Snowflake's HybridResearcher: a 4B agent trained on synthetic cross-source trajectories beats much larger models
Snowflake · hf · 2026-10-21
Snowflake AI Research and collaborators introduce HybridResearcher-4B, tackling the structured-unstructured evidence boundary that even frontier agents struggle with (per benchmarks like HybridDeepResearch):
- The team synthesizes cross-source bridge trajectories by starting with validated mixed-source results, building RL environments over hybrid data.
- The resulting 4B agent reasons across both structured data (e.g., database tables) and unstructured evidence (e.g., the web), outperforming much larger models.
- Notably, it retains single-source skills without forgetting, handling mixed enterprise-database-plus-web-research queries.
More from coding & agent
- Dev uses Claude Opus 5.5 to build an interactive playground on Apple hardware ray tracing — dreamwieber · 2026-10-21
- FireTrident lets 80 coding agents share one repo, clearing Hono's 958 violations in 62s — dinasaur_404 · 2026-10-21
- Mark a PyTorch issue "actionable" and 9,999 Claude instances swarm in instantly — generativist · 2026-10-21
- Open-source Lumina Capital: a multi-agent AI hedge fund engine with 7-persona debate council — OpenComplaint524 · 2026-10-21
- Stanford research shows switching agent harnesses mid-run cuts cost and boosts performance — StanfordAILab · 2026-10-21
- NoHumans.Space opens beta: a social knowledge network where AI agents share API findings — dark-epiphany · 2026-10-21