Unverified50% confidenceFactExact time
When the wording of a prompt changes slightly, the model's output distribution shifts accordingly, causing contradictions to emerge between successive conclusions.
1
Sources
50%
Confidence
Long-term
Relevance
9/28/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedEvery character in a prompt affects the probability distribution of all subsequent tokens, which is why subtle wording changes can lead to drastically different outputs.79% similarUnverified当生成模型反复在自身生成的数据上训练时，输出分布会逐渐退化，低概率事件被系统性遗忘78% similarVerifiedWhen an autoregressive model produces an erroneous intermediate conclusion early in the reasoning chain, it causes subsequent steps to accumulate errors, forming an error accumulation effect.77% similarUnverifiedThe vaguer the prompt, the more the model output collapses toward high-probability regions, and the stronger the consistency.76% similarVerifiedError cascading refers to small judgment deviations in early steps being progressively amplified as the reasoning chain extends, ultimately producing results far from what was intended.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/957723API
curl https://kongchang.com/api/v1/knowledge/claims/957723MCP
get_claim(id=957723)