Low-rank pre-training methods converge to geometrically and spectrally distinct basins from full-rank training and from each other, even at similar validation perplexity.
Same pre-training loss, better downstream: Implicit bias matters for language models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Beyond Perplexity: A Geometric and Spectral Study of Low-Rank Pre-Training
Low-rank pre-training methods converge to geometrically and spectrally distinct basins from full-rank training and from each other, even at similar validation perplexity.