Image de couverture de livre

Diagnosing and Mitigating System Bias in Self-Rewarding RL

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Tan, Chuyi, Yuan, Peiwen, Wang, Xinglin, Li, Yiwei, Feng, Shaoxiong, Zhang, Yueqi, Shi, Jiayi, Zhang, Ji, Pan, Boyuan, Hu, Yao, Li, Kan
Format:	Preprint
Publié:	2025
Sujets:	Machine Learning Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Silencer: From Discovery to Mitigation of Self-Bias in LLM-as-Benchmark-Generator
par: Yuan, Peiwen, et autres
Publié: (2025)

Revisiting Self-Consistency from Dynamic Distributional Alignment Perspective on Answer Aggregation
par: Li, Yiwei, et autres
Publié: (2025)

Make Every Penny Count: Difficulty-Adaptive Self-Consistency for Cost-Efficient Reasoning
par: Wang, Xinglin, et autres
Publié: (2024)

UniCBE: An Uniformity-driven Comparing Based Evaluation Framework with Unified Multi-Objective Optimization
par: Yuan, Peiwen, et autres
Publié: (2025)

Beyond One-Size-Fits-All: Tailored Benchmarks for Efficient Evaluation
par: Yuan, Peiwen, et autres
Publié: (2025)

Every Rollout Counts: Optimal Resource Allocation for Efficient Test-Time Scaling
par: Wang, Xinglin, et autres
Publié: (2025)

Mind the Quote: Enabling Quotation-Aware Dialogue in LLMs via Plug-and-Play Modules
par: Zhang, Yueqi, et autres
Publié: (2025)

From Sub-Ability Diagnosis to Human-Aligned Generation: Bridging the Gap for Text Length Control via MARKERGEN
par: Yuan, Peiwen, et autres
Publié: (2025)

LLM-Powered Benchmark Factory: Reliable, Generic, and Efficient
par: Yuan, Peiwen, et autres
Publié: (2025)

PatternKV: Flattening KV Representation Expands Quantization Headroom
par: Zhang, Ji, et autres
Publié: (2025)

Do Not Waste Your Rollouts: Recycling Search Experience for Efficient Test-Time Scaling
par: Wang, Xinglin, et autres
Publié: (2026)

Speculative Decoding for Multi-Sample Inference
par: Li, Yiwei, et autres
Publié: (2025)

Learning More from Less: Unlocking Internal Representations for Benchmark Compression
par: Zhang, Yueqi, et autres
Publié: (2026)

InsBank: Evolving Instruction Subset for Ongoing Alignment
par: Shi, Jiayi, et autres
Publié: (2025)

On Time, Within Budget: Constraint-Driven Online Resource Allocation for Agentic Workflows
par: Wang, Xinglin, et autres
Publié: (2026)

Focused Large Language Models are Stable Many-Shot Learners
par: Yuan, Peiwen, et autres
Publié: (2024)

Integrate the Essence and Eliminate the Dross: Fine-Grained Self-Consistency for Free-Form Language Generation
par: Wang, Xinglin, et autres
Publié: (2024)

Instruction Embedding: Latent Representations of Instructions Towards Task Identification
par: Li, Yiwei, et autres
Publié: (2024)

CogLM: Tracking Cognitive Development of Large Language Models
par: Wang, Xinglin, et autres
Publié: (2024)

Poor-Supervised Evaluation for SuperLLM via Mutual Consistency
par: Yuan, Peiwen, et autres
Publié: (2024)

BatchEval: Towards Human-like Text Evaluation
par: Yuan, Peiwen, et autres
Publié: (2023)

Escape Sky-high Cost: Early-stopping Self-Consistency for Multi-step Reasoning
par: Li, Yiwei, et autres
Publié: (2024)

Generative Dense Retrieval: Memory Can Be a Burden
par: Yuan, Peiwen, et autres
Publié: (2024)

Unveiling and Mitigating Bias in Audio Visual Segmentation
par: Sun, Peiwen, et autres
Publié: (2024)

Stitch and Tell: A Structured Multimodal Data Augmentation Method for Spatial Understanding
par: Yin, Hang, et autres
Publié: (2025)

Bias Fitting to Mitigate Length Bias of Reward Model in RLHF
par: Zhao, Kangwen, et autres
Publié: (2025)

From Self-Evolving Synthetic Data to Verifiable-Reward RL: Post-Training Multi-turn Interactive Tool-Using Agents
par: Gao, Jiaxuan, et autres
Publié: (2026)

SPARD: Self-Paced Curriculum for RL Alignment via Integrating Reward Dynamics and Data Utility
par: Zhi, Xuyang, et autres
Publié: (2026)

Dynamic Stochastic Decoding Strategy for Open-Domain Dialogue Generation
par: Li, Yiwei, et autres
Publié: (2024)

Reward Shaping to Mitigate Reward Hacking in RLHF
par: Fu, Jiayi, et autres
Publié: (2025)

Mitigating Lost in Multi-turn Conversation via Curriculum RL with Verifiable Accuracy and Abstention Rewards
par: Li, Ming, et autres
Publié: (2025)

A Holistic Approach to Understanding and Mitigating the Rising Burden of Maternal Chronic Conditions
par: Binglin Li, et autres
Publié: (2025)

Can Textual Semantics Mitigate Sounding Object Segmentation Preference?
par: Wang, Yaoting, et autres
Publié: (2024)

On Designing Effective RL Reward at Training Time for LLM Reasoning
par: Gao, Jiaxuan, et autres
Publié: (2024)

Debias Can be Unreliable: Mitigating Bias Issue in Evaluating Debiasing Recommendation
par: Wang, Chengbing, et autres
Publié: (2024)

Evolving-RL: End-to-End Optimization of Experience-Driven Self-Evolving Capability within Agents
par: Fan, Zhiyuan, et autres
Publié: (2026)

Mitigating Distribution Shift in Model-based Offline RL via Shifts-aware Reward Learning
par: Luo, Wang, et autres
Publié: (2024)

Benchmarking Bias Mitigation Toward Fairness Without Harm from Vision to LVLMs
par: Tan, Xuwei, et autres
Publié: (2026)

Mitigating Deceptive Alignment via Self-Monitoring
par: Ji, Jiaming, et autres
Publié: (2025)

When Sharpening Becomes Collapse: Sampling Bias and Semantic Coupling in RL with Verifiable Rewards
par: Fan, Mingyuan, et autres
Publié: (2026)