Reproduce “Teaching LLMs to See Graphs: Unifying Text and Structural Reasoning” 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
Dario Vajda
GTLM gives a pretrained language model graph structure inside attention while preserving node text, using node-local positions and learned biases for graph relationships. On a synthetic four-hop retrieval task with 128 nodes and 512 tokens per node, it reaches 98.0% exact-match accuracy, versus 8.5% for a matched flat-text baseline. This constructed test does not establish general long-context ability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.