On RL envs for math: informal methods generalize, formal ones don't
OfirPress · x · 2026-10-10
Will Dequeant argues formal methods don't generalize to informal domains while informal methods generalize to formal ones — solving "loosely verifiable problems" generally beats solving only Lean-expressible ones. OfirPress follows up asking whether such RL envs are just natural-language prompts with an LLM-as-a-judge verifier, either via formalizing into Lean or reading the proof.
More from Research
- EMNLP paper: diffusion LMs internally encode latent denoising timestep representations — gsarti_ · 2026-10-23
- KU Leuven and VIB open two Group Leader professorships in AI for Biology, start 2027 — jpillowtime · 2026-10-23
- Subspace Tuning curbs LLM catastrophic forgetting by freezing principal singular directions — scottniekum · 2026-10-23
- Perceptic launches PercepticOS, an AI workspace for 'scientific superintelligence' — nathanbenaich · 2026-10-23
- EPFL's Bosselut to talk long context as test-time learning at ELLIS seminar — andre_t_martins · 2026-10-23
- DRATS: Adaptive Task Sampling Boosts Worst-Task Performance in Multi-Task RL — Cohere · 2026-10-23