Research map / Label-efficient and robust learning / Recognition and long-tail
Best of Both Worlds: Transferring Knowledge from Discriminative Learning to a Generative Visual Dialog Model
Read the paperCheck 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 |
|---|---|---|---|
| transfer a pretrained discriminator's perceptual feedback to the generator (ablation) | HCIAE-G-MLE · VisDial v0.9 | R@5: 1.7% | Helped |
| use the HCIAE encoder instead of prior generative encoders (ablation) | Generative visual dialog models · VisDial v0.9 | R@5: 0.7% | Helped |
| use N-pair metric loss instead of standard discriminative cross-entropy (ablation) | HCIAE-D-MLE · VisDial v0.9 | R@5: 0.85% | Helped |
| add self-attentive answer encoding (ablation) | HCIAE-D-NP · VisDial v0.9 | R@5: 0.4% | Helped |
| continue discriminator training adversarially using only the ground-truth and generated answers (ablation) | HCIAE-G-DIS · VisDial v0.9 | R@5: 32.97% | Hurt |
| continue discriminator training with generated samples and incorrect negatives to preserve its learned structure (ablation) | HCIAE-G-DIS · VisDial v0.9 | R@5: 65.28 -> 65.05 | Helped |
Closest papers by meaning
- Visual Dialog (CVPR 2017 · similarity 0.92)
- Two Can Play This Game: Visual Dialog With Discriminative Question Generation and Answering (CVPR 2018 · similarity 0.91)
- The Dialog Must Go On: Improving Visual Dialog via Generative Self-Training (CVPR 2023 · similarity 0.91)
- Structure-Aware Multimodal Sequential Learning for Visual Dialog (AAAI 2024 · similarity 0.89)
- V^2Dial: Unification of Video and Visual Dialog via Multimodal Experts (CVPR 2025 · similarity 0.88)
- From Retrieval to Generation: A Simple and Unified Generative Model for End-to-End Task-Oriented Dialogue (AAAI 2024 · similarity 0.87)
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.