通过对比多模型行为自进化生成程序记忆，显著提升计算机操作成功率。
CONTRAMEM: Learning Self-Evolving Procedural Memory from Contrasting Multi-Model Trajectories
- 利用不同模型在相同任务中的行为差异作为监督信号，构建可自我优化的程序记忆。
- 在GAIA2/ARE测试中成功率达55.3%，较原始模型翻倍以上，跨模型迁移有效。
- 适用于多种智能体，尤其适合需要长期协调决策的复杂任务场景。
自主计算机使用代理在执行长周期任务时，常因误读状态、工具语义或任务进度而失败。程序记忆有助于提升决策一致性并减少重复探索，但无需训练即可构建高质量记忆仍具挑战。本文提出CONTRAMEM，一种无需训练、来源灵活的自进化程序记忆框架：将同一任务下不同模型输出的结果差异（正确性、效率、恢复能力、失败模式）视为监督信号，从中提炼出紧凑的指令级功能卡与任务级技能卡，并通过局部精炼而非简单追加或全量重写实现持续演化。在未见的GAIA2/ARE任务上，该框架使三种目标模型的成功率从26.2%提升至55.3%，各模型均有显著增益（GPT-5.5：27.5%→61.0%；Claude Sonnet 4.6：28.0%→52.5%；DeepSeek V4 Pro：23.0%→52.5%）。同一记忆库直接迁移至未见的Qwen3.7 Plus（18.5%→35.5%），证明其具备可迁移的程序知识。在AppWorld环境中，该方法优于无记忆及单源自记忆版本，对三个中等水平代理均表现更优。在相同轨迹预算下，异构多模型轨迹生成的记忆强于同源或单一模型的多轮次记忆，优势源于对比性行为多样性，而非更强源模型或更多采样。
原文摘要 · Abstract (English)
Autonomous computer-use agents are increasingly applied to long-horizon tasks requiring coordinated application calls, persistent state tracking, and verifier-sensitive writes, yet they remain prone to procedural failures: misreading application state, tool semantics, or task progress. Procedural memory promises more consistent decisions and less redundant exploration, but constructing high-quality memory without model training remains challenging. We introduce CONTRAMEM, a source-flexible, training-free framework for self-evolving procedural memory that treats same-task outcome variation as supervision: differences in correctness, efficiency, recovery, and failure modes expose outcome-relevant procedural distinctions, distilled into a compact bank of app-level Function Cards and task-level Skill Cards that evolves through localized curation rather than append-only accumulation or whole-bank rewriting. On held-out GAIA2/ARE computer-use tasks, CONTRAMEM more than doubles the success rate across the three source-model targets (26.2% to 55.3%), with consistent per-model gains (GPT-5.5: 27.5 to 61.0; Claude Sonnet 4.6: 28.0 to 52.5; DeepSeek V4 Pro: 23.0 to 52.5). The same bank transfers unchanged to the unseen Qwen3.7 Plus (18.5 to 35.5), indicating transferable procedural knowledge rather than model-specific behavior. The same construction carries over unchanged to AppWorld, beating both no memory and its own single-source self-memory variant for all three mid-tier agents on both public test splits. Under a matched trajectory budget, heterogeneous multi-model trajectories yield stronger memory than self- or same-model multi-rollout memory: the margin comes from contrastive behavioral diversity, not stronger source agents or more sampling.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。