Research map / Large language models / Reasoning and chain of thought
Hallucination Detection from Structural Reasoning Model
Jianbo Sun, Pengkun Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a structural reasoning model that extracts a directed acyclic reasoning graph from chain-of-thought traces to detect hallucinations in LLM reasoning.
Problem
Hallucinations in LLM chain-of-thought reasoning are hard to identify because linear traces hide cross-step dependencies and unsupported intermediate claims.
Approach
Model interactions among local reasoning steps as a structure, extracting a directed acyclic reasoning graph from the trace and using it to detect hallucinated intermediate claims, rather than treating the trace as a linear sequence.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use forward flow aggregation instead of reverse InvFlow (ablation) | FLOW · All four benchmarks | AP (%): 66.57 -> 73.69 | Improved |
| use forward flow aggregation instead of MaxErr (ablation) | FLOW · All four benchmarks | AP (%): 55.37 -> 73.69 | Improved |
| use forward flow aggregation instead of MinErr (ablation) | FLOW · All four benchmarks | AP (%): 31.70 -> 73.69 | Improved |
| use forward flow aggregation instead of MeanErr (ablation) | FLOW · All four benchmarks | AP (%): 73.28 -> 73.69 | Improved |
| add DAG adjacency features to step-error features (FULL) (ablation) | logistic regression · MATH | AP (%): 74.74 -> 77.56 | Improved |
| add DAG adjacency features to step-error features (NOSHORTCUT) (ablation) | logistic regression · MATH | AP (%): 72.98 -> 76.05 | Improved |
| use high-fidelity dependency graph instead of parsed raw DAG (ablation) | FLOW · All four benchmarks | AP (%): 73.69 -> 80.15 | Improved |
| score steps locally on a DAG instead of whole-rationale GPT-5.2 judge | FLOW+ · GSM8K | AP (%): 41.55 -> 77.70 | Improved |
| score steps locally on a DAG instead of whole-rationale GPT-5.2 judge | FLOW+ · HumanEval | AP (%): 46.95 -> 83.08 | Improved |
| use FLOW+ aggregation instead of sampling-based detectors | FLOW+ · HumanEval | AP (%): 25.52 -> 83.08 | Improved |
| use FLOW+ instead of PDS | FLOW+ · HotpotQA | AP (%): 88.69 -> 83.23 | Hurt |
| use prefix-step judging vs whole-trace judging, then Step-FLOW with base/high-fidelity DAG (ablation) | Step-FLOW · MATH (Grok) | AP: 53 -> 65.70/72.87 | Improved |
Closest papers by meaning
- Auditing Meta-Cognitive Hallucinations in Reasoning Large Language Models (NeurIPS 2025 · similarity 0.93)
- Joint Evaluation of Answer and Reasoning Consistency for Hallucination Detection in Large Reasoning Models (AAAI 2026 · similarity 0.92)
- Scalable Token-Level Hallucination Detection in Large Language Models (NeurIPS 2026 · similarity 0.91)
- Learning to Reason for Hallucination Span Detection (ICLR 2026 · similarity 0.91)
- MIRAGE: Assessing Hallucination in Multimodal Reasoning Chains of MLLM (NeurIPS 2025 · similarity 0.91)
- Mitigating Hallucinations in Large Language Models via Causal Reasoning (AAAI 2026 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
