Unverified50% confidenceBenchmarkExact time
As the model scales from 8B to 70B, the predictive power of fixed output vector similarity for behavior actually worsens, with a paired mean correlation change of -0.080 (95% confidence interval [-0.127, -0.035]).
1
Sources
50%
Confidence
Long-term
Relevance
9/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedFor the 7B model, accuracy with the solver was 75.31% versus 72.02% for direct calculation, a paired improvement of +3.29 percentage points, with a 95% cluster interval of [-3.49, 10.38] crossing zero, making the advantage non-significant.75% similarUnverified领域微调后的7B模型在任务准确率上往往超越未经微调的70B通用模型，同时推理成本降低一个数量级74% similarVerifiedResearch shows that the output quality of the same model can vary by several times under different prompts.71% similarUnverifiedDifferent hyperparameters such as system prompt, temperature, maximum token count, and number of retries can cause the evaluation results of the same model to differ by more than 10 percentage points.71% similarUnverified该研究发现同一模型不同部署时间点在代码生成和逻辑推理任务上通过率可能出现10%至20%的显著波动，且方向不总是单向改善71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/931313API
curl https://kongchang.com/api/v1/knowledge/claims/931313MCP
get_claim(id=931313)