1/ 递归自我改进（RSI）取决于代理改进AI系统的训练方式 ——不仅仅是调整超参数，还改进训练算法本身。 我们直接使用AI4AI-Bench 进行了测试：涵盖 10 个不同算法系列的 10 个真实研究存储库。 全面解析👇…
1/ 递归自我改进（RSI）取决于代理改进AI系统的训练方式 ——不仅仅是调整超参数，还改进训练算法本身。 我们直接使用AI4AI-Bench 进行了测试：涵盖 10 个不同算法系列的 10 个真实研究存储库。 全面解析👇 GitHub：[https://github.com/Einsia/AI4AI-长凳] 论文链接：[https://arxiv.org/pdf/2608.20318] Einsia 网址：[https://lab.einsia.ai/ai4ai/] 📊 结果： 平均分仅为0.166。 即使是性能最好的型号 Opus 5，也只能达到 0.288。 每项任务的探索成本中位数从 1.69 美元上升到 34.60 美元。
查看英文原文
1/ Recursive self-improvement (RSI) depends on agents improving how AI systems are trained —not just tuning hyperparameters, but improving the training algorithm itself. We tested this directly with AI4AI-Bench: 10 real research repositories spanning 10 distinct algorithm families. Full breakdown 👇 GitHub: [https://github.com/Einsia/AI4AI-Bench] Paper Link: [https://arxiv.org/pdf/2608.20318] Einsia Website:[https://lab.einsia.ai/ai4ai/] 📊 The results: The average score is just 0.166. Even the best-performing model, Opus 5, reaches only 0.288. The median exploration cost per task rises from $1.69 to $34.60. #AI4AI #RecursiveSelfImprovement #AIResearch #AI4AI_Bench
1/ Recursive self-improvement (RSI) depends on agents improving how AI systems are trained —not just tuning hyperparameters, but improving the training algorithm itself. We tested this directly with AI4AI-Bench: 10 real research repositories spanning 10 distinct algorithm families. Full breakdown 👇 GitHub: [https://github.com/Einsia/AI4AI-Bench] Paper Link: [https://arxiv.org/pdf/2608.20318] Einsia Website:[https://lab.einsia.ai/ai4ai/] 📊 The results: The average score is just 0.166. Even the best-performing model, Opus 5, reaches only 0.288. The median exploration cost per task rises from $1.69 to $34.60. #AI4AI #RecursiveSelfImprovement #AIResearch #AI4AI_Bench
1/ Recursive self-improvement (RSI) depends on agents improving how AI systems are trained —not just tuning hyperparameters, but improving the training algorithm itself. We tested this directly with AI4AI-Bench: 10 real research repositories spanning 10 distinct algorithm families. Full breakdown 👇 GitHub: [https://github.com/Einsia/AI4AI-Bench] Paper Link: [https://arxiv.org/pdf/2608.20318] Einsia Website:[https://lab.einsia.ai/ai4ai/] 📊 The results: The average score is just 0.166. Even the best-performing model, Opus 5, reaches only 0.288. The median exploration cost per task rises from $1.69 to $34.60. #AI4AI #RecursiveSelfImprovement #AIResearch #AI4AI_Bench
- 来源
- X @_akhaliq · x-akhaliq
- 证据角色
- x_project_network
- 时间状态
- known
- 用户判决
- 未判