REVIEW 20 cited by
Critique-out-Loud Reward Models
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Critique-out-Loud Reward Models
read the original abstract
Traditionally, reward models used for reinforcement learning from human feedback (RLHF) are trained to directly predict preference scores without leveraging the generation capabilities of the underlying large language model (LLM). This limits the capabilities of reward models as they must reason implicitly about the quality of a response, i.e., preference modeling must be performed in a single forward pass through the model. To enable reward models to reason explicitly about the quality of a response, we introduce Critique-out-Loud (CLoud) reward models. CLoud reward models operate by first generating a natural language critique of the assistant's response that is then used to predict a scalar reward for the quality of the response. We demonstrate the success of CLoud reward models for both Llama-3-8B and 70B base models: compared to classic reward models CLoud reward models improve pairwise preference classification accuracy on RewardBench by 4.65 and 5.84 percentage points for the 8B and 70B base models respectively. Furthermore, CLoud reward models lead to a Pareto improvement for win rate on ArenaHard when used as the scoring model for Best-of-N. Finally, we explore how to exploit the dynamic inference compute capabilities of CLoud reward models by performing self-consistency decoding for reward prediction.
Forward citations
Cited by 20 Pith papers
-
Support Vector Rubrics: Closing the Gap Between Self-Generated and Human Rubrics
SVR learns a bank of contrastive rubrics from preference data via max-margin boundaries and prompt-conditioned selection, narrowing the gap to human rubrics on RubricBench from 24.1 to 0.3 points.
-
Will It Go Viral? Grounding Micro-Video Popularity Prediction on the Open Web
WEBSHORTS dataset and SHORTS-CAST framework ground micro-video popularity prediction in structured open-web context collected at upload time and enable selective online adaptation using delayed labels.
-
Think Twice, Act Once: Verifier-Guided Action Selection For Embodied Agents
VeGAS improves MLLM-based embodied agents by sampling action ensembles and using a verifier trained on LLM-synthesized failure cases, yielding up to 36% relative gains on hard multi-object long-horizon tasks in Habita...
-
POSTCONDBENCH: Benchmarking Correctness and Completeness in Formal Postcondition Inference
POSTCONDBENCH is a new multilingual benchmark that evaluates LLM postcondition generation on real code using defect discrimination to assess completeness beyond surface matching.
-
Text-to-Distribution Prediction with Quantile Tokens and Neighbor Context
Quantile tokens inserted into LLM inputs combined with neighbor retrieval enable direct prediction of full distributions, yielding lower MAPE and narrower intervals than baselines on Airbnb and StackSample tasks.
-
LLM-as-a-Coach: Experiential Learning for Non-Verifiable Tasks
Training a language model by distilling a coach's written experiential knowledge beats training on a scalar rubric score for open-ended tasks, with better out-of-distribution transfer.
-
Weak Critics Make Strong Learners: On-Policy Critique Distillation for Scalable Oversight
Weak models used as critics supplying non-misleading revision directions, distilled on-policy via OPCD, improve frozen and trained strong models on reasoning and alignment benchmarks.
-
STRIDE: Learnable Stepwise Language Feedback for LLM Reasoning
STRIDE co-trains generator and verifier on outcome rewards alone to deliver learnable stepwise language feedback that redirects LLM reasoning trajectories and outperforms scalar-reward baselines.
-
Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria
Auto-Rubric as Reward externalizes VLM preferences into structured rubrics and applies Rubric Policy Optimization to create more reliable binary rewards for multimodal generation, outperforming pairwise models on text...
-
Building a Precise Video Language with Human-AI Oversight
CHAI framework pairs AI pre-captions with expert human critiques to produce precise video descriptions, enabling open models to outperform closed ones like Gemini-3.1-Pro and improve fine-grained control in video gene...
-
PaTaRM: Bridging Pairwise and Pointwise Signals via Preference-Aware Task-Adaptive Reward Modeling
PaTaRM converts pairwise preference data into pointwise reward signals via a novel PAR mechanism and task-adaptive rubrics, reporting 8.7% gains on RewardBench/RMBench and 13.6% relative RLHF improvement.
-
RewardBench 2: Advancing Reward Model Evaluation
RewardBench 2 is a new benchmark that supplies challenging fresh human prompts for reward model evaluation, yielding lower average scores but higher correlation with downstream best-of-N sampling and RLHF training per...
-
When Slower Isn't Truer: Inverse Scaling Law of Truthfulness in Multimodal Reasoning
Slower multimodal reasoning models exhibit inverse scaling in truthfulness by fabricating details under ambiguous visual inputs, while faster models remain more cautious via broader inference.
-
Trust Region On-Policy Distillation
TrOPD stabilizes on-policy distillation for LLMs with trust-region learning, outlier estimation, and off-policy guidance, outperforming prior OPD methods on reasoning and code benchmarks.
-
Teaching Large Language Models When Not to Know: Learning Temporal Critique for Ex-Ante Reasoning
TCFT trains LLMs on temporal critique tasks to reduce post-cutoff knowledge leakage by 37-42 percentage points over prompting and standard SFT on Qwen models.
-
Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges
The paper introduces the Proxy Compression Hypothesis as a unifying framework explaining reward hacking in RLHF as an emergent result of compressing high-dimensional human objectives into proxy reward signals under op...
-
Reward Modeling for Reinforcement Learning-Based LLM Reasoning: Design, Challenges, and Evaluation
A taxonomy-driven survey arguing that reward design is the central mechanism shaping reliable LLM reasoning, with maps of reward paradigms, reward-hacking failure modes, and benchmark pitfalls.
-
A Survey on LLM-as-a-Judge
A survey on LLM-as-a-Judge that reviews reliability strategies, proposes evaluation methods, and introduces a novel benchmark for assessing such systems.
-
A Survey of Reinforcement Learning for Large Reasoning Models
A survey compiling RL methods, challenges, data resources, and applications for enhancing reasoning in large language models and large reasoning models since DeepSeek-R1.
-
Reinforcement Learning from Human Feedback
The book introduces the origins, mathematical setup, and optimization stages of RLHF including reward modeling, reinforcement learning, rejection sampling, and direct alignment algorithms.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.