A sparse binary weight mask inherited from a 4B-token 'evolutionary' pruning loop improves 100M-token language model performance and human alignment compared with random-mask and dense controls.
A deep learning framework for neuroscience
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Model Connectomes: A Generational Approach to Data-Efficient Language Models
A sparse binary weight mask inherited from a 4B-token 'evolutionary' pruning loop improves 100M-token language model performance and human alignment compared with random-mask and dense controls.