Primacy, anchoring, and order-dependence are architecturally necessary in autoregressive models due to causal masking constraints, with supporting evidence from theorems, LLM fits, and human experiments.
Proceedings of the National Academy of Sciences of the United States of America , title =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Bias by Necessity: Impossibility Theorems for Sequential Processing with Convergent AI and Human Validation
Primacy, anchoring, and order-dependence are architecturally necessary in autoregressive models due to causal masking constraints, with supporting evidence from theorems, LLM fits, and human experiments.