PathAnchor: Path-Structured Evidence for Scientific Agents

2026-09-29

AI Summary

This study addresses the loss of functional ordering, source conflation, and conclusion overreach when AI agents retrieve evidence from scientific documents. We propose a bounded reasoning system grounded in a path-structured evidence workspace. Methodologically, independent concepts are replaced with source-linked trajectories to achieve ordered evidence organization and explicit boundary delineation. A read-only tool controller is further introduced to perform trajectory search, cross-source tracing, and precise evidence activation, ensuring the complete preservation of roles, directions, and transitions. Experimental results demonstrate that the proposed system achieves an overall score of 82.6%, outperforming six baselines. It attains a source recall of 82.9% and a citation completeness of 90.0%, while effectively reducing tool-calling overhead.

Abstract

Scientific agents can retrieve relevant passages yet still lose functional order, mix evidence across sources, or state conclusions that exceed the retrieved record. We introduce PathAnchor, a bounded scientific reasoning system built on path-structured evidence workspaces. Instead of treating passages or extracted concepts as independent units, the system retrieves source-linked Material-Sensor-Signal-System trajectories that preserve role, direction, and the evidence supporting each transition. A controller uses three read-only tools to search paper-specific trajectories, trace paths across candidate sources, and open exact evidence before producing a claim-cited answer and an explicit evidence boundary. On 120 single- and cross-paper flexible-sensor questions, PathAnchor scores 82.6% and leads six evaluated systems. Under a matched controller, corpus, and six-call budget, replacing unordered concept graphs with path-structured records raises source recall from 61.3% to 82.9%, increases answers whose claims all cite opened evidence from 69.2% to 90.0%, and reduces tool calls. These results show that evidence organization affects retrieval and citation completeness under fixed agent resources.

Problem

• Scientific agents
• Evidence retrieval
• Source mixing
• Citation completeness
• Functional order

Innovation

• Path-Structured Evidence
• Scientific Agents
• Bounded Reasoning
• Evidence Trajectories
• Citation Completeness