Research map / Vision-language and multimodal / Cross-modal learning
SupCLAP: Controlling Optimization Trajectory Drift in Audio-Text Contrastive Learning with Support Vector Regularization
Jiehui Luo, Yuguo Yin, Yuxin Xie, Jinghan Ru, Xianwei Zhuang, Minghua He, Aofan Liu, Zihan Xiong, Dongchao Yang
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 bidirectional DynamicSVR regularizer | InfoNCE · AudioCaps (monolingual) | T2A R@1: 41.87 -> 44.16 | Improved |
| add bidirectional DynamicSVR regularizer | SigLIP · AudioCaps (monolingual) | T2A R@1: 36.74 -> 43.09 | Improved |
| add bi-DynamicSVR | InfoNCE · ESC-50 zero-shot | top-1 accuracy: 89.6 -> 92.1 | Improved |
| use dynamic radius predictor instead of static learnable radius (ablation) | bi-SVR on InfoNCE · AudioCaps (monolingual) | A2T R@1: 57.77 -> 59.66 | Improved |
| make SVR bidirectional instead of text-to-audio only (ablation) | DynamicSVR · AudioCaps (monolingual) | A2T R@1: 58.51 -> 59.66 | Improved |
| add constraint term L_cons to radius predictor (ablation) | bi-DynamicSVR · AudioCaps (monolingual) | T2A R@1: 44.01 -> 44.16 | Improved |
| add constraint term L_cons to radius predictor (ablation) | uni-DynamicSVR · AudioCaps (monolingual) | A2T R@1: 57.67 -> 58.51 | Improved |
| replace SVR with learnable scaling of positive probability (Simple Reweight) (ablation) | InfoNCE · AudioCaps (monolingual) | A2T R@1: 56.72 -> 51.26 | Hurt |
| add bi-DynamicSVR trained on AudioCaps, tested on Clotho | InfoNCE · AudioCaps -> Clotho | A2T R@1: 18.1643 -> 20.1932 | Improved |
| add bi-DynamicSVR on large-scale WavCaps | InfoNCE · WavCaps | T2A R@1: 7.76 -> 8.30 | Improved |
| use InfoNCE rather than SigLIP loss as base (ablation) | contrastive loss · AudioCaps (monolingual) | T2A R@1: 36.74 -> 41.87 | Improved |
| add bi-DynamicSVR to ATRI-CACL on multilingual | ATRI-CACL · AudioCaps (multilingual) | T2A R@1: 39.44 -> 43.61 | Improved |
Closest papers by meaning
- Revisiting Deep Audio-Text Retrieval Through the Lens of Transportation (ICLR 2024 · similarity 0.86)
- Measuring Audio's Impact on Correctness: Audio-Contribution-Aware Post-Training of Large Audio Language Models (ICLR 2026 · similarity 0.86)
- Omni-Interactive Universal Embedder (NeurIPS 2026 · similarity 0.85)
- MATS: An Audio Language Model under Text-only Supervision (ICML 2025 · similarity 0.85)
- Scaling Language-centric Omnimodal Representation Learning (NeurIPS 2025 · similarity 0.85)
- Unbiased Sliced Wasserstein Kernels for High-Quality Audio Captioning (NeurIPS 2025 · similarity 0.84)
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.
