AI
Incremental Open-Ended Deep Research with Structured Harness
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 07/10/2026, 20:00:00 UTC
Briefing created: 17/11/2026, 22:12:18 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 09/10/2026, 15:18:24 UTC.
Briefings
Incremental Open-Ended Deep Research (Incremental-OEDR) treats research reports as evolving states, using Structured Harness to continuously update reports with new information. Experiments on the DeepResearch Bench and DeepConsult show it achieves up to 0.51 higher content-level ROUGE-L F1, 0.63 higher outline-level EM F1, and 33% lower token consumption compared to existing systems, while maintaining competitive quality.
Key points
- Incremental-OEDR treats research reports as evolving states, continuously updating them with new information rather than generating reports from scratch.
- Structured Harness represents reports as structured collections of outlines, sections, and supporting evidence to enable selective updating and evidence reuse.
Context
Existing Open-Ended Deep Research systems generate reports from scratch, making them inefficient for scenarios where reports need continuous maintenance as new information emerges. This finding matters because it offers a method to maintain research continuity and reduce costs while preserving valid knowledge.
Limitations
The study covers ten years of temporal evaluation using Single-Step and Long-Chain Tasks. Results are from two datasets (DeepResearch Bench and DeepConsult) under both Open-source and Proprietary configurations. Specific city, region, or country details are not supported by the evidence.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 17/11/2026, 22:12:18 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
IPFS archive pending.