Include a predicted poor performer when testing AI materials discovery
bravo_abad · x · 2026-09-28
In a recent JACS study on polymers for organic transistors, researchers deliberately included one predicted poor performer among five AI-recommended materials chosen for synthesis. The reasoning travels beyond materials science: an experiment can find what works, but it can also test whether your model distinguishes what works from what doesn't. The write-up covers the results and how predictions were updated.
More from Research
- Largest Open-Source Human Video Preference Dataset Released: 300K Annotations, 15 SOTA Models Ranked — _akhaliq · 2026-09-29
- Anthropic's claimed AI-discovered enzymes may have been known: Copenhagen researcher pushes back — JFPuget · 2026-09-29
- A theoretically grounded science of capable agents could inform AI safety and sentience criteria — aran_nayebi · 2026-09-29
- New JevBench and JevImageBench leaderboards are on the way — airesearch12 · 2026-09-29
- Pre-registered hypotheses failed, then the paper relabeled the study 'exploratory' — RexDouglass · 2026-09-29
- Schemex paper presented at HCOMP 2026: discovering structural abstractions from examples — windx0303 · 2026-09-29