Editing Just 0.016% of Weights Cures Model Reasoning Loops
The author built an interactive WebGPU walkthrough of "weight-space programming": without gradient descent, editing only 0.016% (131,072 of ~800M) of a small Qwen model's ffn_down output weights across layers 15 and 19 largely cures its self-verification loop. On identical prompts, runs hitting the 32k-token cap fell from 93.0% to 33.1%, while MATH-500 accuracy rose from 7.9% to 48.1%. Unlike broad model souping, the method targets a few specific failure cases and the neurons behind them, and is pitched as a fit for agentic propose-test-feedback loops.
Related Entities
Full Report
Contents4 sections
The base model gets stuck in a loop
With thinking on, small Qwen keeps re-checking itself — “wait, let me verify…” — until it runs out of budget instead of answering.
93.0% Runs that hit the 32k-token limit 2
An LLM edits 0.016% of the weights
It subtracts one direction from 64 neurons' output weights (ffn_down) in layers 15 and 19 — 131,072 of ~800M weights.
layer 0layer 23
Stripes drawn ~25× wider than true scale — at real size they would be invisible.
3
The edited model escapes
Same model, same prompts — it now stops looping and finishes its answers.
Hit the token limit 93.0%→33.1% MATH-500 accuracy 7.9%→48.1%
Where this might fit
Model souping - diffusion-LoRA merging is one case of it - is weight-space editing too. The difference is the target: souping picks merge weights (and sometimes merge order) with a broad benchmark; weight-space programming zeroes in on a few specific failure cases and the neurons behind them. And here's the fun part: it's a promising fit for agents - propose an edit, test it, feed the result back - the same edit-test-repeat loop behind the recent coding-agent wins.
作者发帖说明(HN):
I wanted to build a full weight-space programming loop entirely in the browser with WebGPU: let an LLM inspect Qwen3.5-0.8B, edit its weights, evaluate, and repeat.
the full loop needed too much VRAM. so I turned it into an interactive walkthrough of a concrete example: editing 0.016% of a model's weights to reduce reasoning loops, no gradient descent.
Based on exp8 of “Sparks of Weight-Space Programming.”: https://news.ycombinator.com/item?id=50341174