Research map / Vision-language and multimodal / Human and embodied modeling
UTC: A Unified Transformer With Inter-Task Contrastive Learning for Visual Dialog
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use UTC rather than VDBERT | VisDial v1.0 val discriminative · VisDial v1.0 val | R@1: 54.02 -> 55.48 | Helped |
| Use UTC rather than LTMI | VisDial v1.0 val generative · VisDial v1.0 val | R@1: 40.44 -> 42.56 | Helped |
| Use UTC rather than LTMI | VisDial v1.0 val generative · VisDial v1.0 val | R@10: 69.71 -> 69.51 | Hurt |
| Apply UTC contrastive learning to the VisDial-BERT fine-tuned model | VisDial v1.0 test discriminative · VisDial v1.0 test | NDCG: 63.87 -> 64.60 | Helped |
| Add joint training of the generative task to the ranking task (ablation) | UTC discriminative · VisDial v1.0 val | R@1: 53.94 -> 54.39 | Helped |
| Add joint training of the discriminative task to the generation task (ablation) | UTC generative · VisDial v1.0 val | R@1: 41.39 -> 41.75 | Helped |
| Add answer and context contrastive losses (ablation) | UTC discriminative · VisDial v1.0 val | R@1: 54.39 -> 55.48 | Helped |
| Add answer and context contrastive losses (ablation) | UTC generative · VisDial v1.0 val | R@1: 41.75 -> 42.56 | Helped |
| Add answer contrastive loss (ablation) | UTC generative · VisDial v1.0 val | R@1: 41.75 -> 42.56 | Helped |
| Add context contrastive loss (ablation) | UTC discriminative · VisDial v1.0 val | R@1: 54.39 -> 55.24 | Helped |
| Add context contrastive loss to answer-contrastive training (ablation) | UTC generative · VisDial v1.0 val | R@1: 42.52 -> 42.56 | No clear effect |
| Add answer contrastive loss to context-contrastive training (ablation) | UTC discriminative · VisDial v1.0 val | R@1: 55.24 -> 55.48 | No clear effect |
Closest papers by meaning
- V^2Dial: Unification of Video and Visual Dialog via Multimodal Experts (CVPR 2025 · similarity 0.89)
- Structure-Aware Multimodal Sequential Learning for Visual Dialog (AAAI 2024 · similarity 0.87)
- Tackling Vision Language Tasks through Learning Inner Monologues (AAAI 2024 · similarity 0.87)
- From Retrieval to Generation: A Simple and Unified Generative Model for End-to-End Task-Oriented Dialogue (AAAI 2024 · similarity 0.86)
- F2RVLM: Boosting Fine-grained Fragment Retrieval for Multi-Modal Long-form Dialogue with Vision Language Model (AAAI 2026 · similarity 0.86)
- Knowledge-based Visual Question Answer with Multimodal Processing, Retrieval and Filtering (NeurIPS 2025 · 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.