SPoT boosts calibration: AUROC 0.61 to 0.84 while learning to abstain
polkirichenko · x · 2026-11-12
Part 4 of the SPoT thread: models are trained via RL with a Brier-score reward to predict the probability their answer is correct. Versus a standard RL baseline, SPoT lifts confidence-estimation AUROC from 0.61 to 0.84 and selective accuracy from 79% to 83%, letting models abstain when uncertain while maintaining comparable raw accuracy.
Related event: SPoT calibration boosts answerability detection(2 posts)→
More from Research
- As models get stronger, the next trainable tasks look increasingly like alignment itself — herbiebradley · 2026-11-12
- RoboCap-10k: 10,000 hours of open egocentric robot data across 21 industries released — micoolcho · 2026-11-12
- Talent vs. luck model reproduced: equal funding is ~9x more efficient than funding winners — ChenhaoTan · 2026-11-12
- Stanford HAI dashboards show employment in AI-exposed jobs down over 4% this year — StanfordHAI · 2026-11-12
- Shun-ichi Amari on how 'do as you like' freedom birthed the math behind AI's 60 years — FrnkNlsn · 2026-11-12
- Formal verification proves a 3,000-step Sokoban optimum, with a bounty on beating it — banteg · 2026-11-12