wuyang
1.5 KiB
1.5 KiB
Paper: Compiled Memory: Not More Information, but More Precise Instructions for Language Agents
type: paper title: "Compiled Memory: Not More Information, but More Precise Instructions for Language Agents" authors: James Rhodes, George Kang year: 2026 venue: arXiv url: https://arxiv.org/abs/2603.15666 code_url: source: arxiv collected_at: 2026-07-08 published_at: 2026-03-12 updated_at: 2026-03-12 status: skimmed relevance: high topics:
- memory
- rag methods:
- verified-experience-distillation
- instruction-rewriting
- promotion-gate benchmarks:
- CUAD
- HotpotQA models:
datasets:
- cs.AI related_concepts:
- procedural-memory
- prompt-compilation related_jobs:
related_experiments:
- KC-001-agent-memory-pilot related_projects:
- learning/agent-memory collection_score: 13 collection_queries: language-agent
One-line Takeaway
对稳定重复任务，经过验证的经验可以被编译成行为规则，而不必每次作为 retrieved context 注入。
Pilot Skim
- problem: 有更多历史不代表 Agent 知道哪些经验应改变未来行为。
- method: Atlas 用多层 promotion 和 verification 把经验写成 system prompt 子规则。
- evidence: CUAD token F1 +8.7pp，HotpotQA joint F1 +3.16pp；跨模型复用同一 prompt 时 +2.31pp。
- boundary: evolved prompt 更长且缺少 length-matched control；改进严格受训练信号覆盖范围限制。