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

• URL: http://arxiv.org/abs/2508.06026v1
• Date: Fri, 08 Aug 2025 05:25:54 GMT
• Title: Temporal Self-Rewarding Language Models: Decoupling Chosen-Rejected via Past-Future
• Authors: Yidong Wang, Xin Wang, Cunxiang Wang, Junfeng Fang, Qiufeng Wang, Jianing Chu, Xuran Meng, Shuxun Yang, Libo Qin, Yue Zhang, Wei Ye, Shikun Zhang,
• Abstract summary: 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.<n>We propose textbf Self-Rewarding Language Models that strategically coordinate past, present, and future model generations to sustain learning signals.
• Score: 38.1810626252963
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• Multi-Metric Preference Alignment for Generative Speech Restoration [15.696247605348383]
  We propose a multi-metric preference alignment strategy for generative models.<n>We observe consistent and significant performance gains across three diverse generative paradigms.<n>Our aligned models can serve as powerful ''data annotators'', generating high-quality pseudo-labels.
  arXiv  Detail & Related papers  (2025-08-24T07:05:10Z)
• Self-Correcting Code Generation Using Small Language Models [11.4397549365277]
  Self-correction has demonstrated potential in code generation by allowing language models to revise and improve their outputs through successive refinement.<n>We introduce CoCoS, an approach designed to enhance the ability of small language models for multi-turn code correction.<n>With 1B-scale models, CoCoS achieves improvements of 35.8% on the MBPP and 27.7% on HumanEval compared to the baselines.
  arXiv  Detail & Related papers  (2025-05-29T04:04:44Z)
• Self-Boost via Optimal Retraining: An Analysis via Approximate Message Passing [58.52119063742121]
  Retraining a model using its own predictions together with the original, potentially noisy labels is a well-known strategy for improving the model performance.<n>This paper addresses the question of how to optimally combine the model's predictions and the provided labels.<n>Our main contribution is the derivation of the Bayes optimal aggregator function to combine the current model's predictions and the given labels.
  arXiv  Detail & Related papers  (2025-05-21T07:16:44Z)
• SPaR: Self-Play with Tree-Search Refinement to Improve Instruction-Following in Large Language Models [88.29990536278167]
  We introduce SPaR, a self-play framework integrating tree-search self-refinement to yield valid and comparable preference pairs.<n>Our experiments show that a LLaMA3-8B model, trained over three iterations guided by SPaR, surpasses GPT-4-Turbo on the IFEval benchmark without losing general capabilities.
  arXiv  Detail & Related papers  (2024-12-16T09:47:43Z)
• PRefLexOR: Preference-based Recursive Language Modeling for Exploratory Optimization of Reasoning and Agentic Thinking [0.0]
  PRefLexOR combines preference optimization with concepts from Reinforcement Learning to enable models to self-teach. We focus on applications in biological materials science and demonstrate the method in a variety of case studies.
  arXiv  Detail & Related papers  (2024-10-16T08:46:26Z)
• Self-Exploring Language Models: Active Preference Elicitation for Online Alignment [88.56809269990625]
  We propose a bilevel objective optimistically biased towards potentially high-reward responses to actively explore out-of-distribution regions. Our experimental results demonstrate that when fine-tuned on Zephyr-7B-SFT and Llama-3-8B-Instruct models, Self-Exploring Language Models (SELM) significantly boosts the performance on instruction-following benchmarks.
  arXiv  Detail & Related papers  (2024-05-29T17:59:07Z)
• LLMs can learn self-restraint through iterative self-reflection [57.26854891567574]
  Large Language Models (LLMs) must be capable of dynamically adapting their behavior based on their level of knowledge and uncertainty associated with specific topics. This adaptive behavior, which we refer to as self-restraint, is non-trivial to teach. We devise a utility function that can encourage the model to produce responses only when it is confident in them.
  arXiv  Detail & Related papers  (2024-05-15T13:35:43Z)
• Secrets of RLHF in Large Language Models Part II: Reward Modeling [134.97964938009588]
  We introduce a series of novel methods to mitigate the influence of incorrect and ambiguous preferences in the dataset. We also introduce contrastive learning to enhance the ability of reward models to distinguish between chosen and rejected responses.
  arXiv  Detail & Related papers  (2024-01-11T17:56:59Z)
• Improving Non-autoregressive Generation with Mixup Training [51.61038444990301]
  We present a non-autoregressive generation model based on pre-trained transformer models. We propose a simple and effective iterative training method called MIx Source and pseudo Target. Our experiments on three generation benchmarks including question generation, summarization and paraphrase generation, show that the proposed framework achieves the new state-of-the-art results.
  arXiv  Detail & Related papers  (2021-10-21T13:04:21Z)
• Layer-wise Analysis of a Self-supervised Speech Representation Model [26.727775920272205]
  Self-supervised learning approaches have been successful for pre-training speech representation models. Not much has been studied about the type or extent of information encoded in the pre-trained representations themselves.
  arXiv  Detail & Related papers  (2021-07-10T02:13:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.