BWhat changed
Integrated long-context reasoning RL with reusable partial rollouts, policy optimization and multimodal data, and transferred longer reasoning behavior toward shorter-response models through length control and merging.
Scientific scope
Integrated long-context reasoning RL with reusable partial rollouts, policy optimization and multimodal data, and transferred longer reasoning behavior toward shorter-response models through length control and merging. RL, online mirror descent and distillation/merging have predecessors. Report includes long-CoT warmup SFT, so not a no-supervision recipe. Context scaling is observed on specified experiments, not an unlimited scaling law; inference traces are not proof of faithful internal search.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. RL, online mirror descent and distillation/merging have predecessors. Report includes long-CoT warmup SFT, so not a no-supervision recipe. Context scaling is observed on specified experiments, not an unlimited scaling law; inference traces are not proof of faithful internal search.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- Kimi k1.5: Scaling Reinforcement Learning with LLMs
- Kimi Team: partial-rollout long-context RL and long-to-short transfer — author/institutional project
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction