Research map / Reinforcement learning / RL from feedback
What Makes A Good Story? Designing Composite Rewards for Visual Storytelling
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 the BLEU reward with the composite relevance, coherence, and expressiveness rewards (ablation) | BLEU-RL on VIST · VIST test | CIDEr: 6.7 -> 8.6 | Helped |
| Replace the BLEU reward with the composite relevance, coherence, and expressiveness rewards (ablation) | BLEU-RL on VIST · VIST test | BLEU-4: 14.4 -> 12.4 | Hurt |
| Replace the BLEU reward with the composite relevance, coherence, and expressiveness rewards (ablation) | BLEU-RL on VIST · VIST test | METEOR: 35.2 -> 33.9 | Hurt |
| Replace the BLEU reward with the composite relevance, coherence, and expressiveness rewards (ablation) | BLEU-RL on VIST · VIST test | ROUGE-L: 30.1 -> 29.9 | Hurt |
| Replace the BLEU reward with the composite relevance, coherence, and expressiveness rewards (ablation) | BLEU-RL on VIST · VIST test | Relevance reward: 2.17 -> 10.39 | Helped |
| Replace the BLEU reward with the composite relevance, coherence, and expressiveness rewards (ablation) | BLEU-RL on VIST · VIST test | Coherence reward: 12.40 -> 12.74 | Helped |
| Replace the BLEU reward with the composite relevance, coherence, and expressiveness rewards (ablation) | BLEU-RL on VIST · VIST test | Expressiveness reward: 30.41 -> 39.37 | Helped |
| Use ReCo-RL rather than BLEU-RL | Visual storytelling on VIST · VIST human evaluation | Relevance preference: 17.6% -> 74.5% | Helped |
| Use ReCo-RL rather than BLEU-RL | Visual storytelling on VIST · VIST human evaluation | Coherence preference: 18.9% -> 72.3% | Helped |
| Use ReCo-RL rather than BLEU-RL | Visual storytelling on VIST · VIST human evaluation | Expressiveness preference: 19.5% -> 71.5% | Helped |
| Use ReCo-RL rather than AREL | Visual storytelling on VIST · VIST human evaluation | Relevance preference: 27.6% -> 62.2% | Helped |
| Use ReCo-RL rather than HSRL | Visual storytelling on VIST · VIST human evaluation | Coherence preference: 38.0% -> 51.9% | Helped |
Closest papers by meaning
- VinaBench: Benchmark for Faithful and Consistent Visual Narratives (CVPR 2025 · similarity 0.83)
- ViStoryBench: Comprehensive Benchmark Suite for Story Visualization (CVPR 2026 · similarity 0.83)
- Is Your World Simulator a Good Story Presenter? A Consecutive Events-Based Benchmark for Future Long Video Generation (CVPR 2025 · similarity 0.83)
- Intelligent Grimm - Open-ended Visual Storytelling via Latent Diffusion Models (CVPR 2024 · similarity 0.83)
- LogiStory: A Logic-Aware Framework for Multi-Image Story Visualization (ICLR 2026 · similarity 0.83)
- ContextualStory: Consistent Visual Storytelling with Spatially-Enhanced and Storyline Context (AAAI 2025 · similarity 0.82)
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.