← 研踪 PaperTrailarXiv:2609.33822 · v2
Vestrum: Learning the Verification, Retrieval and Structure of Agent Harnesses from a Few Long Episodes
首次提交 2026-09-28 02:20
版本更新 2026-11-16 11:43
雷达观测 2026-11-17 17:12
摘要
Agent harnesses decide what the model retrieves, decomposes, remembers and verifies, and most people use harnesses that are pre-built and fixed, such as Claude Code or a published memory system. Harness optimizers rewrite their own harness code rather than editing pre-built systems, and they learn from tens to hundreds of tasks, while long-horizon work offers a handful of episodes, each hundreds of tool calls long. We present Vestrum, which improves existing harnesses component by component. It reads full raw traces, states each failure as a recognizable class, and changes the one component where the error arose (a self-verification, a retrieval view, a decomposition step or a memory operation). Each change targets a failure that recurs across tasks of the same kind. Before anything runs, a generalization filter refuses changes that copy facts from the development tasks. Vestrum learns from one to ten long development episodes and improves tasks it has never seen in all five settings. On Claude Code, Vestrum raises Terminal-Bench 4 Hard from 63.7% to 70.3% of test checks at 1.03x the test cost. It raises expert cell-type agreement from 67.5% to 77.8% on held-out sections of one slide. On the GAM memory system, it raises UltraHorizon from 47.6 to 59.8, AMA-Bench from 64.2% to 74.2% and LoCoMo from 71.5% to 74.9%. It beats the GEPA prompt search we ran on the same data in every paired evaluation, and these gains hold under a test model different from the one Vestrum's search used. Across the searches, every self-verification the loop kept acts early: it checks an intermediate result against recorded evidence before the result is committed, or, on Terminal-Bench, requires an invariant and a probe before the first edit. In development, auditors that re-examined every finished answer broke more correct answers than they fixed, and the loop did not keep them.
作者备注:27 pages, 2 figures, 20 tables. Under review
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v2 · 2026-11-16 11:43 · 最新观测版本
Vestrum: Learning the Verification, Retrieval and Structure of Agent Harnesses from a Few Long Episodes
Agent harnesses decide what the model retrieves, decomposes, remembers and verifies, and most people use harnesses that are pre-built and fixed, such as Claude Code or a published memory system. Harness optimizers rewrite their own harness code rather than editing pre-built systems, and they learn from tens to hundreds of tasks, while long-horizon work offers a handful of episodes, each hundreds of tool calls long. We present Vestrum, which improves existing harnesses component by component. It reads full raw traces, states each failure as a recognizable class, and changes the one component where the error arose (a self-verification, a retrieval view, a decomposition step or a memory operation). Each change targets a failure that recurs across tasks of the same kind. Before anything runs, a generalization filter refuses changes that copy facts from the development tasks. Vestrum learns from one to ten long development episodes and improves tasks it has never seen in all five settings. On Claude Code, Vestrum raises Terminal-Bench 4 Hard from 63.7% to 70.3% of test checks at 1.03x the test cost. It raises expert cell-type agreement from 67.5% to 77.8% on held-out sections of one slide. On the GAM memory system, it raises UltraHorizon from 47.6 to 59.8, AMA-Bench from 64.2% to 74.2% and LoCoMo from 71.5% to 74.9%. It beats the GEPA prompt search we ran on the same data in every paired evaluation, and these gains hold under a test model different from the one Vestrum's search used. Across the searches, every self-verification the loop kept acts early: it checks an intermediate result against recorded evidence before the result is committed, or, on Terminal-Bench, requires an invariant and a probe before the first edit. In development, auditors that re-examined every finished answer broke more correct answers than they fixed, and the loop did not keep them.
查看此版本解读 →查看此版本原文 ↗