跳到正文
𝕏 Rohan Paul(x.com)·· 2 天前品質評分62
重點導讀

模型部署後想變得更好,通常得再花大量運算資源訓練一次。

  • 研究團隊讓 Qwen3.5-27B 檢視自己的權重、練習題答案和 Python 執行環境,並阻止它使用梯度。
  • 模型比較答對與答題超時時的內部狀態,再調整兩者之間的差異。在事先固定的 490 題測試中,正確率從 57.2% 提高到 63.8%。不過,2B 模型沒有做出修改,4B 和 9B 模型則把自己弄壞了。作者建議先讓模型提出權重修改,再用保留的測試題確認效果。

AI 伴隨問答

自備 Key 深度分析

針對這篇報導有疑問?可直接點選快捷問題或向 AI 追問背景脈絡與技術細節:

Rohan Paul
Rohan Paul@rohanpaul_ai
·2 天前
𝕏

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.

推文來源 · X (Twitter)在 X 參與討論

讀者討論

理性交流 · 即時討論