Probing the Knowledge Boundary: An Interactive Agentic Framework for Deep Knowledge Extraction

Yuheng Yang, Siqi Zhu, Tao Feng, Ge Liu, Jiaxuan You · 2026

Read Probing the Knowledge Boundary: An Interactive Agentic Framework for Deep Knowledge Extraction by Yuheng Yang, Siqi Zhu, Tao Feng, Ge Liu, Jiaxuan You (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-00959.txt.

Description

Yuheng Yang and colleagues probe large language models as stored knowledge banks, asking what each model holds and where its edge lies. Their interactive agent frame tests knowledge at varied depths with four adaptive search policies. The haul is then cleaned in three stages: vector filters drop repeats, model judges settle fuzzy overlaps, and domain checks keep only valid units. Recursive Taxonomy proves the strongest search policy in their tests. Larger models recover more knowledge, while tuned models start sharp but fade fast, and general models hold steady over runs.

Catalog record

CREATORYuheng Yang, Siqi Zhu, Tao Feng, Ge Liu, Jiaxuan You
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44621
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.00959

Browse the full library · About the Tower of Babel