SearchAtlas: Analyzing Agentic Search Strategies via Evidential Query Graphs¶

简体中文

L1 core-mechanism diagnostic. Builds evidence-propagation edges on a public fixture; does not run the paper's five search agents.

Paper information¶

Field	Details
Paper	SearchAtlas: Analyzing Agentic Search Strategies via Evidential Query Graphs
Organization	First-author affiliation: Duke University
First public release	2026-09-09
Authors' code	https://github.com/DukeNLP/SearchAtlas
Adapter / method	`searchatlas`
Local reproduction code	`src/auto_research/agent_research/latest_20260914.py`

Original paper¶

Motivation and method¶

Turn search trajectories into directed query–evidence–answer graphs, then audit whether evidence covers the question constraints and final answer.

flowchart LR
  I[Input and public state] --> M[SearchAtlas mechanism]
  M --> A[Auditable intermediates]
  A --> O[Output or update]

Key figure from the paper¶

Figure 1: Overview of the paper's method and key components. From the original paper; copyright remains with the authors.

Core computation or operation¶

The local code implements the decisive operation as a deterministic reference kernel and explicitly reports gate, selection, projection, graph-edge, or state-update statistics. Tests cover formula invariants and boundary conditions.

Paper results¶

Automatic graph parsing achieves 86.0% mean edge F1 against human annotations. These are paper results, not directly comparable with the scaled-down local run. No online gains are inferred in the absence of an online A/B test.

Local reproduction¶

See the three-seed metrics. All candidates share the public fixtures, budget, and seeds. The results verify actual execution of core state transitions only.

Reproduction boundaries¶

The local diagnostic builds evidence-propagation edges on a public fixture. The paper's five search agents were not run.