Meta 论文：前沿智能体从经验中学习的能力差异显著
Meta 冻结模型权重测智能体自学能力：GPT-5.6 Sol 每 1000 美元涨 298 分，Claude Fable 5 只有 57 分但胜率涨幅更大，挺有意思的对比。
Meta 发表论文研究自改进智能体从经验中学习的效率差异。实验冻结模型权重，让智能体在 Chess、Go、Hex 和 NetHack 上反复对局，每轮自行改写工具和笔记，再在保留测试局上评分，并以每美元学习成本带来的保留集提升定义可塑性。在困难棋局中，Claude Fable 5 的保留集胜率从 37.5% 升至 73.3%；综合 Chess、Go 和 Hex，GPT-5.6 Sol 每花费 1,000 美元学习成本获得 298 分提升，Claude Fable 5 仅 57 分。结论是表现最好的模型未必是学习效率最高的那个。
New Meta paper shows that frontier agents differ widely in how they learn from experience, and the best performer is not the most efficient learner.
Most agent tests measure what a model can do once. They ignore whether it gets better by building tools, notes and skills from past attempts, and at what cost.
They froze model weights and let agents play chess, Go, Hex and NetHack, rewriting their own tools and notes after each round. They then scored held-out games and defined plasticity as held-out gain per dollar spent learning.
In hard chess, Claude Fable 5 rose from 37.5% to 73.3% on held-out games. Across chess, Go and Hex, GPT-5.6 Sol gained 298 points per $1,000 of learning cost, versus 57 for Claude Fable 5.
When you evaluate self-improving agents, track held-out learning curves and cost, and check whether the agent actually uses what it saved.
