All benchmarks saturated? Someone built an Anime Identification Difficulty Set for LLMs
gnukeith · x · 2026-09-04
gnukeith created the Anime Identification Difficulty Set (AIDS) benchmark, arguing every other benchmark is saturated, and is now ranking model categories on it—a meme-flavored jab at benchmark saturation in AI.
More from Fun
- Dev shows off 'pocket physics' fidget toy that uses gyroscope to pick reactive materials — jh3yy · 2026-09-04
- Matt Shumer Claims GPT-6 Astra Built a Manhattan World in Unreal Engine Over a Week — jasonkneen · 2026-09-04
- David Lynch explains why AI is just like a pencil — UrAWeaselBro · 2026-09-04
- Dev builds a p5js brush that draws schematic diagrams freehand — anselm · 2026-09-04
- AIE Conference Speaker Caught Reading Claude-Generated Script Onstage — himanshustwts · 2026-09-04
- Claim: a 129-digit number divides RSA-260, shocking the AI crowd — rickasaurus · 2026-09-04