AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

From Detection to Diagnosis: Advancing Hallucination Analysis with Automated Data Synthesis

Yanyi Liu, Qingwen Yang, Tiezheng Guo, Feiyu Qu, Jun Liu, Yingyou Wen

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Data and hallucination detection · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
GRPO training with HDG-synthesized data (reasoning mode)Qwen3-4B · detection average of sevenmacro F1: 76.74 -> 79.65Improved
GRPO training with HDG data (non-reasoning mode)Qwen3-4B · detection average of sevenmacro F1: 64.92 -> 70.13Improved
enable reasoning mode (ablation)Qwen3-4B · detection average of sevenmacro F1: 64.92 -> 76.74Improved
enable reasoning mode (ablation)HDM-4B-RL · detection average of sevenmacro F1: 70.13 -> 79.65Improved
HDM-4B-RL vs specialized detector Lynx-8BHDM-4B-RL · detection average of sevenmacro F1: 77.25 -> 79.65Improved
HDM-4B-RL vs Lynx-8B on HaluEvalHDM-4B-RL · HaluEvalmacro F1: 86.96 -> 83.96Hurt
HDM-4B-RL vs GPT-4.1 promptingHDM-4B-RL · detection average of sevenmacro F1: 75.37 -> 79.65Improved
HDM-4B-RL vs o4-mini promptingHDM-4B-RL · detection average of sevenmacro F1: 82.62 -> 79.65Hurt
multi-step pipeline instead of single prompt (ablation)Qwen3-32B · SummEval diagnosislocalization hit rate: 64.85 -> 76.97Improved
single-prompt HDM-4B-RL vs single-prompt GPT-4.1 and o4-miniHDM-4B-RL · SummEval diagnosisdetection accuracy: 77.54 (GPT-4.1) -> 92.28Improved
single-prompt HDM-4B-RL vs Qwen3-32B single promptHDM-4B-RL · SummEval diagnosismitigation AlignScore: 70.98 -> 69.16Mixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy