字节跳动论文：保留试错历史比生成摘要更能提升推理性能
New ByteDance paper shows for test-time improvement, keeping the messy history o…
Agent 开发必读，这篇论文用数据证明保留原始试错轨迹比强行压缩记忆更有效，直接优化了 Agent 的上下文策略与成本控制。
New ByteDance paper shows for test-time improvement, keeping the messy history of attempts can work better than turning that history into a neat memory summary.
字节跳动最新论文表明，在测试时改进方面，保留杂乱无章的尝试历史可能比将其整理为简洁的记忆摘要效果更好。
Chain-of-Experience keeps earlier attempts and feedback in context, then asks the model to try again. Across 6 math, coding, and knowledge benchmarks, self-feedback averaged 71.0%, versus 66.8% with iterative solving but no feedback; correctness or executor feedback reached 79.3%.
Chain-of-Experience（经验链）将早期的尝试和反馈保留在上下文中，然后要求模型再次尝试。在6项数学、编码和知识基准测试中，自我反馈的平均准确率为71.0%，而迭代求解但无反馈的情况为66.8%；使用正确性或执行器反馈则达到79.3%。
The paper also reports 5.6% overall improvement with 19% lower API cost across tasks and models when feedback is used.
该论文还报告称，在使用反馈的情况下，整体性能提升了5.6%，且跨任务和模型的API成本降低了19%。
No weights change here, so this is contextual adaptation rather than persistent learning. Self-feedback also hurt on BrowseComp-Plus when solving the task required external search.
此处没有权重变化，因此这属于上下文适应而非持久学习。当任务求解需要外部搜索时，自我反馈在BrowseComp-Plus上反而造成了负面影响。
For agents, preserve the trajectory, add reliable feedback, and compress only when you know what can safely disappear.
对于智能体而言，应保留轨迹，添加可靠的反馈，仅在明确知道哪些内容可以安全移除时才进行压缩。
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