PAG: Multi-Turn Reinforced LLM Self-Correction with Policy as Generative Verifier
This addresses the scalability and reliability issues in LLM self-correction for complex reasoning tasks, representing a novel method rather than an incremental improvement.
The paper tackles the problem of LLMs struggling to verify their own outputs by proposing PAG, a framework that enables LLMs to self-correct through a multi-turn RL approach with selective revision, resulting in improved reasoning and verification accuracy across benchmarks.
Large Language Models (LLMs) have demonstrated impressive capabilities in complex reasoning tasks, yet they still struggle to reliably verify the correctness of their own outputs. Existing solutions to this verification challenge often depend on separate verifier models or require multi-stage self-correction training pipelines, which limit scalability. In this paper, we propose Policy as Generative Verifier (PAG), a simple and effective framework that empowers LLMs to self-correct by alternating between policy and verifier roles within a unified multi-turn reinforcement learning (RL) paradigm. Distinct from prior approaches that always generate a second attempt regardless of model confidence, PAG introduces a selective revision mechanism: the model revises its answer only when its own generative verification step detects an error. This verify-then-revise workflow not only alleviates model collapse but also jointly enhances both reasoning and verification abilities. Extensive experiments across diverse reasoning benchmarks highlight PAG's dual advancements: as a policy, it enhances direct generation and self-correction accuracy; as a verifier, its self-verification outperforms self-consistency.