Unverified50% confidenceFactExact time
In language model evaluation, position bias typically manifests as a preference for a fixed position (such as first or second), and this preference is unrelated to the text content.
1
Sources
50%
Confidence
Long-term
Relevance
11/5/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedWhen generating text, language models default to selecting word sequences with higher conditional probability.71% similarUnverifiedThe output of large language models is inherently variable, fundamentally different from the relatively stable deterministic rankings of traditional search engines under the same conditions.69% similarUnverifiedThe core difference between decision models and general-purpose large language models is that the former focuses more on outputting actionable judgments or recommendations rather than mere text generation.67% similarUnverifiedUnder two of the four prompts, for about two-fifths of the non-equivalent statement pairs, the language model proves both directions.66% similarUnverifiedLarge language models tend to assign greater attention weight to recent inputs and high-frequency words, a phenomenon known as recency bias.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1138240API
curl https://kongchang.com/api/v1/knowledge/claims/1138240MCP
get_claim(id=1138240)