Unverified50% confidenceFactExact time
The experiments use the recurrent model Ouro-1.4B and the standard Transformer Qwen2.5-1.5B, constructing minimal-pair questions that differ by only one number.
1
Sources
50%
Confidence
Long-term
Relevance
10/21/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe experiments used two models: RWKV-4-Raven-7B (recurrent architecture) and Tulu-Pythia-6.9b (Transformer architecture).65% similarUnverifiedTev1-4B-experimental is built on top of the Qwen3.5 4B base model.63% similarUnverifiedQwen-Image-2.1-Turbo maintains the same 7B parameter architecture as the base model, with the number of denoising steps reduced by 5×62% similarUnverifiedThe tests used Transformers.js 4.2.0 to run three models on WebGPU with 4-bit weights: SmolLM2 360M Instruct, Qwen2.5 0.5B Instruct, and Llama 3.2 1B Instruct (q4f16).60% similarUnverifiedQwen底层架构基于Transformer技术，参数规模从70亿到数千亿不等59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1078902API
curl https://kongchang.com/api/v1/knowledge/claims/1078902MCP
get_claim(id=1078902)