Pretrained LLM layers can be skipped/repeated per input to build custom paths, but the search uses ground-truth answers, so the accuracy gains are fitted, not predicted.
Faith and fate: Limits of transformers on compositionality
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.LG 1years
2025 1verdicts
REJECT 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Skip a Layer or Loop it? Test-Time Depth Adaptation of Pretrained LLMs
Pretrained LLM layers can be skipped/repeated per input to build custom paths, but the search uses ground-truth answers, so the accuracy gains are fitted, not predicted.