Research map / Video / Video-language understanding
TRANSPORTER: Transferring Visual Semantics from VLM Manifolds
Alexandros Stergiou
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 |
|---|---|---|---|
| couple generator and VLM embeddings with learned network instead of decode-and-re-encode (ablation) | TRANSPORTER · Caption-target alignment (VideoLLaMA 3) | cosine similarity: 0.19 -> 0.26 | Improved |
| combine Phi1 and Phi2 via learnable OT instead of using Phi1 only (ablation) | TRANSPORTER · Caption-target alignment (VideoLLaMA 3) | cosine similarity: 0.16 -> 0.26 | Improved |
| use learnable entropic OT instead of Sliced Gromov-Wasserstein (ablation) | TRANSPORTER · Caption-target alignment (VideoLLaMA 3) | cosine similarity: 0.14 -> 0.26 | Improved |
| increase OT projections from 50 to 400 (ablation) | TRANSPORTER · Caption-target alignment (VideoLLaMA 3) | cosine similarity: 0.24 -> 0.26 | No clear effect |
| use bounded Hellinger distance instead of KL divergence for logit divergence (ablation) | TRANSPORTER concept bank · Caption-target alignment (VideoLLaMA 3) | cosine similarity: 0.19 -> 0.26 | Improved |
| use Hellinger distance instead of JS divergence (ablation) | TRANSPORTER concept bank · Caption-target alignment (VideoLLaMA 3) | cosine similarity: 0.22 -> 0.26 | Improved |
| generate videos with TRANSPORTER instead of activation-maximization baselines | VideoLLaMA 3 logits · CLIPv score | CLIPv score: 16.74 -> 35.44 | Improved |
| generate videos with TRANSPORTER instead of feature-visualization baseline | Gemma 3 logits · FVD | FVD: 2.18e3 -> 1.05e2 | Improved |
Closest papers by meaning
- VideoHallu: Evaluating and Mitigating Multi-modal Hallucinations on Synthetic Video Understanding (NeurIPS 2025 · similarity 0.89)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.89)
- PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding (NeurIPS 2025 · similarity 0.89)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.88)
- Language Model as Visual Explainer (NeurIPS 2024 · similarity 0.88)
- ViLMA: A Zero-Shot Benchmark for Linguistic and Temporal Grounding in Video-Language Models (ICLR 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.
