Research map / Vision-language and multimodal / Cross-modal learning
DiscoVL: Unveiling Disentangled Cross-Modal Representation Learning via Orthogonal Adversarial Regularization for Vision-Language Models
Mengping Dong, Jinbao Li, Fei Li
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 multi-branch residual aligner (ablation) | MMRL baseline · Base-to-novel average (11 datasets) | HM: 81.20 -> 81.24 | Improved |
| add bidirectional cross-modal feedback (ablation) | DiscoVL · Base-to-novel average (11 datasets) | HM: 81.24 -> 81.32 | Improved |
| add adversarial representation triplet loss (ablation) | DiscoVL · Base-to-novel average (11 datasets) | HM: 81.32 -> 81.46 | Improved |
| add orthogonal regularization (ablation) | DiscoVL · Base-to-novel average (11 datasets) | HM: 81.46 -> 81.58 | Improved |
| add L_atr^k (ablation) | MMRL baseline · Base-to-novel: StanfordCars | HM: 78.06 -> 78.26 | Improved |
| add L_atr^k' (ablation) | DiscoVL · Base-to-novel: StanfordCars | HM: 78.26 -> 78.33 | Improved |
| add orthogonal regularization L_ortho (ablation) | DiscoVL · Base-to-novel: StanfordCars | HM: 78.33 -> 78.42 | Improved |
| raise gamma from 0.2 to 0.3 (ablation) | DiscoVL · Base-to-novel: Food101 | HM: 90.98 -> 91.20 | Improved |
| raise alpha from 0.0 to 0.7 (ablation) | DiscoVL · Base-to-novel: UCF101 | HM: 84.31 -> 84.46 | Improved |
| raise beta from 0.0 to 0.5 (ablation) | DiscoVL · Base-to-novel: FGVCAircraft | HM: 41.29 -> 41.60 | Improved |
| use cosine instead of L1 regularization to frozen CLIP (ablation) | DiscoVL · Base-to-novel: EuroSAT | HM: 88.00 -> 88.32 | Improved |
| increase branches E from 2 to 4 (ablation) | DiscoVL · Base-to-novel: EuroSAT | HM: 88.17 -> 88.32 | Improved |
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 3 of 3 shared results
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
- Measuring Cross-Modal Synergy: A Benchmark for VLM Explainability (NeurIPS 2026 · similarity 0.87)
- VL-SAE: Interpreting and Enhancing Vision-Language Alignment with a Unified Concept Set (NeurIPS 2025 · similarity 0.86)
- Same Concept, Different Directions: Cross-Modal Feature Heterogeneity in Sparse Autoencoders (NeurIPS 2026 · similarity 0.86)
- Dynamic Adapter with Semantics Disentangling for Cross-lingual Cross-modal Retrieval (AAAI 2025 · similarity 0.86)
- Vision-Language Models Create Cross-Modal Task Representations (ICML 2025 · similarity 0.86)
- CrossMAE: Cross-Modality Masked Autoencoders for Region-Aware Audio-Visual Pre-Training (CVPR 2024 · similarity 0.86)
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.
