FractAL: Soft Strategy Selection for Batch-Mode Active Learning in AI-Guided Experiments
anshulkundaje · x · 2026-10-08
Researchers in the Kundaje group (Stanford) note that getting started with AI-guided experimental design in the wet lab is like shooting in the dark: the end goal, the predictive problem setup, and the acquisition strategy are all open questions. Rushil Gupta introduces FractAL, a soft-strategy selection method for batch-mode active learning — instead of committing to a single acquisition strategy before labeling, FractAL adapts across strategies, since the best one varies with dataset, model, and task. A preprint is available.
More from Research
- Stanford paper: decentralized multi-agent DeLM beats Claude Code and Codex, 2.49× faster — rohanpaul_ai · 2026-10-08
- Reason to Play dataset officially released on Hugging Face — somnirons · 2026-10-08
- RoomTSC: a scrappy SF lab using AI models to chase room-temperature superconductors — CatAstro_Piyush · 2026-10-08
- OpenSLA shows zero-shot prediction of unseen actions and future physiological states — yang_yuzhe · 2026-10-08
- UCLA Team Releases OpenSLA: Unifying Sensor Time Series, Language and Actions — yang_yuzhe · 2026-10-08
- Don't Trust a Single Score: 7 Questions to Ask Before Believing Any AI Benchmark — goyalshaliniuk · 2026-10-08