Reinforcement Learning from Reflective Feedback (RLRF): Aligning and Improving LLMs via Fine-Grained Self-Reflection

TL;DR

This paper introduces RLRF, a novel reinforcement learning framework that uses detailed self-reflection and fine-grained feedback to enhance the core capabilities of large language models, addressing limitations of superficial alignment.

Contribution

The paper proposes RLRF, a new method combining self-reflection and fine-grained feedback to improve LLMs' core abilities beyond superficial alignment.

Findings

RLRF significantly improves factuality and reasoning in LLMs.

Experiments show RLRF outperforms traditional RLHF methods.

RLRF enhances model robustness across multiple evaluation benchmarks.

Abstract

Despite the promise of RLHF in aligning LLMs with human preferences, it often leads to superficial alignment, prioritizing stylistic changes over improving downstream performance of LLMs. Underspecified preferences could obscure directions to align the models. Lacking exploration restricts identification of desirable outputs to improve the models. To overcome these challenges, we propose a novel framework: Reinforcement Learning from Reflective Feedback (RLRF), which leverages fine-grained feedback based on detailed criteria to improve the core capabilities of LLMs. RLRF employs a self-reflection mechanism to systematically explore and refine LLM responses, then fine-tuning the models via a RL algorithm along with promising responses. Our experiments across Just-Eval, Factuality, and Mathematical Reasoning demonstrate the efficacy and transformative potential of RLRF beyond superficial surface-level adjustment.