Pretrained LLMs exhibit human-like working memory interference via entangled representations, where successful recall requires suppressing task-irrelevant content.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Human-like Working Memory Interference in Large Language Models
Pretrained LLMs exhibit human-like working memory interference via entangled representations, where successful recall requires suppressing task-irrelevant content.