Language model representations beat descriptors in multi-objective reaction optimization
pschwllr · x · 2026-10-02
A new arXiv paper (2609.11790) from Philippe Schwaller's group skips hand-crafted reaction featurization: a fine-tuned language model encodes textual descriptions of reaction conditions, trained jointly with Gaussian process surrogates inside a multi-objective Bayesian optimization loop to yield task-adaptive representations.
Across nickel- and palladium-catalysed cross-couplings in both sequential and parallel experimentation regimes, the approach reaches optimization convergence in fewer experiments than descriptor libraries or one-hot encodings, and was applied prospectively to a pa... process. Commenters note that putting natural language and tabular data on equal footing for inference could yield gains in many other domains.
More from Research
- Pinocchio: a lightweight model that adds calibrated confidence to frontier LLM outputs — micahgoldblum · 2026-10-02
- Extropic claims 100x-10,000x efficiency in first results on thermodynamic recursive intelligence — beffjezos · 2026-10-02
- ARC Prize finds Qwen3.8-27B's chat template injects different instructions per reasoning effort — GregKamradt · 2026-10-02
- DataScalar: The Failed 90s Memory-Centric Architecture That Aged Well — lauriewired · 2026-10-02
- AAAI review reform targets 'micro result' papers of minor ablations, says Dietterich — tdietterich · 2026-10-02
- Eval builder explains anti-benchmaxxing: rotating all questions and constantly changing methodology — airesearch12 · 2026-10-02