Research map / Vision-language and multimodal / Multimodal LLMs
MMRL++: Parameter-Efficient and Interaction-Aware Representation Learning for Vision-Language Models
Yuncheng Guo, Xiaodong Gu
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 SRRA (shared plus low-rank residual aligners) to MMRL (ablation) | MMRL · Base-to-novel (11 datasets avg) | HM: 81.20 -> 81.27; trainable params 4.992M -> 0.813M | Improved |
| add PRC (progressive composition) to MMRL (ablation) | MMRL · Base-to-novel (11 datasets avg) | HM: 81.20 -> 81.39 | Improved |
| add PRC on top of SRRA (ablation) | MMRL++ · Base-to-novel (11 datasets avg) | HM: 81.27 -> 81.77 | Improved |
| add representation tokens to the text branch (ablation) | MMRL++ · Base-to-novel (11 datasets avg) | HM: 77.30 -> 81.77 | Improved |
| add representation tokens to the vision branch (ablation) | MMRL++ · Base-to-novel (11 datasets avg) | HM: 78.28 -> 81.77 | Improved |
| use representation features for base-class inference (decoupling strategy) (ablation) | MMRL++ · Base-to-novel (11 datasets avg) | HM: 80.64 -> 81.77 | Improved |
| use only class token features for novel classes (decoupling) (ablation) | MMRL++ · Base-to-novel (11 datasets avg) | HM: 81.00 -> 81.77 | Improved |
| add shared representation space for both modalities (ablation) | MMRL++ · Base-to-novel (11 datasets avg) | Novel accuracy: 76.94 -> 78.32 | Improved |
| use balanced shared space instead of biased MaPLe-like mapping (ablation) | MMRL++ · Base-to-novel (11 datasets avg) | HM: 80.91 -> 81.77 | Improved |
| vary balance weight alpha from 0.0 / 1.0 to 0.7 (ablation) | MMRL++ · Base-to-novel (11 datasets avg) | HM: 79.01 -> 81.77; 80.29 -> 81.77 | Improved |
| replace MMRL by MMRL++ for stability | MMRL++ · EuroSAT novel (3 seeds) | std of novel accuracy: 5.05 -> 0.54 | Improved |
| increase low-rank dimension r2 beyond 64 (ablation) | MMRL++ · Base-to-novel (11 datasets avg) | HM: 81.77 -> 81.04 (r2=256) and 80.42 (r2=512) | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MMRL — MMRL: Multi-Modal Representation Learning for Vision-Language Models (CVPR 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MMRL: Multi-Modal Representation Learning for Vision-Language Models (CVPR 2025 · similarity 0.98)
- MMICL: Empowering Vision-language Model with Multi-Modal In-Context Learning (ICLR 2024 · similarity 0.90)
- MMA: Multi-Modal Adapter for Vision-Language Models (CVPR 2024 · similarity 0.90)
- VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks (ICLR 2025 · similarity 0.90)
- MMInference: Accelerating Pre-filling for Long-Context Visual Language Models via Modality-Aware Permutation Sparse Attention (ICML 2025 · similarity 0.90)
- VL-SAE: Interpreting and Enhancing Vision-Language Alignment with a Unified Concept Set (NeurIPS 2025 · similarity 0.89)
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.
