RL scaling exponent ~0.3 is twice what the rule-of-thumb predicts
tobyordoxford · x · 2026-11-09
Toby Ord adds that the 0.3 RL scaling exponent he derived is about twice as high as what his loose rule of thumb — RL having one third the exponent of pre-training — would give, suggesting RL post-training compute may be more efficient than naive extrapolation suggests.
More from Research
- New benchmarks DecoyCount and ScienceVision probe whether AI sees or reasons — liangchen5518 · 2026-11-09
- AI chip design demo: one prompt for a 64-bit RISC-V processor under 3W TDP — satnam6502 · 2026-11-09
- Video-guided machine translation paper lifts low-resource Indic languages by +1.75 BLEU — kalyan_kpl · 2026-11-09
- Peer-reviewed paper applies truth maintenance systems to LLM memory belief revision — letandrewcook · 2026-11-09
- A cathedral that sings: 1D acoustic sim shows voices fighting their own resonant body — Sauers_ · 2026-11-09
- Survey by 30+ authors maps modern NLP tokenization, from subword wins to its flaws — rdesh26 · 2026-11-09