Story generation aims to automatically produce coherent, structured, and engaging narratives. Although large language models (LLMs) have significantly advanced text generation, stories generated by LLMs still diverge from human-authored works regarding complex narrative structure and human-aligned preferences. A key reason is the absence of effective modeling of human story preferences, which are inherently subjective and under-explored. In this work, we systematically evaluate the modeling of human story preferences and introduce StoryRMB, the first benchmark for assessing reward models on story preferences. StoryRMB contains

1,133

high-quality, human-verified instances, each consisting of a prompt, one chosen story, and three rejected stories. We find existing reward models struggle to select human-preferred stories, with the best model achieving only

66.3\%

accuracy. To address this limitation, we construct roughly

100,000

high-quality story preference pairs across diverse domains and develop StoryReward, an advanced reward model for story preference trained on this dataset. StoryReward achieves state-of-the-art (SoTA) performance on StoryRMB, outperforming much larger models. We also adopt StoryReward in downstream test-time scaling applications for best-of-n (BoN) story selection and find that it generally chooses stories better aligned with human preferences. We will release our dataset, model, and code to facilitate future research. Related code and data are available at https://github.com/THU-KEG/StoryReward.

Haotian Xia, Hao Peng, Yunjia Qi +4

Department of Computer Science and Technology, Tsinghua University

Reward Modeling Language Model Generation Evaluation Human Preference Modeling Text Generation Reward Model Evaluation

May 3, 2026cs.CL

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 Modeling LLM Evaluation Reward Model Evaluation

May 1, 2026cs.SE

Themis: Training Robust Multilingual Code Reward Models for Flexible Multi-Criteria Scoring

Reward models (RMs) have become an indispensable fixture of the language model (LM) post-training playbook, enabling policy alignment and test-time scaling. Research on the application of RMs in code generation, however, has been comparatively sparse, with existing work largely focusing on execution feedback. This choice constrains post-training to optimizing functional correctness over self-contained executable code. In this work, we examine the training and evaluation of multilingual, multi-criteria code RMs. To this end, we first compile Themis-CodeRewardBench, a benchmark to evaluate code RMs across five preference dimensions (i.e., criteria) and eight programming languages, on which we profile 50+ code, math, and general-purpose RMs. Observing the limited proficiency of current RMs beyond scoring for functional correctness, we develop Themis-CodePreference, the largest open-source collection of code preferences to date (more than 350k preference pairs), and use it to train Themis-RM, a suite of multilingual code reward models for flexible multi-criteria scoring, ranging in size from 600M to 32B parameters. Our experiments and ablations demonstrate positive scaling trends, strong cross-lingual transfer when training on diverse preferences, and the importance of multi-criteria training for reliable code reward modeling.

Indraneil Paul, Goran Glavaš, Iryna Gurevych

UKP Lab, TU Darmstadt and National Research Center for Applied Cybersecurity ATHENE · Center for Artificial Intelligence and Data Science, University of Würzburg

Reward Modeling Reward Model Evaluation

Apr 30, 2026cs.CL

From Coarse to Fine: Benchmarking and Reward Modeling for Writing-Centric Generation Tasks

Large language models have achieved remarkable progress in text generation but still struggle with generative writing tasks. In terms of evaluation, existing benchmarks evaluate writing reward models coarsely and fail to measure performance from the perspective of specific requirements. In terms of training, existing training methods either use LLM-as-a-judge approaches or train coarse-grained reward models, lacking fine-grained requirement-adherence reward modeling. To address these issues, we propose a fine-grained evaluation pipeline WEval for writing reward models and a fine-grained reinforcement learning training framework WRL. The evaluation data of WEval covers multiple task categories and requirement types, enabling systematic evaluation of writing reward models by measuring the correlation between the rankings of the reward model and gold rankings. WRL constructs positive and negative samples by selectively dropping instruction requirements, allowing for more precise reward model training. Experiments show that our models achieve substantial improvements across various writing benchmarks and exhibit strong generalization. The code and data are publicly available at https://github.com/Rainier-rq1/From_Coarse_to_Fine.

Qingyu Ren, Tianjun Pan, Xingzhou Chen +1

Shanghai Key Laboratory of Data Science, College of Computer Science and Artificial Intelligence, Fudan University · 2Shanghai Artificial Intelligence Laboratory

Reward Modeling RL for Language Models LLM Evaluation Language Model Generation Evaluation Instruction Following Reward Model Evaluation

Apr 28, 2026cs.LG

When Errors Can Be Beneficial: A Categorization of Imperfect Rewards for Policy Gradient

Training language models via reinforcement learning often relies on imperfect proxy rewards, since ground truth rewards that precisely define the intended behavior are rarely available. Standard metrics for assessing the quality of proxy rewards, such as ranking accuracy, treat incorrect rewards as strictly harmful. In this work, however, we highlight that not all deviations from the ground truth are equal. By theoretically analyzing which outputs attract probability during policy gradient optimization, we categorize reward errors according to their effect on the increase in ground truth reward. The analysis establishes that reward errors, though conventionally viewed as harmful, can also be benign or even beneficial by preventing the policy from stalling around outputs with mediocre ground truth reward. We then present two practical implications of our theory. First, for reinforcement learning from human feedback (RLHF), we develop reward model evaluation metrics that account for the harmfulness of reward errors. Compared to standard ranking accuracy, these metrics typically correlate better with the performance of a language model after RLHF, yet gaps remain in robustly evaluating reward models. Second, we provide insights for reward design in settings with verifiable rewards. A key theme underlying our results is that the effectiveness of a proxy reward function depends heavily on its interaction with the initial policy and learning algorithm.

Shuning Shang, Hubert Strauss, Stanley Wei +2

Princeton Language and Intelligence, Princeton University

Reward Modeling Policy Gradient Policy Learning Reward Design for RL Reward Model Evaluation RL from Human Feedback

Oct 1, 2025cs.AI

Rethinking Reward Models for Multi-Domain Test-Time Scaling

The reliability of large language models (LLMs) during test-time scaling is often assessed with \emph{external verifiers} or \emph{reward models} that distinguish correct reasoning from flawed logic. Prior work has studied both outcome reward models (ORMs), which assess only the final answer, and process reward models (PRMs), which score intermediate reasoning steps. Although PRMs are often viewed as advantageous due to their finer-grained supervision, much of the supporting evidence comes from math-adjacent settings, and their relative benefits across broader domains remain unclear. We present the first unified evaluation of four reward model variants, discriminative ORM and PRM (dORM, dPRM) and generative ORM and PRM (gORM, gPRM), across 14 diverse domains. Contrary to conventional wisdom, we find that (i) dORM performs on par with dPRM, (ii) gPRM is not competitive, and (iii) overall, gORM is the most robust, yielding significant and consistent gains across every tested domain. We attribute the worse performance of gPRM to the stepwise scoring process, which inherits label noise from LLM-based automatic labeling, leading to difficulties in evaluating long reasoning trajectories, including those involving self-correcting reasoning. Both our theoretical analysis and empirical observations indicate that stepwise aggregation compounds errors as reasoning length increases. These findings challenge the common assumption that fine-grained supervision is always better and support generative outcome verification for multi-domain deployment. Our \href{https://github.com/db-Lee/Multi-RM}{\underline{code}} is publicly available to facilitate future research in multi-domain settings.

Dong Bok Lee, Seanie Lee, Sangwoo Park +12

KAIST · TH Nürnberg · Microsoft Research Asia

Reward Modeling LLM Evaluation Process Reward Models Test-Time Scaling LLM Reliability Reward Model Evaluation