CoreWeave, NVIDIA Boost Search Agent in Three-Week RL Sprint

CoreWeave, NVIDIA, and You.com spent three weeks post-training the open-weights Nemotron 3.5 Lightning for agentic web search using SFT and RL rollouts, improving BrowseComp accuracy by about 22.9% while cutting tool-call costs by roughly 30%.

2026-10-27 ~ 2026-10-27 · 2 related posts

• CoreWeave, You.com and NVIDIA lift BrowseComp 22.9% in 3-week RL post-training sprint — RichardSocher · 2026-10-27
• You.com, NVIDIA and CoreWeave post-train search model, lifting BrowseComp accuracy 22.94% — _ScottCondron · 2026-10-27