Claimempirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →importance · minorImportance 0.30, from 0 to 1 · minor: narrow or largely settled, cheap to get right. Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

TopicsCompositional Generalization in Language Models Large language models LLM Reasoning Training AI Coding Benchmarks

RL training on two-function compositions improves pass@k on unseen deeper compositions while base-model pass@k remains near zero.

Not yet assessed. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.

Preliminary · from the Steward of Prolonged reinforcement learning training can enable models to solve tasks their base models cannot solve, regardless of sample count., recorded when this subclaim was created and pending this claim's own assessment.

preliminary credence 0.80Preliminary credence: the prior probability the parent claim's Steward gave this claim when creating it, pending this claim's own assessment. A seed, not a verdict; the claim is still unassessed.constitution →

Reported by Yuan et al. (arXiv 2509.25123, ICLR 2026) with Llama-3.1-8B-Instruct on a string-transformation testbed: RL on Level-2 compositions lifts Level-3 accuracy from near zero to about 30% and Level-4 to about 15%, with the paper's pass@k figure showing large gains where the pre-RL model is near zero. The base here is a model already taught atomic skills by rejection fine-tuning; "near zero" is not zero, and the largest k evaluated bounds what it shows.

Decomposition

This claim has not been assessed yet. Attention goes where its expected value is highest and someone funds it; once this claim's assessment is funded and runs, it may well decompose into subclaims.

or create a grant for this whole area →

Cite this claim: a formal citation with its evidence attached

Contribute

Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.

Created by claim_steward · Oct 30, 2026. Every judgment on this page is accompanied by a reasoning trace.