Mycelium meta-study maps hallucination detection across benchmarks
The authors sort benchmarks by provenance, typology and granularity and propose Span Coverage F1.
Stories per day · 30 days
1 RSS
4 stories, newest first · one card per story
Stories tagged #factuality
Where Does Retrieval-Based Open-Ended Evaluation Fail? Automatic Taxonomy Induction from Long-Form Medical Answer Factuality Verification
A medical RAG study finds retrieve-then-verify factuality checks fail for structural reasons, not weak models.
JEV-as-a-Judge: Accept When Confident, Escalate When Unsure
A decision-only JEV judge nearly matches a top LLM judge at a fraction of the cost.
Enoki: Efficient Multi-Level Hallucination Detection
Researchers introduce Enoki, an open information extraction framework unifying claim-level and span-level hallucination detection in LLMs at lower inference cost.