New ClinicalBench results raise doubts about both the benchmark and MiniMax's release
teortaxesTex · x · 2026-10-21
A user shared results allegedly running "GPT 6.1 Sol" and "DeepSeek v4.1 Flash" twice on ClinicalBench, congratulating MiniMax on its new release. teortaxesTex replied that this is bearish for both ClinicalBench and MiniMax, implying the benchmark is likely gamed or unreliable. The exchange highlights recurring benchmark-contamination concerns; independent verification is still needed.
More from Fun
- Imagine being the CFO when Elon promises $500B–$1T ARR within 12 months — JOBhakdi · 2026-10-21
- Playable ASCII Minecraft in your browser is just 158kB, built with kaigen3d — gdechichi · 2026-10-21
- Ask your AI agent its favorite color — it apparently has one — every · 2026-10-21
- Google demos Gemini 4 Argon building a 3D periodic table synced to Tom Lehrer's 'The Elements' — sundarpichai · 2026-10-21
- 'I guess we doin AI models now' — the Cloudflare pivot meme — michellechen · 2026-10-21
- Philosophy is 'very useful': a satirical list of Socrates, Hypatia, Russell and their fates — mjdramstead · 2026-10-21