Unverified50% confidenceFactExact time
Sona stacks a 4-layer Transformer on top of Qwen's hidden states to further refine the representations.
1
Sources
50%
Confidence
Long-term
Relevance
10/6/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedQwen (Tongyi Qianwen) is developed by Alibaba Cloud and is based on the Transformer Decoder architecture.66% similarUnverifiedQwen3-4B 版本拥有约40亿参数，采用 Transformer 解码器架构，支持多语言理解与生成65% similarUnverifiedSince its first release in 2023, the Qwen series has gone through multiple iterations and adopts a Transformer decoder-only architecture.64% similarUnverifiedQwen底层架构基于Transformer技术，参数规模从70亿到数千亿不等63% similarUnverifiedLoRA is primarily applied to the four projection matrices Q, K, V, and O in the Transformer.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/980327API
curl https://kongchang.com/api/v1/knowledge/claims/980327MCP
get_claim(id=980327)