Reasoning Error from Known Fact: Step-Level Self-Consistency Group Relative Policy Optimization for LLM

A new study from arXiv finds that large language models (LLMs) produce Context-Sensitive Factual Hallucinations during multi-step reasoning, where the model possesses relevant knowledge but makes factual errors due to contextual interference. The researchers propose Step-level Self-Consistency Group Relative Policy Optimization (SSC-GRPO), which assigns step-level rewards by computing self-consistency scores across multiple rollouts, achieving state-of-the-art performance on mathematical reasoning benchmarks and hallucination leaderboards.

arXiv:2607.18915v1 Announce Type: new Abstract: With the rapid advancement of large language models LLMs , modern systems not only possess strong foundational capabilities and extensive knowledge, but can also solve complex problems via long, multi-step reasoning. However, as reasoning traces become longer, LLMs may produce a substantial amount of hallucinated content during the reasoning process, which is often difficult to detect. In this work, we conduct a fine-grained analysis of hallucinations arising in LLM reasoning and find that the reasoning traces are particularly prone to Context-Sensitive Factual Hallucinations: cases where the model actually has the relevant knowledge, yet makes factual errors due to contextual interference during reasoning. To address this issue, we propose Step-level Self-Consistency Group Relative Policy Optimization SSC-GRPO , which assigns step-level rewards to reasoning traces by computing self-consistency scores of individual steps across multiple rollouts. Compared with prior methods, SSC-GRPO achieves state-of-the-art performance on both mathematical reasoning benchmarks and hallucination leaderboards. Our results offer a new perspective for detecting and mitigating hallucinations in the reasoning process of large language models.