// the find
EdinburghNLP/awesome-hallucination-detection
List of papers on hallucination detection in LLMs.
A plain-markdown bibliography of hallucination-detection and mitigation papers for LLMs and VLMs, maintained by Edinburgh NLP. It's a reading list, not a tool — no code, no API, just annotated paper links with metrics and datasets for each entry. Useful for someone doing a literature review on the topic, not for someone trying to ship detection into a pipeline.
Genuinely active — entries dated October 2026 sit next to EMNLP/ACL/NeurIPS 2025 work, so it isn't an abandoned list someone started once and forgot. Each entry goes beyond a bare link: metrics used, datasets, and a few sentences on the actual contribution, which saves you from opening every arXiv PDF just to figure out relevance. Comes from a credible source (Edinburgh NLP group) rather than an SEO-farmed awesome-list clone.
Zero structure — 140+ entries in one long flat list with no sections for detection vs. mitigation, text vs. multimodal, training-based vs. training-free, so finding what applies to your use case means skimming the whole thing. No code anywhere in the repo itself; it's a bibliography, so there's nothing to install, run, or build on. No editorial filter — every entry is summarized with the same uncritical 'introduces X, achieves Y%' abstract-lifted tone, with no indication of which methods are actually load-bearing in practice versus a one-off benchmark paper nobody built on. Entry quality is inconsistent — some are rigorous NeurIPS/ACL papers, others are unreviewed arXiv preprints or a Zenodo self-published 'cognometry' piece, with no flagging of peer-review status.