Unverified50% confidenceDecision RuleExact time
A practical criterion is: if a field is used in more than 80% of reasoning, prefetch it; if its usage rate is low and its size is large, leave it to MCP.
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedResearch shows that in certain tasks, simply changing '请判断' to '请分析' in a prompt can cause accuracy fluctuations of 5–10 percentage points.69% similarUnverifiedThe composite score in this study places 80% weight on the large language model's judgment of factual entailment and 20% on lexical overlap.69% similarUnverifiedDemo three introduces a decision rule: the top option must reach at least 80% and exceed the second-best option by 20 percentage points; otherwise, the output is "request clarification."68% similarVerifiedThe Mixture of Experts (MoE) architecture activates only approximately 10%–30% of total parameters during inference, significantly reducing FLOPs and GPU memory usage per inference68% similarUnverifiedIn 10%–25% of cases, reasoning about the hypothetical grader pulled the agent's work away from the user's original specification.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1086255API
curl https://kongchang.com/api/v1/knowledge/claims/1086255MCP
get_claim(id=1086255)