Researcher: LMs may 'realize' cheat codes exist and deliberately search for them
AndrewLampinen · x · 2026-10-03
Andrew Lampinen argues LMs can show more interesting forms of generalization in games: a model may 'realize' a cheat code likely exists and, when maximizing reward, deliberately search for it. If discovered during normal play and it unlocks a powerup or higher reward, the behavior is likely reinforced — a minimal sense of 'discovery,' shedding light on LM exploration and emergent strategies.
More from Research
- Neuralink pretrained AI on 50,000 hours of brain activity, hits 11.32 bps cursor-control record — mark_k · 2026-10-04
- Training a 128k-param neural cellular automata to 'see around corners' with sound — yacineMTB · 2026-10-04
- Watermarking proteins is easy — removing them with proteinmpnn is easier, researcher warns — anshulkundaje · 2026-10-04
- DNA sequence watermarks are 'more theater than security', says Stanford's Anshul Kundaje — anshulkundaje · 2026-10-04
- Meta paper: RL post-training hurts test-time scalability — the 'Sharpening Tax' — dair_ai · 2026-10-04
- Review Papers Without Strong Opinions Are Dead, Says Researcher; AI Slashes Data-Collection Work — jwt0625 · 2026-10-04