Unverified50% confidenceTradeoffExact time
The paper shows that model-level interventions—such as adding instructions to acknowledge uncertainty in the system prompt—can improve epistemic humility but often reduce task accuracy.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe effort parameter controls how much the model thinks rather than how much it says; reducing effort does not reliably shorten visible responses, and controlling response length requires explicit prompting.75% similarUnverifiedFor high-reliability scenarios such as evidence-based medicine, a small model that honestly expresses uncertainty may be more practical than an overconfident large model—this represents a tradeoff.75% similarUnverifiedThe model's tool-triggering decision is essentially a probabilistic threshold problem: when confidence exceeds the threshold the model tends to answer directly, and only triggers retrieval when confidence falls below it.75% similarUnverified该项目引入主动消歧机制：当不确定性超过阈值时智能体不直接输出，而是主动向用户提出澄清问题74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/999286API
curl https://kongchang.com/api/v1/knowledge/claims/999286MCP
get_claim(id=999286)