Primary sourceLatent.Space
View sourceEvent ·Multi-source
MIT 研究：编辑 0.016% 权重让 Qwen3.5-0.8B 的 MATH-500 成绩从 7.9% 升至 48.1%
MIT 的一项研究中，GPT-6 Astra 分析了 Qwen3.5-0.8B 在数学题上耗尽 32K token 仍无答案的原因，将问题定位到两层循环。研究团队随后编辑了该模型 0.016% 的权重。在采样条件固定的情况下，Qwen3.5-0.8B 在 MATH-500 上的成绩从 7.9% 提升到 48.1%。
Current read
Qwen3.5-0.8B 做数学题经常卡在死循环里烧 32K token，MIT 找到出问题的两层，只改 0.016% 权重就把 MATH-500 从 7.9 分拉到 48.1 分。
11 sources21° AI heatLast updated 2026/11/3 05:40:15
Evidence
11 sourcesRelated source 1rohanpaul_ai
View sourceRelated source 2宝玉
View sourceRelated source 3shao__meng
View sourceRelated source 4The Rundown AI
View sourceRelated source 5orange.ai
View sourceRelated source 6IT之家
View sourceRelated source 7李继刚
View sourceRelated source 8kimmonismus
View sourceRelated source 9Jerry Liu
View sourceRelated source 10Thomas Wolf
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.