TDiMS: fragment distances beat large pretrained models at chromophore property prediction
bravo_abad · x · 2026-09-18
Hamada and coauthors propose TDiMS for predicting molecular optical properties: instead of letting a large pretrained model discover long-range relationships from scratch, they explicitly represent pairs of molecular fragments and their distances along the molecular graph.
- Key insight: molecules with similar chemical groups can behave very differently depending on where those groups sit; chromophore optical properties depend on functional-group arrangement.
- For Stokes shift prediction (difference between absorbed and emitted light), this simple distance-based representation outperforms the second-best descriptor method.
- Takeaway: explicit structural inductive bias wins over purely learned representations when relationships matter.
A solid AI-for-science empirical result.
More from Research
- MoDA: RL alignment method fights LLM mode collapse while preserving output quality — stanfordnlp · 2026-09-19
- Stanford study finds the brain is actually two separate organs working together — Dr_Singularity · 2026-09-19
- Apple researchers propose probe guidance, cutting guidance cost for diffusion LMs with no extra forward pass — itsbautistam · 2026-09-19
- New Science paper shows disorder and heterogeneity can stabilize complex networks — wgilpin0 · 2026-09-19
- Token Superposition Training Cuts Pretraining Compute 2.5x on 10B MoE — gordic_aleksa · 2026-09-19
- First-of-its-kind AI x Med Chem Hackathon in Boston Ends; Compounds Head to Synthesis — generativist · 2026-09-19