Research map / Image generation / GANs
Are You Talking to Me? Reasoned Visual Dialog Generation Through Adversarial Learning
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 |
|---|---|---|---|
| Replace MLE-only training with adversarial training using global rewards (ablation) | CoAtt-G-MLE · VisDial v0.9 | R@5: 63.82 -> 64.17 | Helped |
| Add intermediate rewards instead of using only a global reward (ablation) | CoAtt-GAN-w/o R_inte · VisDial v0.9 | R@5: 64.17 -> 65.16 | Helped |
| Add teacher forcing after adversarial learning (ablation) | CoAtt-GAN-w/ R_inte · VisDial v0.9 | R@5: 65.16 -> 65.69 | Helped |
| Use sequential co-attention to jointly encode image, question, and dialogue history (ablation) | CoAtt-G-MLE · VisDial v0.9 | R@5: 62.33 -> 63.82 | Helped |
| Use the full generative model rather than the strongest reported generative baseline | VisDial v0.9 generative · VisDial v0.9 | R@5: 63.55 -> 65.69 | Helped |
| Use the proposed discriminative model rather than HCIAE-NP-ATT | VisDial v0.9 discriminative · VisDial v0.9 | R@5: 78.75 -> 80.71 | Helped |
| Use the final model instead of CoAtt-G-MLE | VisDial v0.9 human study · VisDial v0.9 | Turing Test pass rate: 0.46 -> 0.49 | Helped |
| Condition the discriminator on the question attention memory rather than the original question (ablation) | Visual dialogue discriminator · VisDial v0.9 | discriminator result | Hurt |
Closest papers by meaning
- Posterior-GAN: Towards Informative and Coherent Response Generation with Posterior Generative Adversarial Network (AAAI 2020 · similarity 0.91)
- Two Can Play This Game: Visual Dialog With Discriminative Question Generation and Answering (CVPR 2018 · similarity 0.91)
- Divide-and-Conquer: Tree-structured Strategy with Answer Distribution Estimator for Goal-Oriented Visual Dialogue (AAAI 2025 · similarity 0.86)
- NaturalBench: Evaluating Vision-Language Models on Natural Adversarial Samples (NeurIPS 2024 · similarity 0.85)
- Tackling Vision Language Tasks through Learning Inner Monologues (AAAI 2024 · similarity 0.85)
- MMDU: A Multi-Turn Multi-Image Dialog Understanding Benchmark and Instruction-Tuning Dataset for LVLMs (NeurIPS 2024 · similarity 0.85)
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.