An AUC 0.62 classifier cut 65,578 candidates to 145 DFT runs and found 6 new discoveries
bravo_abad · x · 2026-09-24
A case from Ran et al. in AI for Science: a classifier with AUC of just 0.62 reduced 65,578 candidates to 145 DFT calculations and still yielded six new discoveries.
The takeaway: in scientific tools, the model doesn't need to be maximally accurate — it just needs to rank candidates well enough to make the expensive step dramatically more efficient. A pattern worth carrying beyond the original papers.
More from Research
- DeepMind unveils AlphaGenome 2, a cell-type-conditional genomics model — pushmeet · 2026-11-11
- AlphaGenome 2 commercial API enters Google Cloud private preview, weights promised later — anshulkundaje · 2026-11-11
- Pooling-aware training cuts retrieval vectors 4x while boosting nDCG to 97.8% of unpooled — lateinteraction · 2026-11-11
- Lazarus attack explained: hijacking LLM output by reprogramming one MLP neuron — DimitrisPapail · 2026-11-11
- Nearly 30% of US STEM PhD dissertations now have AI-generated abstracts, NBER study finds — joshgans · 2026-11-11
- Stanford's DyCraft Uses Agentic Decomposition to Cut Physical Errors in Long-Horizon 4D Scene Generation — StanfordUniversity · 2026-11-11