Two GP models turn failed battery experiments into signal, additive holds 90.3% after 50k cycles
bravo_abad · x · 2026-11-10
Researchers used two Gaussian process models for battery electrolyte search: one predicting whether a formulation yields a measurable result, the other predicting performance from valid experiments only. This lets failed experiments still inform the search. The identified low-dose electrolyte additive retained 90.3% capacity after 50,000 high-rate cycles in lab tests — and the modeling choice of treating feasibility as its own signal is the broader lesson.
More from Research
- MIT's McDermott lab: humans infer physical properties from sound using internalized statistics of how objects make sound — JoshHMcDermott · 2026-11-10
- Philosophy Paper Poses Embodiment Challenge: True Cognitive AI Requires Artificial Life First — ArtificialOther · 2026-11-10
- WorldFoundry: open-source world-model infrastructure with 270 models and 74 benchmarks — pmttyji · 2026-11-10
- Amazon FAR's Pick Anything: single policy grasps 66 objects, zero teleop data — notmahi · 2026-11-10
- LeCun clarifies JEPA: the real split is joint embedding vs generative, not the predictor — jbhuang0604 · 2026-11-10
- Stanford AI100 2026 report ships with a section on recursive self-improvement — conitzer · 2026-11-10