RL vs pretraining compute have different power-law exponents on benchmarks
tobyordoxford · x · 2026-11-14
Toby Ord (Oxford) argues in a thread with Ramez Naam, Ben Todd and Allan Dafoe that the slopes of benchmark score vs log(compute) differ for RL compute, inference compute and pretraining compute, implying different exponents in the resulting power laws. Consequently, the odds of succeeding at a random benchmark task grow exponentially in log(compute) — i.e., odds follow a power law in compute.
Related event: Oxford's Toby Ord: RL Compute Needs 10,000x Boost to Lift Benchmark Scores(17 posts)→
More from Research
- NeurIPS 2025 best paper: diffusion models start memorizing at ~300 steps per image — Mastbubbles · 2026-11-14
- TestPrune: pruning regression tests to unblock AI-driven autonomous bug fixing — burkov · 2026-11-14
- Toddlers learn new verbs only from responsive partners; a call to rerun the test with AI — AnnaCiaunica · 2026-11-14
- Context beats model choice: Databricks benchmark shows 91% vs 58% for data agents — CautiousUse8597 · 2026-11-14
- MIT professor to showcase autonomous AI agent swarms for metamaterials discovery — ProfBuehlerMIT · 2026-11-14
- Walden talk: base models improving fast, mid-training data commoditizing — ZeYanjie · 2026-11-14