The authors hypothesize, with evidence from merging and freezing experiments, that the attention output projection is the primary locus of mathematical reasoning in large language models.
Deep reinforcement learning from human preferences
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.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Who Reasons in the Large Language Models?
The authors hypothesize, with evidence from merging and freezing experiments, that the attention output projection is the primary locus of mathematical reasoning in large language models.