Research map / Vision-language and multimodal / Human and embodied modeling
RobustRDP: Advancing Reaction Diagram Parsing via Synthetic-to-Real Data Scaling and Robustness-Oriented Training
Jianting Tang, zezhong wu, Linli Xu
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 DPO stage (ablation) | RobustRDP · RxnScribe-test | hard match F1: 80.9 -> 81.7 | Improved |
| add DPO stage (ablation) | RobustRDP · RxnScribe-test | soft match F1: 90.0 -> 90.8 | Improved |
| add synthetic-data pretraining (ablation) | RobustRDP · RxnScribe-test | hard match F1: 78.8 -> 80.9 | Improved |
| add synthetic-data pretraining (ablation) | RobustRDP · RxnScribe-test | soft match F1: 88.5 -> 90.0 | Improved |
| add RGRP and PPRP auxiliary SFT tasks (ablation) | RobustRDP · RxnScribe-test | hard match F1: 74.6 -> 78.8 | Improved |
| add RGRP and PPRP auxiliary SFT tasks (ablation) | RobustRDP · RxnScribe-test | soft match F1: 85.9 -> 88.5 | Improved |
| add PPRP alone (prefix-perturbed parsing) (ablation) | RobustRDP w/o DPO and pretraining · RxnScribe-test | hard match F1: 74.6 -> 77.6 | Improved |
| add RGRP alone (region-guided parsing) (ablation) | RobustRDP w/o DPO and pretraining · RxnScribe-test | hard match F1: 74.6 -> 75.7 | Improved |
| add 3,000 newly annotated RobustRDP-train images (ablation) | RobustRDP · RxnScribe-test | hard match F1: -4.2% hard F1 without them | Improved |
| replace RxnCaption-VL with RobustRDP | RobustRDP · RxnScribe-test | hard match F1: 72.2 -> 81.7 | Improved |
| replace RxnScribe with RobustRDP | RobustRDP · RobustRDP-test | hard match F1: 65.5 -> 82.5 | Improved |
| replace RxnIM with RobustRDP | RobustRDP · RobustRDP-test | hard match F1: 53.6 -> 82.5 | Improved |
Closest papers by meaning
- ReactGPT: Understanding of Chemical Reactions via In-Context Tuning (AAAI 2025 · similarity 0.87)
- TinyChemVL: Advancing Chemical Vision-Language Models via Efficient Visual Token Reduction and Complex Reaction Tasks (AAAI 2026 · similarity 0.85)
- Uncertainty-Aware Yield Prediction with Multimodal Molecular Features (AAAI 2024 · similarity 0.85)
- RETRO SYNFLOW: Discrete Flow-Matching for Accurate and Diverse Single-Step Retrosynthesis (NeurIPS 2025 · similarity 0.84)
- Reaction Graph: Towards Reaction-Level Modeling for Chemical Reactions with 3D Structures (ICML 2025 · similarity 0.84)
- Reaction Prediction via Interaction Modeling of Symmetric Difference Shingle Sets (NeurIPS 2025 · similarity 0.84)
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.
