Research map / Video / Video-language understanding
StoryTeller: Training-Free Narrative Grounding for Long-Form Audio Description
Seung Hahm, Minh Dinh, SouYoung Jin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add structured fact extraction, identity, verification and memory over VLM-only narration (ablation) | Qwen3-VL narration · MAD-Eval | CIDEr: 11.4 -> 21.4 | Improved |
| use structured facts instead of scene summaries (ablation) | StoryTeller · MAD-Eval | CIDEr: 15.4 -> 21.4 | Improved |
| add cross-clip narrative memory (ablation) | StoryTeller · MAD-Eval | CIDEr: 18.0 -> 21.4 | Improved |
| add persistent identity graph (ablation) | StoryTeller · MAD-Eval | CIDEr: 17.9 -> 21.4 | Improved |
| add public IMDb metadata retrieval (ablation) | StoryTeller · MAD-Eval | SPICE: 4.8 -> 6.7 | Improved |
| add narrative memory on StoryAD-QA Track A 240s (ablation) | StoryTeller · StoryAD-QA Track A (240) | accuracy: 0.891 -> 0.972 | Improved |
| add components over VLM-only on short segments (Track A 30s) (ablation) | StoryTeller · StoryAD-QA Track A (30) | accuracy: 0.908 -> 0.892 | No clear effect |
| add identity graph on Track B 30+30 (ablation) | StoryTeller · StoryAD-QA Track B (30+30) | accuracy: 0.794 -> 0.754 | No clear effect |
| use StoryTeller instead of AutoAD-Zero on Track B 90+30 | AutoAD-Zero · StoryAD-QA Track B (90+30) | accuracy: 0.673 -> 0.788 | Improved |
| use StoryTeller instead of AutoAD-Zero on Track A 30s | AutoAD-Zero · StoryAD-QA Track A (30) | accuracy: 0.840 -> 0.892 | Improved |
| use StoryTeller (VideoLLaMA2) instead of AutoAD-Zero on MAD-Eval | AutoAD-Zero · MAD-Eval | CIDEr: 22.4 -> 19.1 (CIDEr); SPICE 7.3 -> 9.0; ROUGE-L 14.4 -> 16.0 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MM-Narrator — MM-Narrator: Narrating Long-form Videos with Multimodal In-Context Learning (CVPR 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- NarrativeTrack: Evaluating Entity-Centric Reasoning for Narrative Understanding (ECCV 2026 · similarity 0.88)
- DistinctAD: Distinctive Audio Description Generation in Contexts (CVPR 2025 · similarity 0.88)
- Contextual AD Narration with Interleaved Multimodal Sequence (CVPR 2025 · similarity 0.88)
- AutoAD III: The Prequel - Back to the Pixels (CVPR 2024 · similarity 0.87)
- MM-Narrator: Narrating Long-form Videos with Multimodal In-Context Learning (CVPR 2024 · similarity 0.87)
- StoryGPT-V: Large Language Models as Consistent Story Visualizers (CVPR 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.
