{"slug": "nemo-dcr-bit-exact-delta-compressed-refit-for-scalable-agentic-rl-at-trillion", "title": "NeMo-DCR: Bit-Exact Delta-Compressed Refit for Scalable Agentic RL at Trillion-Parameter Scale", "summary": "NVIDIA's NeMo-DCR introduces bit-exact delta-compressed refit for agentic reinforcement learning at trillion-parameter scale, addressing the 87.5-minute transfer time for a full 1T checkpoint between two AWS regions. The technique targets weight synchronization between training and rollout clusters in disaggregated agentic RL pipelines.", "body_md": "Agentic reinforcement learning (RL) disaggregates training from rollout, so each policy update must reach the rollout clusters before the next batch. Transferring a full 1T checkpoint for such weight synchronization (refit) takes 87.5 min between two AWS regions. Measurements of BF16 training show t", "url": "https://wpnews.pro/news/nemo-dcr-bit-exact-delta-compressed-refit-for-scalable-agentic-rl-at-trillion", "canonical_source": "https://aiflash.com/news/132563/", "published_at": "2026-10-07 07:00:00+00:00", "updated_at": "2026-10-07 07:16:45.373599+00:00", "lang": "en", "topics": ["ai-infrastructure", "machine-learning", "large-language-models", "ai-research"], "entities": ["NVIDIA", "NeMo-DCR", "AWS"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/nemo-dcr-bit-exact-delta-compressed-refit-for-scalable-agentic-rl-at-trillion", "markdown": "https://wpnews.pro/news/nemo-dcr-bit-exact-delta-compressed-refit-for-scalable-agentic-rl-at-trillion.md", "text": "https://wpnews.pro/news/nemo-dcr-bit-exact-delta-compressed-refit-for-scalable-agentic-rl-at-trillion.txt", "jsonld": "https://wpnews.pro/news/nemo-dcr-bit-exact-delta-compressed-refit-for-scalable-agentic-rl-at-trillion.jsonld"}}