DR Tulu | An open, end-to-end training recipe for long-form deep research @allenai
DR Tulu | An open, end-to-end training recipe for long-form deep research @allenai
Uploaded November 2025 | Updated September 2026, 1 day ago
Deep Research Tulu (DR Tulu), the first open model that is directly trained for long-form deep research tasks through an end-to-end training recipe that combines supervised fine-tuning (SFT) and Reinforcement Learning with Evolving Rubrics (RLER). DR Tulu starts from a strong base model and progresses through multiple training stages: SFT on high-quality, naturally occurring information-seeking queries, followed by online RL with RLER tailored to long-form research.
DR Tulu | An open, end-to-end training recipe for long-form deep researchStructure Modeling in Language Models❓ Molmo AMA: Your Questions Answered!Simulation and Generalization in VLA Models for Robotic Manipulation🥽 Molmo Vision Pro Demo - Augmenting how we see with AIReliable Evaluation and High-Quality Data: Building Blocks for Helpful Question Answering SystemsOlmo 3 | A family of leading fully open LMs and complete model flowDiscrete diffusion with planned denoisingMolmo 2 | Robotics ApplicationsThe Future is Hear: Advances in Computational Audition and Sound ManipulationSkill it! A Data-Driven Skills Framework for Understanding and Training Language ModelsAsta | an agentic ecosystem that advances scientific discovery
Ai2 |

DR Tulu | An open, end-to-end training recipe for long-form deep research

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER