Qwen Business Unit ORG

Official codebase of the paper -- Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills

STAR: Similarity-guided Teacher-Assisted Refinement for Super-Tiny Function Calling Models

No description.

Implementation of "CollectionLoRA: Collecting 50 Effects in 1 LoRA via Multi-Teacher On-Policy Distillation"

Research code for CollectionLoRA, a technique that merges 50 or more AI image-effect adapters into one without losing quality. A single combined adapter handles many visual styles and can blend two effects at once without extra training.

CLIPO: Contrastive Learning in Policy Optimization Generalizes RLVR

Open Rubric System: Scaling Reinforcement Learning with Pairwise Adaptive Rubric

Search Self-Play: Pushing the Frontier of Agent Capability without Supervision

No description.

Skill-RM: Unifying Heterogeneous Evaluation Criteria via Agent Skill

No description.

Code of the research paper "SiameseNorm: Breaking the Barrier to Reconciling Pre/Post-Norm"

No description.

No description.

No description.

RiT: Rubrics-in-Thinking Reinforcement Learning for Improved Reasoning in Large Language Models