MagicBeanStalk paper: recursive synthesis of hard problems boosts gpt-oss-120b by 8.6 points on 8 math benchmarks
hyunw_kim · x · 2026-10-16
The MagicBeanStalk paper analyzes from first principles what makes synthetic data useful for RLVR, introducing recursive synthesis that turns hard problems into useful training data. Across 8 math benchmarks it delivers +8.6 points for gpt-oss-120b and +5.4 for DeepSeek v4 Flash. Team includes GXiming Xue, Aviral Kumar, Yejin Choi, and Jan Kautz.
More from Research
- Hugging Face launches AI4Science heatmap tracking open models from 79 orgs, AI2 and OpenMed lead — anshulkundaje · 2026-10-16
- MIT's Neil Thompson: AI comes for tasks before jobs, and fixed costs favor big firms — bigdata · 2026-10-16
- UMD's PAIR framework bridges perception and action in VLA models via bridge tokens — Kaixi Feng · 2026-10-16
- CI-MSE offline metric hits -0.87 Spearman correlation with real robot task success vs weak raw MSE — chris_j_paxton · 2026-10-16
- Colossal Biosciences uses AI to decode genetic switches behind the extinct dodo — castrotech · 2026-10-16
- Colossal Biosciences uses AI to identify gene switches in dodo de-extinction push — castrotech · 2026-10-16