自我改进智能体因记忆奖励膨胀而退化
A self-improving agent can keep its weights frozen and still get worse through t…
A self-improving agent can keep its weights frozen and still get worse through the memories it learns to trust.
一个自我改进的智能体可以保持其权重冻结，但仍然会通过它学会信任的记忆而变得更糟。
These agents store past episodes, score them with an LLM, and reuse them later as precedents without changing model weights.
这些智能体存储过去的片段，用LLM对它们进行评分，并在以后将它们作为先例重用，而不改变模型权重。
Wrong episodes can still receive high self-scores; across the paper's factual banks, the tested models endorsed 31% to 54% of their own wrong answers as correct.
错误的片段仍然可以获得高的自我评分；在论文的事实库中，测试的模型认可了它们自己31%到54%的错误答案作为正确。
Once that score enters persistent memory, the mistake can influence future decisions instead of disappearing after one bad answer.
一旦该分数进入持久记忆，错误就可能影响未来的决策，而不是在一次错误回答后消失。
The authors call this the Echo Gap.
作者称之为回声差距。
Stronger or different LLMs did not reliably repair it, because their grading errors often remained correlated with the original self-grading bias.
更强或不同的LLM并不能可靠地修复它，因为它们的评分错误往往与原始的自我评分偏差保持相关。
– arxiv. org/abs/2608.00017
– arxiv.org/abs/2608.00017
Title: "Memory Reward Inflation in Self-Improving LLM Agents"
标题：“自我改进LLM智能体中的记忆奖励膨胀”
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力