Harness^2: Recursive Agent Harnessing for an Open World

arXiv:2610.14224v1 Announce Type: new Abstract: LLM agents' capabilities depend not only on their underlying models but also on their execution harnesses. Recursive self-improvement (RSI) offers a promising approach to automatically evolve these harnesses. However, deploying RSI in open-world settings still faces two critical challenges. First, open-world tasks are typically unseen during the harness design phase and novel at test time, leaving no ground-truth verifiers to reliably guide online harness edits. Second, the RSI process remains highly task-specific and fails to generalize to novel scenarios. We introduce Harness^2, a plug-and-play, recursive agent harnessing framework that addresses these challenges through two connected levels of recursion. Task-level recursion refines the current task's harness using a contrastive Propose-Probe-Compose step that reliably guides edits without ground-truth verifiers, and can be scaled further via parallel or sequential recursion. Harnessing-level recursion then extracts experience from these edits to update a continual improvement harness, enabling the agent to better adapt to future tasks. Across three professional benchmarks and four model-harness configurations, a single step consistently improves GLM-5 and Gemini 3.5 Flash over their base harnesses, yielding gains of up to 11.5 points on WorkBuddy-Bench, 4.5 points on JobBench, and 11.8 percentage points on LAB, with further gains as the task-level recursion budget increases.

Harness^2 improves AI performance in open-world settings, but implications for human flourishing are unclear.

This is an incremental capability paper on recursive agent self-improvement, evidenced by concrete benchmark numbers but on narrow, unspecified benchmarks without peer review. It has mild long-term relevance to autonomy and oversight questions but no immediate societal implication beyond standard capability advancement.

This paper introduces a recursive self-improvement framework for AI agents, allowing them to adapt to novel tasks without human intervention or predefined ground truth. While framed as an enhancement, the development of autonomous agents capable of continual, unverified self-improvement in open-world settings presents material societal risks. The gains are measured in points on benchmarks, indicating improved capability.

SOURCE     arXiv cs.AI (tier 1)
   ↓
DOCUMENT   c3f16532-4d86-4f91-ba98-9c9a118d07bc
           https://arxiv.org/abs/2610.14224
   ↓
EVIDENCE   1 extracted excerpt
   ↓
MODEL RUN  3 runs, methodology 1.0.1
   ↓
SCORE      -50  (Adverse)
   ↓
CONFIDENCE 72%