Research map / Vision-language and multimodal / Cross-modal learning
MultiLoReFT: Decoupling Shared and Modality-Specific Subspaces in Multimodal Learning via Low-Rank Representation Fine-Tuning
Sana Tonekaboni, Viktoria Schuster, Caroline Uhler
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adaptive rank pruning (ablation) | MultiLoReFT · Simulation I (M1 label from z_s) | accuracy (lower = less leakage): 57.5 -> 37.8 | Improved |
| add GradNorm loss weighting (ablation) | MultiLoReFT · Simulation I (M1 label from z_s) | accuracy (lower = less leakage): 82.2 -> 37.8 | Improved |
| add MI retention loss (ablation) | MultiLoReFT · Simulation I (M1 label from z_s) | accuracy (lower = less leakage): 68.7 -> 37.8 | Improved |
| add independence loss (ablation) | MultiLoReFT · Simulation I (M1 label from z_s) | accuracy (lower = less leakage): 34.3 -> 37.8 | Mixed |
| add orthogonality loss (ablation) | MultiLoReFT · Simulation I (M1 label from z_s) | accuracy (lower = less leakage): 72.6 -> 37.8 | Improved |
| use stronger unimodal encoders (MViT-V2-S, WavLM-Base+) (ablation) | MultiLoReFT · Crema-D emotion | accuracy: 46.0 -> 76.0 | Improved |
| fine-tune with MultiLoReFT versus pretrained embeddings (3 modalities) (ablation) | MultiLoReFT · CMU-MOSI | accuracy: 54.81 -> 56.84 | Improved |
| use MultiLoReFT versus APOLLO/DRIM-U for decoupling shared factor | MultiLoReFT · Simulation I | gap Delta: 0.1 -> 18.0 | Improved |
| MultiLoReFT versus late fusion on UR-Funny | MultiLoReFT · UR-Funny | accuracy: 58.8 -> 61.1 | Improved |
| MultiLoReFT versus cross attention on UR-Funny raw video | MultiLoReFT · UR-Funny (raw video) | accuracy: 58.2 -> 58.3 | No clear effect |
Closest papers by meaning
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.90)
- MSP: Modality Self-Play for Training Multimodal LLMs Without Paired Cross-Modal Data (NeurIPS 2026 · similarity 0.90)
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 2025 · similarity 0.89)
- Balancing Multimodal Training Through Game-Theoretic Regularization (NeurIPS 2025 · similarity 0.89)
- Mitigating Pervasive Modality Absence Through Multimodal Generalization and Refinement (AAAI 2025 · similarity 0.88)
- Unity by Diversity: Improved Representation Learning for Multimodal VAEs (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
