Unverified70% confidenceTradeoffExact time
Writing scoring rules into a prompt in natural language has drawbacks: wording variations across different calls may cause the model's understanding to drift, and the scoring logic cannot be directly parsed or reused by programs.
2
Sources
70%
Confidence
Long-term
Relevance
9/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhen a prompt simultaneously requires mutually conflicting formats such as natural language expression and structured scoring, the quality of both types of output usually declines.80% similarVerifiedPrompt injection exploits the model's tendency to obey natural language instructions, and it is difficult to eradicate with simple input filtering.80% similarUnverifiedThe output of large language models inherently appears plausible (fluency bias), meaning that even when the reasoning chain is wrong, the generated natural language remains grammatically correct and natural in tone, rendering traditional monitoring approaches based on output format or exception capture ineffective.79% similarUnverifiedNatural language generation has a non-reproducibility problem: the same prompt may produce different behavior across multiple generations.77% similarUnverifiedLoulan believes that the idea that writing prompts completely will definitely produce good results is a popular misconception, and that the evaluation of natural language interaction cannot be summed up by simple numbers.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/927300API
curl https://kongchang.com/api/v1/knowledge/claims/927300MCP
get_claim(id=927300)