Microsoft 提出可解释权重空间编程:Claude Opus 5.5 直接写权重,无需梯度训练
微软和 DAIR.AI 都在推的这篇论文挺有意思:让 Claude Opus 5.5 当智能体直接写 transformer 权重,不做梯度训练,算法任务 26 次全对,最小只用 166 个参数,结果人能逐行读懂。做可解释性研究或神经科学数据分析的读者可以看看它对脑数据的用法。
Microsoft 及合作者发布论文,提出 interpretable weight-space programming:让 Claude Opus 5.5 智能体直接以代码形式编写 transformer 权重,完全不用梯度训练。智能体提出一个计算方案,把它写成显式权重和特征,测试后修改,迭代 30 到 50 轮,产出的结果人类可以逐行阅读。
在算法类任务上,全部 26 次运行在留出数据上都达到 100% 准确率,通常是把已知算法直接写进权重;解决任务的最小模型中位参数量只有 166。模型规模影响明显:Sonnet 5.5 中位需要 41,600 参数,Haiku 5.5 需要 100,000。语言任务则困难得多,模型停留在约 190,000 参数且准确率不如学习得到的基线。
团队把同一套循环应用到脑数据上:在视觉皮层预测上优于现有可解释模型,在语言 fMRI 数据中还发现了数据集里的一个混杂因素。论文地址 arxiv.org/abs/2611.18114。
Recommend reading following up on the idea of weight-space programming. Bigger models (Opus 5.5) tend to do better and be more efficient. DAIR.AI @dair_ai Super interesting work from Microsoft and colleagues. They take a closer look at the exciting research area of weight-space programming. They had Claude Opus 5.5 write transformer weights directly in code, with no gradient training. The smallest models that solved the algorithmic tasks had a median of 166 parameters. The method is called interpretable weight-space programming. An agent proposes a computation, writes it as explicit weights and features, tests it, and revises it for 30 to 50 rounds. A person can read the result line by line. All 26 runs on algorithmic tasks reached 100% accuracy on held-out data, often by writing a known algorithm into the weights. They find that the model matters. Sonnet 5.5 needed a median of 41,600 parameters, and Haiku 5.5 needed 100,000. Language tasks were harder. Those models stayed near 190,000 parameters and lost accuracy against learned baselines. They then used the same loop on brain data. In the visual cortex, it predicted responses better than existing interpretable models. In language fMRI, it found a confound in the dataset. arxiv.org/abs/2611.18114 Chat with Paper: academy.dair.ai/papers/interpr… 🔗 View Quoted Tweet📊 8 ⚡
本条情报归属 AITOP「论文」核心栏目,重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:Claude Opus 5.5、权重空间编程、可解释性、transformer,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「elvis」(@omarsar0),经 AITOP 多维评分管线结构化提炼。