Open-Source Training for Long-Horizon Research Trajectories
tom_doerr · x · 2026-07-15
The author generated 96,000 high-quality deep research trajectories and provided a fully open-source training recipe to train agentic language models on long-horizon web research tasks. Key highlights:
- The training goal is to enable models to handle long-horizon web research independently, without relying on external APIs
- The solution emphasizes no external API dependencies, making it easy to reproduce and run localized experiments
- This data and recipe are highly valuable references for building research agents, browser agents, and automated retrieval tasks
More from coding & agent
- A deleted NanoGPT PR still propagated into later world-record submissions — yacineMTB · 2026-07-21
- Soft Clamp cuts tool-call overuse in multi-teacher distillation, from 13.7% to 9.0% — antgroup · 2026-07-21
- Agent harness memory loss and compaction are still a major usability problem — adityaag · 2026-07-21
- SpecJudge runs locally on Ollama to pick the right-sized AI model for your project — jokiruiz · 2026-07-21
- A developer maps out six design rules for CLIs that humans and AI agents can both use — yujiezha · 2026-07-21
- A coding-agent skill that forces ADHD-friendly, answer-first output — ayghri · 2026-07-21