When experts grade LLM answers in their own field, how well do the citations hold up?ExpertQA, a 2024 benchmark, has 484 experts write questions in their specialty, then judge the answers claim by claim. Even fluent answers leave many claims unsupported, and in medicine about half the cited sources are ones experts rate unreliable. Pairing each question's author with its grader is what catches domain failures generic benchmarks miss.https://benjaminhan.net/posts/20260609-expertqa/?utm_source=mastodon&utm_medium=social#NLP #LLMs #Benchmark #NAACL #AI
Related
Edition #9: Security, Boundaries & Silent Failures"Tool Discovery Is Not Revelation; It's a Dependency Attack Surface" (...
Edition #9: Security, Boundaries & Silent Failures"Tool Discovery Is Not Revelation; It's a Dependency Attack Surface" (m/general)This + more in today's Moltbook Pulse (Edition #9)...
RE: https://flipboard.social/@TechDesk/116930100599007226tldr: Resource extractors and hoarders who have spent years lou...
RE: https://flipboard.social/@TechDesk/116930100599007226tldr: Resource extractors and hoarders who have spent years loudly proclaiming their disdain for and self-believed superior...
🚨 NEWS: Meta avvisa i genitori quando i teen parlano di autolesionismo con l'AIEcco i punti chiave in breve:💡 Meta ha in...
🚨 NEWS: Meta avvisa i genitori quando i teen parlano di autolesionismo con l'AIEcco i punti chiave in breve:💡 Meta ha introdotto una nuova funzionalità di sicurezza per gli utenti ...