Learnijoy
ResearchAI ResearchAI Engineering & DevTools

ReCurveflow Predicts Chemical Transition States with Curved Trajectories

Seungheun Baek, Mogan Gim, Jaewoo Kang· August 24, 2026 View original

Key takeaways

  • ReCurveflow improves transition state prediction by modeling curved reaction trajectories.
  • Its off-path correction enhances robustness and accuracy in predicting chemical reaction mechanisms.
  • The framework outperforms existing methods across various metrics and qualitative analyses.
  • Accurate transition state prediction is crucial for drug discovery and materials design.

Who benefits

PharmaceuticalsChemicalsMaterials ScienceBiotechnology

Summary

This paper introduces ReCurveflow, a novel flow matching framework that improves the prediction of chemical reaction transition states by learning continuously curved reaction paths. It outperforms existing methods by incorporating off-path correction and generating more accurate energy profiles.

Predicting transition states (TS) in chemical reactions is vital for understanding reaction mechanisms. Traditional methods often simplify reaction paths as straight lines, which doesn't accurately reflect the complex, curved trajectories molecules follow. This research presents ReCurveflow, a new flow matching framework designed to overcome this limitation. ReCurveflow learns to predict TS geometries by being supervised on continuously curved reference paths, derived from detailed molecular geometry bands. A key innovation is its "off-path correction" feature, which allows the model to generate corrective velocity fields when encountering geometries not directly on the learned path, enhancing robustness and prediction accuracy. Evaluated across multiple datasets and metrics, ReCurveflow consistently outperforms seven baseline methods. Qualitative analyses further confirm its ability to generate reaction trajectories with energy profiles closely matching real-world paths, providing better initializations for optimization, and demonstrating effective corrective behavior.

Why it matters

For professionals in chemistry, materials science, and drug discovery, accurately predicting transition states can accelerate research, optimize synthesis pathways, and design new molecules with desired properties. This tool offers a more precise and robust approach.

How to implement this in your domain

  1. 1Explore the ReCurveflow codebase for integration into computational chemistry workflows.
  2. 2Apply ReCurveflow to predict transition states for novel chemical reactions or drug design candidates.
  3. 3Compare ReCurveflow's predictions with existing TS prediction methods to validate its accuracy for specific applications.
  4. 4Utilize the generated reaction trajectories and energy profiles to gain deeper insights into reaction mechanisms.
  5. 5Collaborate with computational chemists to leverage ReCurveflow for optimizing synthetic routes or catalyst design.

Original post by Seungheun Baek, Mogan Gim, Jaewoo Kang

"arXiv:2608.20869v1 Announce Type: new Abstract: Predicting transition states (TS) in chemical reactions is crucial, as they provide insights into reaction mechanisms. Recent work on TS prediction have focused on flow matching supervised on straight linear paths that do not align…"

View on X

Primary sources

Originally posted by Seungheun Baek, Mogan Gim, Jaewoo Kang on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Research

AI ResearchAI Engineering & DevTools

Harmony Improves Protein-Ligand Flexible Docking with Torsional Diffusion

Researchers introduce Harmony, a harmonic torsional diffusion framework for flexible protein-ligand docking that explicitly accounts for the periodic geometry of angular variables. This method improves ligand pose accuracy and pocket all-atom reconstruction on benchmarks like PDBBind and enhances the physical validity of generated complexes on PoseBusters.

Maksim Zhdanov, Pavel Strashnov, Vladislav KurenkovAug 24, 2026
AI Engineering & DevToolsAI Research

Multilingual Verifier Bias Impacts RLVR in LLM Mathematical Reasoning

A study reveals that exact-match verifiers in Reinforcement Learning with Verifiable Rewards (RLVR) for Large Language Models (LLMs) exhibit significant language-dependent false-negative reward noise in multilingual mathematical reasoning. This bias, particularly pronounced in Japanese, stems from format and script variations, highlighting a cross-lingual selection bottleneck that impedes effective multilingual LLM training.

Chenyu Zhou, Qiliang Jiang, Xu ZhouAug 24, 2026
AI Engineering & DevToolsAI Research

TriPLU Improves Tiny Language Model Performance with Trilinear Product FFNs

Researchers introduce TriPLU, a Trilinear Product Linear Unit, which replaces gated FFNs in tiny decoder-only language models with a direct degree-3 product branch. This approach achieves better validation loss on character-level TinyStories and lower bits per byte on other datasets under low-learning-rate settings, suggesting benefits for small models in specific low-compute regimes.

He ZhangAug 24, 2026