LongHarness: Advancing Agentic Reasoning over 100M-Token Corpora via Stateful Evidence Management

Key points

• Existing agent harnesses enable incremental access to these corpora, but acquired evidence and evolving conclusions often remain entangled in a growing interaction history.1
• This implicit state management can lead to two coupled failure modes: unbounded exploration, where retrieval expands the interaction history without resolving decision-relevant information gaps, and evidence-state drift, where previously acquired evidence or subsequent corrections fail to inform the current conclusion.1
• We present LongHarness, a model-agnostic harness that makes the interaction between acquired evidence and evolving conclusions explicit.1
• Compared with direct reasoning using the same backbone, LongHarness improves GLM-5.3-Flash exact match on OfficeQA Pro from 15.8% to 60.2% on V1 and from 4.4% to 61.1% on V2.1

Sources (1)

• [1]LongHarness: Advancing Agentic Reasoning over 100M-Token Corpora via Stateful Evidence Management
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 06:32 AM

  Corpus-level long-context reasoning requires agents to acquire and integrate evidence across document collections that exceed their context windows.

  Existing agent harnesses enable incremental access to these corpora, but acquired evidence and evolving conclusions often remain entangled in a growing interaction history.

Extractive summary: sentences quoted from the sources.