Unverified50% confidenceFactExact time
Among the three tested models, only Qwen2.5-3B can reliably refuse to answer when the evidence does not support an answer.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedQwen3-1.7B's overall causal faithfulness is only 54.8%, far below that of Qwen3-4B.75% similarUnverifiedThis test is a single-source personal test; the model designation is labeled as Qwen 3.8 27B with naming ambiguity, and details such as the actual configuration, quantization method, and generation time have not been disclosed, leaving reproducibility to be verified.72% similarUnverifiedThe author noted that other models previously tested, such as 35B and 3B variants, actually performed worse than the Q2 version of Qwen3 27B on the Kanban task.70% similarVerifiedQwen3-4B is already on par with DeepSeek V3 or GPT-4o on multiple reasoning evaluation metrics.68% similarUnverifiedThe author has Qwen review the generated results with a maximum of 3 retries, returning the best-matching version if all three attempts fail to meet the standard.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/999245API
curl https://kongchang.com/api/v1/knowledge/claims/999245MCP
get_claim(id=999245)