Loading…
Newsroom
Hypernetworks outscale LoRA for train-time knowledge injection in LLMsReleases · 7/15/2026Staleness-Adaptive Trust Region cuts asynchronous RL performance loss to 3% at 8× policy lagReleases · 7/15/2026Distilled RL transfers knowledge across model families without unconditional imitationResearch · 7/15/2026Qwen-Music generates full vocal songs from text and lyricsReleases · 7/15/2026LongStraw trains RL models at 2.1M tokens on eight H20 GPUsReleases · 7/15/2026ShortOPD cuts pruned LLM recovery time by 75% while raising generation quality 9×Releases · 7/15/2026