Event ·Multi-source

谷歌 DeepMind 展示 AI 递归自我改进闭环

谷歌 DeepMind 的 Dream-RSI 模型能回放过去发现过程，低成本测试数千种探索策略，再将更优策略用于下一轮。在算法设计、数学优化和 GPU 内核工程中，它保持或提高了结果质量，最高将 Agent 调用次数减少 162 倍。

Current read

谷歌 DeepMind 新的 Dream-RSI 模型，能自己改进自己，在算法设计、数学优化这些领域效果更好，还大幅减少了调用次数。

3 sources78° AI heatLast updated 2026/9/18 03:13:09

Evidence

3 sources

Primary sourceAI Will

2026/9/18 03:13:09View source

Related source 1Yangyi

2026/9/18 05:47:42View source

Related source 2Decoder

2026/9/19 11:08:06View source

Conflict 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.

Open the story