Unverified50% confidenceFactExact time
Microsoft conducted post-training on top of Alibaba's open-source Qwen3.5-9B weights, specifically to perform single-pass decision scoring tasks.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMicrosoft's team is using reinforcement learning methods to train models in order to improve performance degradation behavior in multi-turn conversations.65% similarUnverifiedThe Qwen2.5-Max-0902 update conducted targeted post-training optimization focused on two areas: programming and collaborative office work.63% similarUnverifiedE5 由微软研究院推出，采用对比学习训练范式61% similarUnverified智谱GLM-5.3基于743B（7430亿参数）基座进行后训练，主打编程与智能体两大核心场景61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/997994API
curl https://kongchang.com/api/v1/knowledge/claims/997994MCP
get_claim(id=997994)