AIHugging Faceabout 1 hour ago
Hugging Face Releases Novelty Evaluation Benchmark
8 min read
Hugging Face releases Old Ideas, Novel Problems, a study showing large language models fail to reliably judge research novelty. Testing six LLMs, the team found that changing prompts shifts accuracy by more than 50 percentage points on AI-generated ideas, dropping from 93.5% to 40.9%.
Level
Hype check
Tap to vote and see what everyone thinks.
