arXiv Artificial Intelligence· Nazmus Sakib Ahmed·· 2 天前AI 评分42
Claude Opus 5.5 能否真正推理神经网络权重，还是仅靠记忆？Weight-Space Programming 探究
Can an LLM Actually Reason and Program about Neural Network's Weights or is it just memorization on steroids? Exploring "Weight-Space Programming"
AI 导读
研究测试 Claude Opus 5.5 能否对约 125,000 参数的小型卷积网络进行权重空间编程，模型在无工具、无前向传播、无数据集名称条件下能准确完成局部计算，如从 batch-normalization 统计推断约三次优化器更新。
来源：arXiv Artificial Intelligence · arxiv.org