Research map / Vision-language and multimodal / Visual question answering
Answerer in Questioner's Mind: Information Theoretic Approach to Goal-Oriented Visual Dialog
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 |
|---|---|---|---|
| use information-gain question selection instead of random question selection (ablation) | AQM on GuessWhat?! · GuessWhat?! test, 5 turns | accuracy: 46.36 ($\pm$ 0.91) -> 72.89 ($\pm$ 0.70) | Helped |
| use answerer-generated responses to train the questioner's likelihood model rather than training both models separately on the same data (ablation) | AQM on GuessWhat?! · GuessWhat?! test, 5 turns | accuracy: 66.73 ($\pm$ 0.76) -> 72.89 ($\pm$ 0.70) | Helped |
| share training data between the answerer and the questioner's likelihood model (ablation) | AQM on GuessWhat?! · GuessWhat?! test, 5 turns | accuracy: 61.64 ($\pm$ 0.97) -> 66.73 ($\pm$ 0.76) | Helped |
| use the countQ candidate-question sampler rather than random sampling (ablation) | AQM on GuessWhat?! · GuessWhat?! test, 5 turns | accuracy: 65.66 ($\pm$ 0.55) -> 66.73 ($\pm$ 0.76) | Helped |
| use seq2seq-generated candidate questions for the first turn instead of countQ questions (ablation) | AQM on GuessWhat?! · GuessWhat?! test, 2 turns | accuracy: 49.79\% -> 51.07\% | Helped |
| use seq2seq-generated candidate questions for the first turn instead of countQ questions (ablation) | AQM on GuessWhat?! · GuessWhat?! test, 5 turns | accuracy: 72.89\% -> 70.74\% | Hurt |
| use the true answerer distribution as the questioner's likelihood instead of an approximation (ablation) | AQM on GuessWhat?! · GuessWhat?! test, 3 turns | accuracy: 63.63\% -> 63.76\% | No clear effect |
| use the true answerer distribution as the questioner's likelihood instead of an approximation (ablation) | AQM on GuessWhat?! · GuessWhat?! test, 10 turns | accuracy: 78.72\% -> 81.96\% | Helped |
| increase property-recognition accuracy (ablation) | AQM on MNIST Counting Dialog · MNIST Counting Dialog, 6 turns | goal-oriented dialog accuracy: 39\% -> 51\% | Helped |
Closest papers by meaning
- Large-Scale Answerer in Questioner's Mind for Visual Dialog Question Generation (ICLR 2019 · similarity 0.92)
- Divide-and-Conquer: Tree-structured Strategy with Answer Distribution Estimator for Goal-Oriented Visual Dialogue (AAAI 2025 · similarity 0.90)
- Learning Better Visual Dialog Agents With Pretrained Visual-Linguistic Representation (CVPR 2021 · similarity 0.90)
- Two Can Play This Game: Visual Dialog With Discriminative Question Generation and Answering (CVPR 2018 · similarity 0.89)
- Visual Dialog (CVPR 2017 · similarity 0.88)
- Thinker: Learning to Think Fast and Slow (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.