一项研究称,Qwen3.5-27B可通过推理修改自身权重Chinese only
研究人员让Qwen3.5-27B查看自身权重和已评分的练习答案,在禁用梯度的Python环境中自行提出修改。材料称,这次修改使其在预先确定的490道未见题目上的准确率从57.2%升至63.8%;较小的2B模型未作修改,4B和9B模型的修改则导致性能受损。
1 developments2 reports1 sources
Event developments 1
Study reports Qwen3.5-27B can edit its own weights through reasoning
Researchers gave Qwen3.5-27B access to its weights and graded practice answers in a Python session with gradients blocked. The account says its 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, while the 4B and 9B models broke themselves.
View source reports · 1 sources · 2 reports
- Rohan Paul (@rohanpaul_ai) · Study reports Qwen3.5-27B can edit its own weights through reasoning
- Rohan Paul (@rohanpaul_ai) · Research report describes Qwen3.5-27B editing its own weights through reasoning
All reports 2
Research report describes Qwen3.5-27B editing its own weights through reasoning
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.
1 more source reported this
- Rohan Paul (@rohanpaul_ai)
Study reports Qwen3.5-27B can edit its own weights through reasoning
Researchers gave Qwen3.5-27B access to its weights and graded practice answers in a Python session with gradients blocked. The account says its 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, while the 4B and 9B models broke themselves.
1 more source reported this
- Rohan Paul (@rohanpaul_ai)