reward-models

Tag

Cards List
#reward-models

CSPF: A Constrained Shared-Private Fusion Method for Non-Verifiable Preference Evaluation

arXiv cs.CL · 2026-07-24 Cached

CSPF proposes a constrained shared-private fusion method to integrate representations from multiple reward models for non-verifiable preference evaluation, outperforming baselines.

0 favorites 0 likes
#reward-models

Reward Models Can Be Too Sensitive (22 minute read)

TLDR AI · 2026-06-29 Cached

This paper argues that reward models in RL are often oversensitive, assigning different scores to equally good responses, and proposes a training-free discretization algorithm using Monte Carlo dropout to reduce oversensitivity, improving policy quality.

0 favorites 0 likes
#reward-models

Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents

Hugging Face Daily Papers · 2026-06-24 Cached

This paper introduces 'progress advantage', an implicit advantage function derived from reinforcement learning post-training that enables effective step-level scoring for LLM agents without requiring dedicated reward model training. It outperforms confidence-based baselines and trained reward models across multiple benchmarks and model families.

0 favorites 0 likes
#reward-models

Discretizing Reward Models

Hugging Face Daily Papers · 2026-06-19 Cached

This paper identifies oversensitivity in continuous reward models for reinforcement learning, where equally good responses receive different scores, and proposes a discretization technique using Monte Carlo dropout to reduce this oversensitivity while maintaining discriminative ability, leading to better policies and less reward hacking.

0 favorites 0 likes
#reward-models

Steerable Cultural Preference Optimization of Reward Models

arXiv cs.CL · 2026-06-18 Cached

Introduces SCPO, a novel reward model training algorithm that incorporates diverse cultural preferences in a balanced manner, achieving up to 7 points improvement and 280% data efficiency over baselines.

0 favorites 0 likes
#reward-models

From Demonstrations to Rewards: Test-Time Prompt Optimization for VLM Reward Models

arXiv cs.LG · 2026-06-02 Cached

Proposes Demo2Reward, a test-time prompt optimization technique for VLM reward models using a few expert demonstrations, significantly reducing false positives and improving policy learning in robotics without additional model training.

0 favorites 0 likes
#reward-models

The Flip Side of RLHF: On-Policy Feedback for Reward Model Self-Supervised Improvement

Hugging Face Daily Papers · 2026-05-29 Cached

The SAVE framework improves reward model training by using value functions to grade on-policy responses and update models through contrastive objectives, achieving outperforming results across six benchmarks.

0 favorites 0 likes
#reward-models

Preference Instability in Reward Models: Detection and Mitigation via Sparse Autoencoders

arXiv cs.LG · 2026-05-19 Cached

This paper investigates preference instability in reward models for LLMs, where subtle input variations cause contradictory preference assignments. The authors propose two SAE-based mitigation strategies—SAE Feature Steering and SAE Residual Correction—to reduce incorrect preference assignments without retraining.

0 favorites 0 likes
#reward-models

Test-Time Personalization: A Diagnostic Framework and Probabilistic Fix for Scaling Failures

arXiv cs.LG · 2026-05-13 Cached

This paper introduces Test-Time Personalization (TTP), a framework that improves LLM personalization by scaling inference-time computation through candidate sampling and reward-based selection. It diagnoses failure modes in standard reward models and proposes a probabilistic personalized reward model to mitigate them.

0 favorites 0 likes
#reward-models

Scaling laws for reward model overoptimization

OpenAI Blog · 2022-10-19 Cached

OpenAI researchers empirically study how reward model overoptimization affects performance, establishing scaling laws that show the relationship between proxy reward optimization and ground truth performance varies by optimization method and scales predictably with model size.

0 favorites 0 likes
← Back to home

Submit Feedback