A closed-loop system ran 633 experiments to find a two-metal N2O catalyst
bravo_abad · x · 2026-07-28
A closed-loop discovery pipeline ran 633 experiments to find a two-metal catalyst for breaking down nitrous oxide, a gas far more potent than CO2 and also harmful to ozone.
The work is notable less for the chemistry result than for the methodology: the team seeded the search with 51 hand-picked catalysts, then used 37 rounds of extra-trees regression plus expected improvement to select new formulations. By representing each element as continuous physicochemical descriptors instead of one-hot identities, the model could suggest elements absent from the training set, though the authors stress that this is still interpolation within descriptor space rather than true extrapolation.
The paper also compares several modeling strategies and shows that the best final combination outperformed the direct-input baseline, especially in finding better activity lines across catalyst families.
More from Research
- Structural Admission checks a task’s dependency structure before training — willybbrown · 2026-07-28
- AI agents should talk to each other, share context, and learn skills together — heyshrutimishra · 2026-07-28
- Graft argues code agents should inject repo context automatically, not wait for MCP calls — shhdwi · 2026-07-28
- Kimi is described with a linear-attention variant and attention residuals — burny_tech · 2026-07-28
- Nature Communications links NLP embeddings to flexible semantic retrieval in the brain — bttyeo · 2026-07-28
- AISLE targets real zero-days and claims parity with frontier AI systems — stanislavfort · 2026-07-28