aifollow.news
Back Rohan Paul
Rohan Paul· @rohanpaul_ai · X· · Original publication time AI score61

Research report describes Qwen3.5-27B editing its own weights through reasoningMachine translation

AI introduction

A post describing the research says Qwen3.5-27B was given access to its weights, graded practice answers and a Python session, with gradients blocked. Its edit raised accuracy from 57.2% to 63.8% on a preselected test of 490 unseen problems; the 2B model made no edit, while the 4B and 9B models damaged their performance.

Original post · Original

@rohanpaul_ai

– https://t.co/Jqlke4C7nz

Title: "Language Models Can Rewrite Their Own Weights"

ReplyRohan Paul@rohanpaul_ai
New Stanford+other top labs paper finds that a 27B open model can improve itself by editing its own weights through reasoning, with no training or gradients. Improving a deployed model today means another training run on large clusters. Prompts and extra inference compute help, but the gain disappears once the task ends. They gave Qwen3.5-27B its own weights, its graded practice answers, and a Python session, with gradients blocked. It compared its hidden states on answers that finished against answers that ran out of tokens, then weakened the direction separating them. Its own edit raised accuracy from 57.2% to 63.8% on 490 unseen problems, in a test fixed in advance. The 2B model made no edit, and the 4B and 9B models broke themselves. If you run a large open model, let it propose weight edits for its own failures, but keep only edits that pass held-out tests.
View this post on X ↗

Source: Rohan Paul · x.com

Research
Found an error? Send a correction