AI Is Making Mistakes Before It Even 'Sees' Correctly. A New Benchmark Testing Perception ...
Moonshot AI's PerceptionBench is a 3000-question benchmark that strips ... hallucination . Conversely, Gemini 3.5 Flash, which ranks mid-pack ...
Read the original at XenoSpectrum ↗
- Topics
- Hallucinations & Fabrication
- Mentions
- GeminiGoogleMoonshot AI
- Tracked via
- “AI hallucination”