Fable 5.2 scores just 21% on MazeBench, mocked in training-cheating comic
patience_cave · x · 2026-11-18
A reposted comic mocks Fable 5.2's 21% score on MazeBench: the character breeze through a maze with a memorized cheat sheet, then protests "I thought this was another training session!" — a jab at models that ace benchmarks by memorizing training data and fail on real tests. Funny, and it nails the benchmark-contamination problem.
More from Fun
- Cute AI-generated plush toy travel photos made with fal hit Instagram — noahsolomon · 2026-11-18
- GPT-6.2 Astra Ultra put to the test: turn a dream-house PDF into a Blender model and playable game — BLCNYY · 2026-11-18
- /prune-tests: a Claude command to hunt down useless tests — teropa · 2026-11-18
- Pinecone co-founder slams AI community's outage panic as media-driven hysteria — econoar · 2026-11-18
- GPT-6.2 Astra-generated shaders power an interactive museum called Light and Matter — pvncher · 2026-11-18
- AI-generated 'village shamans' dancing in a snowy village becomes a meme — Megalania34 · 2026-11-18