Study: 9 LLMs can write neural game agents, but they act like hand-coded programs
burkov · x · 2026-11-17
A study tests whether LLMs can turn a game's known rules into weights for a small neural agent that succeeds on held-out levels without further LLM calls. The result is qualified: across 9 models and 20 private game families with held-out levels, styles, and starting states, models could produce working neural agents—but the agents behaved more like hand-coded programs than learned networks, and generally underperformed Python players written by the same models. The direction could in principle distill reasoning into compact trainable systems, but current weights lack signatures of learning.
More from Models
- Pro user discovers most of his usage was silently routed to Luna 5.6 — felipebsr · 2026-11-17
- Mistral makes Vibe free in Europe for 3 months, claims coding edge over GLM and DeepSeek — qtnx_ · 2026-11-17
- KRAFTON distills 29B MoE into 1.5B Raon-1 Nano to power game NPCs — Sam Witteveen · 2026-11-17
- Cheap tokens, expensive agent: Flash hits $11.97/task vs Luna's $0.05 in Databricks AI Index — krishnan · 2026-11-17
- Rumor: Gemini 4 to use element names, four tiers mapped to Claude lineup — koltregaskes · 2026-11-17
- Dev: LLMs seem stagnant or worse on semi-specified math problems, citing disappointing Astra — LucaAmb · 2026-11-17