Expired50% confidenceFactTime unknown
Research shows that the same base LLM model can vary by 30-50 percentage points in performance under different system prompts.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 9/30/2026(expired)
Sources
Deep Dive into Hermes Agent Self-Evolution: The GIPA Algorithm and Engineering Practice
bilibili探索未至之境
Related Claims
Unverified同一个基座模型在不同系统提示词下的表现差异可以高达30-50个百分点79% similarUnverifiedResearch shows that, for the same task, an optimized prompt can improve model output quality by 30%–50% compared to arbitrary input.71% similarUnverified研究表明，即使在 GSM8K 或 MATH 数据集上接近满分的模型，面对稍作变形或需要跨领域组合的问题时性能下降可达30-50个百分点71% similarUnverifiedThe average evaluation metrics of a new LLM may be better than those of an older model, but regressions can occur in niche critical scenarios.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/55461API
curl https://kongchang.com/api/v1/knowledge/claims/55461MCP
get_claim(id=55461)