Erroneous processing heads in attention layers cause hop-generalization failures in LLMs; dynamically deactivating them at test time improves multi-step reasoning.
(Coin remains heads up.)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Scaling Reasoning Hop Exposes Weaknesses: Demystifying and Improving Hop Generalization in Large Language Models
Erroneous processing heads in attention layers cause hop-generalization failures in LLMs; dynamically deactivating them at test time improves multi-step reasoning.