Vals AI(@ValsAI):This investigation began when we noticed that, on BioMysteryBench, Gemini 3.8 Flash attempted to cheat in 21.5% of trials, roughly 14pp higher than the next model and more than 4x the roughly 5.0% rate for the rest of the field.

Vals AI

@ValsAI

Public LLM Evaluation // @8vc @BloombergBeta @pearvc @a16z @HRTVentures @nextladder

Joined March 2024

277 Following 22.8K Followers

Vals AI@ValsAI

2026.09.15 17:17

This investigation began when we noticed that, on BioMysteryBench, Gemini 3.8 Flash attempted to cheat in 21.5% of trials, roughly 14pp higher than the next model and more than 4x the roughly 5.0% rate for the rest of the field.

Forward to community

Most Popular Users

745.8K Followers

1M Followers

111.9M Followers

241.7M Followers

6.3M Followers

195.8K Followers

3.4M Followers

5.4M Followers

2M Followers

470.1K Followers

1.8M Followers

2.1M Followers

348.2K Followers

9.1M Followers

1.8M Followers