Gradually increasing the probability that LoRA adapters stay active during fine-tuning improves generalization, merging, and pruning robustness.
For image classification, we follow Zanella & Ben Ayed (2024) and use 11 datasets: • Aircraft (Maji et al.,
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
-
Come Together, But Not Right Now: A Progressive Strategy to Boost Low-Rank Adaptation
Gradually increasing the probability that LoRA adapters stay active during fine-tuning improves generalization, merging, and pruning robustness.