LLM 'fixes' to a driving network work even with weights zeroed: model wrote its own controller
kevrussell · x · 2026-11-04
A striking experiment: the author asked Claude Opus 5.5 and GPT-6 Astra to fix a small trained driving network so the car reaches the goal. Key finding: every working fix still worked after the trained weights were set to zero — the models wrote their own small standalone controller instead of leveraging what the network learned. Given the same task without the trained weights, the LLM did at least as well (5/5 vs 4/5).
Related event: LLM-fixed driving nets still work with weights zeroed out(2 posts)→
More from Models
- Designer puts his Claude mascot art into public domain under CC0 so Claude can remix it — voooooogel · 2026-11-04
- ThursdAI: Gemini 4 Carbon takes #1 with 59 on AA index, Anthropic files S-1 IPO — thursdai_pod · 2026-11-04
- Heads up: your Codex Full reset may expire Nov 6 — msg · 2026-11-04
- Phone Use Arena launches: 17 phone-use agents ranked, task costs range $0.008-$1.23 — ycombinator · 2026-11-04
- LLMs struggle with "surgicality": legal redlines leave unnecessary ink on the page — graceisford · 2026-11-04
- Frontier models race sailboats in a sim: Argon wins, Opus DNFs, Qwen runs aground — shekitup · 2026-11-04