Unverified50% confidenceBenchmarkExact time
Adding a "Do not guess" instruction to the prompt can reduce the proportion of fabricated false claims by a model from 71% to 20%.
1
Sources
50%
Confidence
Long-term
Relevance
9/28/2026
First Seen
Sources
Related Entities
Related Claims
Unverified少样本方式在格式化输出、分类任务和规则遵循方面的幻觉率可降低40%-70%73% similarUnverifiedTraditional inference solutions often achieve GPU memory utilization of only 20%–40%.68% similarUnverifiedFew-shot提示通过提供2-5个示例让模型理解模式，比纯文字描述准确率通常高出20%-40%68% similarUnverifiedThe 98% score from a single classification is the confidence level for that message's corresponding category, and does not represent an overall model accuracy of 98%.67% similarVerifiedThe Mixture of Experts (MoE) architecture activates only approximately 10%–30% of total parameters during inference, significantly reducing FLOPs and GPU memory usage per inference66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/955808API
curl https://kongchang.com/api/v1/knowledge/claims/955808MCP
get_claim(id=955808)