Latent Space Factorization in LoRA
This addresses a bottleneck in parameter-efficient finetuning for machine learning practitioners, offering incremental improvements in performance and robustness.
The paper tackles the problem of disambiguating task-relevant information in Low-rank adaptation (LoRA) for parameter-efficient finetuning, proposing Factorized Variational Autoencoder LoRA (FVAE-LoRA) which consistently outperforms standard LoRA in experiments on text, audio, and image tasks and improves robustness under distribution shifts.
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