Verbose ListOps embeds deterministic nested arithmetic in narrative camouflage and shows frontier LLMs collapse from near-perfect raw ListOps accuracy to 25 to 55 percent at just 10k tokens.
the lights flicker
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Context Is Not Comprehension
Verbose ListOps embeds deterministic nested arithmetic in narrative camouflage and shows frontier LLMs collapse from near-perfect raw ListOps accuracy to 25 to 55 percent at just 10k tokens.