Spaces:
Running
Please release Training pipeline and related Docs for the Qwen open models
asking Alibaba to consider releasing:
Training code — the full pre-training and post-training pipelines used to create Qwen2.5, Qwen3, and Qwen3.5
Training datasets — or at minimum, detailed documentation of the data curation process, sources, and filtering methodology
RLHF / alignment pipeline — the reinforcement learning and instruction-tuning processes that turn base models into usable assistants
Evaluation frameworks — the internal benchmarks and testing suites used during development
Architecture research — the reasoning behind key design decisions across model generations
It looks like Qwen's commitment to Open source AI maybe changing, if the open model project is ended can we have the details above to adapt / continue model dev as a Open community