paper-with-me

홈 › Papers

Temporal Self-Rewarding Language Models: Decoupling Chosen-Rejected via Past-Future

2025-08-08 · Yidong Wang, Xin Wang, Cunxiang Wang, Junfeng Fang, Qiufeng Wang, Jianing Chu, Xuran Meng, Shuxun Yang, Libo Qin, Yue Zhang, Wei Ye, Shikun Zhang arxiv

Self-Rewarding Language Models propose an architecture in which the Large Language Models(LLMs) both generates responses and evaluates its own outputs via LLM-as-a-Judge prompting, dynamically improving its generative capabilities through iterative Direct Preference Optimization (DPO). However, our analysis reveals a critical limitation in existing Self-Rewarding paradigms: the synchronized improvement of chosen and rejected responses progressively narrows the representational difference between contrasting samples, undermining effective preference learning. We propose \textbf{Temporal Self-Rewarding Language Models} that strategically coordinate past, present, and future model generations to sustain learning signals. Our dual-phase framework introduces: (1) \textit{Anchored Rejection} - fixing rejected responses using the past initial model's outputs and (2) \textit{Future-Guided Chosen} - dynamically curating chosen samples using next-generation model predictions. Extensive experiments across three model families (Llama, Qwen, Mistral) and different model sizes (Llama3B/8B/70B) demonstrate significant improvements when trained with our method compared to Self-Rewarding using same computation resources. For example, Llama3.1-8B reaches a 29.44 win rate on AlpacaEval 2.0 with our method, outperforming the Self-Rewarding baseline (19.69) by 9.75. Notably, our method also demonstrates superior out-of-distribution generalization across mathematical reasoning (GSM8K), knowledge-based QA (ARC, TruthfulQA), and code generation (HumanEval) tasks, even though we do not specifically collect such training data.

📄 PDF Abstract BibTeX arXiv:2508.06026

Code (0)

등록된 구현이 없습니다.

Tasks

Mathematical ReasoningCode Generation

Similar Papers 제목 키워드 기반

Process-based Self-Rewarding Language Models

2025-03-05 · Shimao Zhang, Xiao Liu, Xin Zhang, Junxiao Liu 외

Large Language Models have demonstrated outstanding performance across various downstream tasks and have been widely applied in multiple scenarios. Human-annotated preference data is used for training to further improve …

Mathematical Reasoning

CREAM: Consistency Regularized Self-Rewarding Language Models

2024-10-16 · Zhaoyang Wang, Weilei He, Zhiyuan Liang, Xuchao Zhang 외

Recent self-rewarding large language models (LLM) have successfully applied LLM-as-a-Judge to iteratively improve the alignment performance without the need of human annotations for preference data. These methods commonl…

Language ModelingLanguage Modelling

Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models

2025-08-01 · Zizhuo Zhang, Jianing Zhu, Xinmu Ge, Zihua Zhao 외 arxiv

While reinforcement learning with verifiable rewards (RLVR) is effective to improve the reasoning ability of large language models (LLMs), its reliance on human-annotated labels leads to the scaling up dilemma, especiall…

Self-Supervised LearningReinforcement LearningMathematical Reasoning

Decouple to Generalize: Context-First Self-Evolving Learning for Data-Scarce Vision-Language Reasoning

2025-12-07 · Tingyu Li, Zheng Sun, Jingxuan Wei, Siyuan Li 외 arxiv

Recent vision-language models (VLMs) achieve remarkable reasoning through reinforcement learning (RL), which provides a feasible solution for realizing continuous self-evolving large vision-language models (LVLMs) in the…

Reinforcement Learning

Self-Rewarding Sequential Monte Carlo for Masked Diffusion Language Models

2026-02-02 · Ziwei Luo, Ziqi Jin, Lei Wang, Lidong Bing 외 arxiv

This work presents self-rewarding sequential Monte Carlo (SMC), an inference-time scaling algorithm enabling effective sampling of masked diffusion language models (MDLMs). Our algorithm stems from the observation that m…