Research map / Large language models / Agents, code and math
LLMTM: Benchmarking and Optimizing LLMs for Temporal Motif Analysis in Dynamic Graphs
Bing Hao, Minglai Shao, Zengyi Wo, Yunlong Chu, Yuhang Liu, Ruijie Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add tool-augmented agent (graph algorithms) | GPT-4o-mini · LLMTM Multi-Motif Detection | accuracy: 18.80% -> 98% | Improved |
| add tool-augmented agent | GPT-4o-mini · LLMTM Multi-Motif Count | accuracy: 13.07% -> 99% | Improved |
| add tool-augmented agent | GPT-4o-mini · LLMTM Multi-Motif Detection | avg tokens: 2716.59 -> 12298.63 | Hurt |
| route queries with Structure-Aware Dispatcher instead of random | GPT-4o-mini · Motif Detection 3-star | accuracy: 80.0% -> 88.0% | Improved |
| route queries with dispatcher instead of direct LLM | GPT-4o-mini · Motif Detection 4-clique | accuracy: 50.0% -> 97.5% | Improved |
| dispatcher vs full agent | Structure-Aware Dispatcher · Motif Detection 4-clique | accuracy: 100.0% -> 97.5% | Hurt |
| dispatcher generalizes to held-out triangle motif | Structure-Aware Dispatcher · Motif Detection triangle | accuracy: 77.8% -> 84.8% | Improved |
| dispatcher trained on same motif (Single) vs held-out transfer | Structure-Aware Dispatcher · Motif Detection triangle | accuracy: 86.5% -> 84.8% | Hurt |
| add one-shot example + CoT (ablation) | DeepSeek-Qwen-7B · Motif Classification 3-star | accuracy: 50% -> 90% | Improved |
| add zero-shot CoT only (ablation) | DeepSeek-Qwen-7B · Motif Classification 3-star | accuracy: 50% -> 50% | No clear effect |
| add one-shot example + CoT (ablation) | DeepSeek-Qwen-7B · Motif Classification 4-clique | accuracy: 65% -> 45% | Hurt |
| add one-shot example (ablation) | DeepSeek-Qwen-7B · Motif Classification 4-clique | accuracy: 65% -> 35% | Hurt |
Closest papers by meaning
- How Do Large Language Models Understand Graph Patterns? A Benchmark for Graph Pattern Comprehension (ICLR 2025 · similarity 0.89)
- GraphChain: Large Language Models for Large-scale Graph Analysis via Tool Chaining (NeurIPS 2025 · similarity 0.88)
- Attention Mechanisms Perspective: Exploring LLM Processing of Graph-Structured Data (ICML 2025 · similarity 0.87)
- GraphArena: Evaluating and Exploring Large Language Models on Graph Computation (ICLR 2025 · similarity 0.87)
- GraphRouter: A Graph-based Router for LLM Selections (ICLR 2025 · similarity 0.87)
- DyG-Mamba: Continuous State Space Modeling on Dynamic Graphs (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
