Reproduce “HyperIntent: Tracking Intent Shifts for Efficient Reasoning on Dynamic Textual Graphs” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yilin Yao, Bin Lu, Xinhui Li, Meng Jin, Xinbing Wang

HyperIntent groups timestamped interactions into intent hyperedges, then checks semantic and structural novelty against history before updating a node state with LLMs only when a shift is detected. On Googlemap CT, its DyGFormer version reaches 36.02 ± 0.54% inductive Hits@1, versus 1.78 ± 0.06% for DyGFormer alone, while LLM tokens fall 54.4–95.9% versus DyGRASP across six datasets; results are task- and dataset-specific.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.