Newest papers

HGM-Dual: Lineage-Aware Guided Agent Harness Optimization for Long-Horizon Planning

Sudipta Paul, Pengrui Quan, Aounon Kumar, Nicholas Tzou, Vivek Kulkarni, Srinivas Chappidi · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.30042 · cited by 0 works (OpenAlex) · open-access PDF

Korean

냉동 또는 닫힌 소스 LLM에 구축 된 에이전트를 향상시키는 점점 더 효과적인 방법은 대상 요원, 도구 및 작업 흐름을 향상시키는 구성 요소 최적화입니다. Darwin--Godel Machine 및 Huxley--Godel Machine (HGM)와 같은 선도적 인 접근 방식은 반복적 자기 개선에 의존하며 코딩을 위해 주로 개발되었으며 에이전트를 향상시키는 데 필요한 능력은 능력과 일치합니다. 에이전트가 상호 작용하는 제약을 공동으로 만족시키는 도구 라인 호출의 시퀀스를 생산해야하는 기본적인 계획 작업은 HGM이 반복적으로 효과적이지 않은 변화를 탐구하고 상대적인 평가 예산을 소비합니다. 우리는 예산 효율적인 계획 에이전트를

Machine translation (Meta M2M100); the English original follows.

Abstract

Harness optimization, which improves an agent's prompts, tools, and workflows, is an increasingly effective way to enhance agents built on frozen or closed-source LLMs. Leading approaches such as the Darwin--Godel Machine and the Huxley--Godel Machine (HGM) rely on recursive self-improvement and were developed mainly for coding, where the competence needed to improve the agent coincides with the competence being improved. Planning tasks, in which an agent must produce a sequence of tool calls that jointly satisfies interacting constraints, break this alignment. There, open-ended self-modification repeatedly explores ineffective changes and consumes substantial evaluation budgets. We introduce HGM-Dual, a diagnostic-guided framework for budget-efficient harness evolution for planning agents. First, HGM-Dual represents agent evolution as lineage-aware context. This context records what each ancestor changed, which behaviors improved or regressed, and which modifications were ineffective, giving the editor explicit edit-outcome attribution instead of an unstructured archive of past agents. Second, HGM-Dual introduces error-conditioned dual-path child generation. For each selected parent, it generates a general child that preserves open-ended exploration and specialized children that target the harness components most likely responsible for the parent's highest-ranked constraint violations. HGM-Dual outperforms the strongest baseline by up to 14.45% relative improvement in case accuracy while matching HGM variants' final performance with 2x-4x fewer evaluation budget.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Competence (human resources), Computer science, Context (archaeology), Artificial intelligence, Sequence (biology), Matching (statistics)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange