Query-Aware Learnable Graph Pooling Tokens as Prompt for Large Language Models

Wooyoung Kim, Byungyoon Park, Wooju Kim · 2025

Read Query-Aware Learnable Graph Pooling Tokens as Prompt for Large Language Models by Wooyoung Kim, Byungyoon Park, Wooju Kim (2025) 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-2501-17549.txt.

Description

Wooyoung Kim, Byungyoon Park, and Wooju Kim wrote this paper in two thousand twenty-five. It feeds graph data to large language models as a small set of learned pooling tokens, shaped by the user's query.

Node-by-node projection does not scale, while whole-graph projection washes out detail. The learned tokens strike a balance, holding both fine and broad traits of the graph. A trick called Early Query Fusion blends the question in before the graph code is built. That way the code keeps what the question needs. On the GraphQA benchmark the method gains over four percent. The language model itself is left untrained. The work suits social, citation, and knowledge graphs full of text.

Catalog record

CREATORWooyoung Kim, Byungyoon Park, Wooju Kim
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-17549

Browse the full library · About the Tower of Babel