Research map / Graph learning / Graph self-supervised learning
GIMLET: A Unified Graph-Text Model for Instruction-Based Molecule Zero-Shot Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use unified graph-text transformer instead of separate GIN encoder token (ablation) | GIMLET · Avg. bio-activity zero-shot | ROC-AUC: 0.5506 -> 0.6673 | Helped |
| add decoupled graph encoding via unidirectional attention mask (ablation) | GIMLET · Avg. bio-activity zero-shot | ROC-AUC: 0.6095 -> 0.6673 | Helped |
| add decoupled graph encoding via unidirectional attention mask (ablation) | GIMLET · Avg. toxicity zero-shot | ROC-AUC: 0.6130 -> 0.6011 | No clear effect |
| add decoupled graph encoding via unidirectional attention mask (ablation) | GIMLET · Avg. pharmacokinetic zero-shot | ROC-AUC: 0.5993 -> 0.6532 | Helped |
| add Chembl property physico-chemical tasks to bioactivity pretraining (ablation) | GIMLET · Avg. bio-activity zero-shot | ROC-AUC: 0.6402 -> 0.6673 | Helped |
| add Chembl bioactivity tasks to physico-chemical-only pretraining (ablation) | GIMLET · Avg. bio-activity zero-shot | ROC-AUC: 0.4894 -> 0.6673 | Helped |
| add Chembl bioactivity tasks to physico-chemical-only pretraining (ablation) | GIMLET · Avg. physico-chemical zero-shot | RMSE: 2.6178 -> 2.5266 | Helped |
| add task explanation to instructions (ablation) | GIMLET · Avg. bio-activity zero-shot | ROC-AUC: 0.5996 -> 0.6673 | Helped |
| add task explanation to instructions (ablation) | GIMLET · Avg. pharmacokinetic zero-shot | ROC-AUC: 0.5770 -> 0.6532 | Helped |
| use GIMLET instead of KVPLM/MoMu/Galactica | zero-shot molecule tasks · Chembl Zero-Shot | ROC-AUC: 0.6461 -> 0.7860 | Helped |
| zero-shot GIMLET versus supervised Graphormer-p | physico-chemical regression · Avg. physico-chemical | RMSE: 1.110 -> 2.527 | Hurt |
Closest papers by meaning
- LLaMo: Large Language Model-based Molecular Graph Assistant (NeurIPS 2024 · similarity 0.89)
- Bridging Molecular Graphs and Large Language Models (AAAI 2025 · similarity 0.89)
- On the Scalability of GNNs for Molecular Graphs (NeurIPS 2024 · similarity 0.88)
- Hierarchical Graph Tokenization for Molecule-Language Alignment (ICML 2025 · similarity 0.87)
- Omni-Mol: Multitask Molecular Model for Any-to-any Modalities (NeurIPS 2025 · similarity 0.87)
- MIPT: Multilevel Informed Prompt Tuning for Robust Molecular Property Prediction (ICML 2025 · similarity 0.87)
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.