Thinking, Faithful and Stable: Mitigating Hallucinations in LLMs
This addresses the issue of hallucinations in LLMs for users relying on accurate and faithful reasoning, though it is incremental as it builds on existing uncertainty and reinforcement learning methods.
The paper tackles the problem of hallucinations in large language models during multi-step reasoning by developing a self-correcting framework that uses fine-grained uncertainty signals to detect and mitigate unreliable reasoning in real time, resulting in improved final answer accuracy and reasoning calibration.
This project develops a self correcting framework for large language models (LLMs) that detects and mitigates hallucinations during multi-step reasoning. Rather than relying solely on final answer correctness, our approach leverages fine grained uncertainty signals: 1) self-assessed confidence alignment, and 2) token-level entropy spikes to detect unreliable and unfaithful reasoning in real time. We design a composite reward function that penalizes unjustified high confidence and entropy spikes, while encouraging stable and accurate reasoning trajectories. These signals guide a reinforcement learning (RL) policy that makes the model more introspective and shapes the model's generation behavior through confidence-aware reward feedback, improving not just outcome correctness but the coherence and faithfulness of their intermediate reasoning steps. Experiments show that our method improves both final answer accuracy and reasoning calibration, with ablations validating the individual contribution of each signal.