Iris: Open-Source Search Agents Hit SOTA via SFT + RL Against Live Search
AllSpark-Research · hf · 2026-09-07
AllSpark Research releases Iris, two large-scale search agents trained with a multi-stage pipeline combining supervised fine-tuning and reinforcement learning against live search.
Key engineering details include rigorous trajectory filtering and inference-time context management. The models achieve state-of-the-art open-source results on complex web search benchmarks.
More from coding & agent
- Codex built and runs a local anime pipeline: SD + LoRA → Wan 2.2 on RTX 5060 Ti — Wonderful_Sample6291 · 2026-09-07
- Markdown or JSON for handoffs between coding agents? Community debates template design — RocketSeven · 2026-09-07
- Ex-Stripe engineer's prompt trick: tell Codex agents to "make it surprisingly great" — blakesamic · 2026-09-07
- Yacine says his homegrown AI CAD tool now just works for his needs — yacineMTB · 2026-09-07
- Browser-tab real-time particle VFX, no engine: why AI building games as code changes everything — TheMoonMidas · 2026-09-07
- Dev swaps Codex Sol for Astra: better writing, planning and viz with mostly deleted config — steipete · 2026-09-07