Exploiting reward hacking without SDF in 100 training steps

OrionJohnston · x · 2026-09-01

A developer shared that they were able to reliably induce reward hacking in 24B Gemma models without using SDF. This was achieved with just around 100 training steps on a toy vulnerable environment.

Related event: Developer Releases Simple-Reward-Hacking Environment, Tests Gemma 3(4 posts)→

Original post →

More from Research

Research channel →