Latent Space Factorization in LoRA
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908606243274752 |
|---|---|
| author | Kumar, Shashi Kaloga, Yacouba Mitros, John Motlicek, Petr Kodrasi, Ina |
| author_facet | Kumar, Shashi Kaloga, Yacouba Mitros, John Motlicek, Petr Kodrasi, Ina |
| contents | Low-rank adaptation (LoRA) is a widely used method for parameter-efficient finetuning. However, existing LoRA variants lack mechanisms to explicitly disambiguate task-relevant information within the learned low-rank subspace, potentially limiting downstream performance. We propose Factorized Variational Autoencoder LoRA (FVAE-LoRA), which leverages a VAE to learn two distinct latent spaces. Our novel Evidence Lower Bound formulation explicitly promotes factorization between the latent spaces, dedicating one latent space to task-salient features and the other to residual information. Extensive experiments on text, audio, and image tasks demonstrate that FVAE-LoRA consistently outperforms standard LoRA. Moreover, spurious correlation evaluations confirm that FVAE-LoRA better isolates task-relevant signals, leading to improved robustness under distribution shifts. Our code is publicly available at: https://github.com/idiap/FVAE-LoRA |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_19640 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Latent Space Factorization in LoRA Kumar, Shashi Kaloga, Yacouba Mitros, John Motlicek, Petr Kodrasi, Ina Machine Learning Low-rank adaptation (LoRA) is a widely used method for parameter-efficient finetuning. However, existing LoRA variants lack mechanisms to explicitly disambiguate task-relevant information within the learned low-rank subspace, potentially limiting downstream performance. We propose Factorized Variational Autoencoder LoRA (FVAE-LoRA), which leverages a VAE to learn two distinct latent spaces. Our novel Evidence Lower Bound formulation explicitly promotes factorization between the latent spaces, dedicating one latent space to task-salient features and the other to residual information. Extensive experiments on text, audio, and image tasks demonstrate that FVAE-LoRA consistently outperforms standard LoRA. Moreover, spurious correlation evaluations confirm that FVAE-LoRA better isolates task-relevant signals, leading to improved robustness under distribution shifts. Our code is publicly available at: https://github.com/idiap/FVAE-LoRA |
| title | Latent Space Factorization in LoRA |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2510.19640 |