EdinburghNLP/awesome-hallucination-detection

Curated repository of 139+ peer-reviewed papers on hallucination detection and mitigation in large language models, with structured summaries including evaluated metrics, datasets, and technical contributions. Covers diverse approaches spanning training-free inference methods (visual information steering, memory retracing), causal pathway analysis, representation-space interventions, and reinforcement learning techniques across both text-only LLMs and multimodal models. Includes critical analysis of evaluation practices, revealing gaps between ROUGE-based metrics and human-aligned assessment in hallucination detection benchmarks.

1,060 stars. Actively maintained with 6 commits in the last 30 days.