Tag
Presents GraphGym, a semi-automatic framework for constructing complex graph reasoning benchmarks for LLMs, covering five complexity dimensions and evaluating models across text-based, code-based, and augmented settings.
This paper presents a graph-to-answer mechanism-tracing case study for Graph-PRefLexOR-8B, a materials-science hypothesis generation model, using visualization and activation-based diagnostics to localize where mechanism support is lost or recovered during generation.
The paper introduces Operation-wise TableQA, a new task with fine-grained question taxonomy, and proposes SkillTGR, a skill-augmented table graph reasoning framework that uses graph traversal and a hierarchical SkillBank for self-evolving reasoning, achieving superior performance and efficiency.
This paper introduces RSF-GLLM, a framework that decouples differentiable graph reasoning from LLM generation to address the semantic gap in multi-hop knowledge graph question answering, achieving competitive performance with superior inference efficiency.
Introduces GRASP, a framework that combines LLM planning with graph algorithms to generate high-fidelity related work sections by modeling inter-paper relationships through a two-layer graph structure and Steiner tree pruning.
This paper introduces SGR-BIM, a graph-driven semantic reasoning framework that dynamically aligns regulatory intent with BIM geometry to automate geometry-intensive compliance checks, achieving 84.3% accuracy on fire safety code queries.
This paper explores using visual graph mind maps as reasoning scaffolds for LLMs, finding that visual guidance remains effective even without direct answer hints, while textual flattening of graphs loses benefits.
TRN-R1-Zero introduces a post-training framework that enables LLMs to perform zero-shot reasoning on text-rich networks using only reinforcement learning, without supervised fine-tuning or chain-of-thought data.