Apple 提出 SelfCompact 自适应上下文压缩
Apple Machine Learning Research 发布论文《Self-Compacting Language Model Agents》,提出 SELFCOMPACT 脚手架:模型自己调用一个压缩工具来总结已累积的上下文,并配合一份轻量评分规则,规定何时触发压缩(某个子任务已完成,或轨迹正在收敛)、何时抑制(推导中途,或陷入停滞)。论文称两者缺一不可——只用工具时开放权重模型调用得并不均匀,只用规则则无法行动——且在六个基准(竞赛数学与智能体搜索)和七个模型上测试,效果达到或超过固定间隔摘要,相对无摘要基线在数学上最高提升 18.1 分、智能体搜索上提升 5–9 分,同时每题成本降低 30–70%,无需微调或外部监督。作者中部分人带有 Johns Hopkins University 标注,论文归入 NeurIPS 内容。
长任务的成本与可靠性很大程度取决于上下文如何被压缩,这项研究提示“何时压缩”或许不必靠训练,而可由脚手架提供。
research area Speech and Natural Language Processingconference NeurIPS
content type paperpublished October 2026
Self-Compacting Language Model Agents
AuthorsTianjian Li†, Jingyu Zhang†, William Jurayj†, Xi Wang†, Chuanyang Jin†, Mehrdad Farajtabar, Eric Nalisnick†, Daniel Khashabi†
Long agent traces composed of chains of thought and tool calls accumulate stale content that anchor subsequent generations, and eventually outgrow the context window. Existing scaffolds mitigate it with fixed-interval compaction triggered at a token threshold. Such triggers pay no heed to trajectory structure, risking discard of partial results mid-derivation or mid-search. We propose SELFCOMPACT, a scaffold that allows the model itself to decide when and how to compact. Specifically, it pairs two inference-time elements: (i) a compaction tool the model invokes to summarize the accumulated context, and (ii) a lightweight rubric specifying when to fire (a sub-task has resolved, or the trajectory is converging) and when to suppress (mid-derivation, or when stuck). Both are needed. The tool alone is unevenly used across open-weight models, often invoked at unhelpful moments or not at all; the rubric alone cannot act. Together, they elicit effective adaptive compaction without any fine-tuning or external supervision. We present empirical results on six benchmarks (competitive math and agentic search) and seven models. Our results show that SELFCOMPACT matches or exceeds fixed-interval summarization at a fraction of the token cost, improving over a no-summarization baseline by up to 18.1 points on math and 5–9 points on agentic search at 30–70% lower per-question cost. Our results expose a meta-cognitive gap: although unprompted models cannot reliably tell when their own context is rotting, a lightweight rubric closes this gap, reframing when to compact as a capability that scaffolds can supply without training.
#Apple#Apple Machine Learning Research#Johns Hopkins University#SelfCompact