LLMs directly edit model weights, matching gradient-based training
Playground and Francois Chaubard introduced Weight-Space Programming, letting frontier LLMs directly edit model weights without gradients or new data, matching gradient-based optimization across 87 settings and boosting a robotics task's success rate from 68% to 88% by changing only 2,048 parameters.
2026-11-07 ~ 2026-11-08 · 2 related posts
- Episode 1: MIT Paper: GPT-6 Directly Edits 0.016% of Weights, Boosting Small Model Math 6x(2026-11-03, 159 posts)
- Episode 2: Editing 0.016% of weights fixes Qwen reasoning loops, lifting MATH-500 to 48.1%(2026-11-03, 2 posts)
- Episode 3: Old hands-on guide resurfaces as MIT weight-editing claims spark interest(2026-11-03, 2 posts)
- Episode 4: MIT's Weight-Space Programming Claim Meets Wave of Independent Scrutiny(2026-11-03, 13 posts)
- Episode 5: LLM-fixed driving nets still work with weights zeroed out(2026-11-04, 2 posts)
- Episode 6: MIT's Weight-Space Programming: GPT-6 Directly Edits Other Models' Weights(2026-11-05, 9 posts)
- Episode 7: Weight Editing Fixes Real Model Bugs but Vanishes After 20 Fine-tune Steps(2026-11-07, 3 posts)
- Episode 8: LLMs directly edit model weights, matching gradient-based training(2026-11-07, 2 posts)
- LLMs Edit Robot Policy Weights Directly: 2,048 Parameter Tweak Lifts Success from 68% to 88% — zhaoran_wang · 2026-11-07
- LLM named as co-author on weight-space programming study matching gradient optimization in 30/87 settings — lateinteraction · 2026-11-08