A Self-Supervised Reinforcement Learning Approach for Fine-Tuning Large Language Models Using Cross-Attention Signals

• URL: http://arxiv.org/abs/2502.10482v1
• Date: Fri, 14 Feb 2025 01:44:04 GMT
• Title: A Self-Supervised Reinforcement Learning Approach for Fine-Tuning Large Language Models Using Cross-Attention Signals
• Authors: Andrew Kiruluta, Andreas Lemos, Priscilla Burity,
• Abstract summary: We propose a novel reinforcement learning framework for large language models that does not rely on human in the loop feedback. Instead, our approach uses cross attention signals within the model itself to derive a self supervised reward.
• Score: 0.0
• License:
• Abstract: We propose a novel reinforcement learning framework for post training large language models that does not rely on human in the loop feedback. Instead, our approach uses cross attention signals within the model itself to derive a self supervised reward, thereby guiding iterative fine tuning of the model policy. By analyzing how the model attends to the input prompt during generation, we construct measures of prompt coverage, focus, and coherence. We then use these measures to rank or score candidate responses, providing a reward signal that encourages the model to produce well aligned, on topic text. In empirical comparisons against standard policy gradient methods and RL fine tuning with synthetic preference models, our method shows significant gains in prompt relevance and consistency over a non RL baseline. While it does not yet match the performance of fully human supervised RLHF systems, it highlights an important direction for scaling alignment with minimal human labeling. We provide a detailed analysis, discuss potential limitations, and outline future work for combining cross-attention based signals with smaller amounts of human feedback.

Related papers

• LoRE-Merging: Exploring Low-Rank Estimation For Large Language Model Merging [10.33844295243509]
  We propose a unified framework for model merging based on low-rank estimation of task vectors without the need for access to the base model, named textscLoRE-Merging. Our approach is motivated by the observation that task vectors from fine-tuned models frequently exhibit a limited number of dominant singular values, making low-rank estimations less prone to interference.
  arXiv  Detail & Related papers  (2025-02-15T10:18:46Z)
• PILAF: Optimal Human Preference Sampling for Reward Modeling [14.336058926701432]
  We propose Policy-Interpolated Learning for Aligned Feedback (PILAF), a novel response sampling strategy for preference labeling. PILAF explicitly aligns preference learning with maximizing the underlying oracle reward.
  arXiv  Detail & Related papers  (2025-02-06T18:09:00Z)
• Disentangling Length Bias In Preference Learning Via Response-Conditioned Modeling [87.17041933863041]
  We introduce a Response-conditioned Bradley-Terry (Rc-BT) model that enhances the reward model's capability in length bias mitigating and length instruction following. We also propose the Rc-DPO algorithm to leverage the Rc-BT model for direct policy optimization (DPO) of large language models.
  arXiv  Detail & Related papers  (2025-02-02T14:50:25Z)
• Low-Rank Contextual Reinforcement Learning from Heterogeneous Human Feedback [12.373566593905792]
  Reinforcement learning from human feedback (RLHF) has become a cornerstone for aligning large language models with human preferences. We propose a Low-rank Contextual RLHF framework that integrates contextual information to better model heterogeneous feedback. Our approach builds on a contextual preference model, leveraging the intrinsic low-rank structure of the interaction between user contexts and query-answer pairs.
  arXiv  Detail & Related papers  (2024-12-27T04:02:46Z)
• Learning Reward and Policy Jointly from Demonstration and Preference Improves Alignment [58.049113055986375]
  We develop a single stage approach named Alignment with Integrated Human Feedback (AIHF) to train reward models and the policy. The proposed approach admits a suite of efficient algorithms, which can easily reduce to, and leverage, popular alignment algorithms. We demonstrate the efficiency of the proposed solutions with extensive experiments involving alignment problems in LLMs and robotic control problems in MuJoCo.
  arXiv  Detail & Related papers  (2024-06-11T01:20:53Z)
• Multi-turn Reinforcement Learning from Preference Human Feedback [41.327438095745315]
  Reinforcement Learning from Human Feedback (RLHF) has become the standard approach for aligning Large Language Models with human preferences. Existing methods work by emulating the preferences at the single decision (turn) level. We develop novel methods for Reinforcement Learning from preference feedback between two full multi-turn conversations.
  arXiv  Detail & Related papers  (2024-05-23T14:53:54Z)
• RLHF Deciphered: A Critical Analysis of Reinforcement Learning from Human Feedback for LLMs [49.386699863989335]
  Training large language models (LLMs) to serve as effective assistants for humans requires careful consideration. A promising approach is reinforcement learning from human feedback (RLHF), which leverages human feedback to update the model in accordance with human preferences. In this paper, we analyze RLHF through the lens of reinforcement learning principles to develop an understanding of its fundamentals.
  arXiv  Detail & Related papers  (2024-04-12T15:54:15Z)
• Nash Learning from Human Feedback [86.09617990412941]
  We introduce an alternative pipeline for the fine-tuning of large language models using pairwise human feedback. We term this approach Nash learning from human feedback (NLHF) We present a novel algorithmic solution, Nash-MD, founded on the principles of mirror descent.
  arXiv  Detail & Related papers  (2023-12-01T19:26:23Z)
• SALMON: Self-Alignment with Instructable Reward Models [80.83323636730341]
  This paper presents a novel approach, namely SALMON, to align base language models with minimal human supervision. We develop an AI assistant named Dromedary-2 with only 6 exemplars for in-context learning and 31 human-defined principles.
  arXiv  Detail & Related papers  (2023-10-09T17:56:53Z)
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model [119.65409513119963]
  We introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods.
  arXiv  Detail & Related papers  (2023-05-29T17:57:46Z)
• Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback [8.409764908043396]
  We apply preference modeling and reinforcement learning from human feedback to finetune language models to act as helpful assistants. We find this alignment training improves performance on almost all NLP evaluations. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data.
  arXiv  Detail & Related papers  (2022-04-12T15:02:38Z)

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.