Research map / Multi-agent systems and games / LLM agents
Narrative-Driven Paper-to-Slide Generation via ArcDeck
Tarik Can Ozden, SACHIDANAND VISHNUKUMAR SARMINI, Furkan Horoz, Ozgur Kara, Junho Kim, James Rehg
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 Discourse Parser (ablation) | ArcDeck outline generation · ArcBench outline ablation (Narrative Flow, 0-10 scale) | Narrative Flow: 7.50 -> 9.70 | Improved |
| add Global Commitment (ablation) | ArcDeck outline generation · ArcBench outline ablation (Narrative Flow, 0-10 scale) | Narrative Flow: 7.52 -> 9.70 | Improved |
| add Narrative Refinement Loop (ablation) | ArcDeck outline generation · ArcBench outline ablation (Narrative Flow, 0-10 scale) | Narrative Flow: 8.68 -> 9.70 | Improved |
| add Discourse Parser (pairwise preference) (ablation) | ArcDeck · ArcBench pairwise narrative flow | win rate vs ablated: 89.1 | Improved |
| add Global Commitment (pairwise preference) (ablation) | ArcDeck · ArcBench pairwise narrative flow | win rate vs ablated: 94.5 | Improved |
| add Narrative Refinement Loop (pairwise preference) (ablation) | ArcDeck · ArcBench pairwise narrative flow | win rate vs ablated: 61.8 | Improved |
| run Narrative Refinement Loop for 3 iterations instead of 1 (ablation) | ArcDeck outline · ArcBench NRL study (5 papers) | Narrative Flow: 4.18 -> 7.34 | Improved |
| run Narrative Refinement Loop for 2 iterations instead of 1 (ablation) | ArcDeck outline · ArcBench NRL study (5 papers) | Narrative Flow: 4.18 -> 7.00 | Improved |
| use ArcDeck instead of SlideGen (ablation) | GPT-4o backbone · ArcBench VLM Q/A quiz (open) Hard | accuracy (%): 66.15 -> 72.63 | Improved |
| use ArcDeck instead of SlideGen | Qwen3-VL-32B backbone · ArcBench VLM Q/A quiz (open) Hard | accuracy (%): 76.57 -> 83.22 | Improved |
| use ArcDeck instead of SlideGen (narrative flow A/B) | Qwen3-VL-32B backbone, GPT-5 judge · ArcBench A/B narrative flow | win rate (%): 48.5 | No clear effect |
| use ArcDeck instead of SlideGen (narrative flow A/B) | GPT-5 backbone, GPT-5 judge · ArcBench A/B narrative flow | win rate (%): 50.6 | No clear effect |
Closest papers by meaning
- AutoPresent: Designing Structured Visuals from Scratch (CVPR 2025 · similarity 0.86)
- Agents' Room: Narrative Generation through Multi-step Collaboration (ICLR 2025 · similarity 0.84)
- PPT-Eval: A Benchmark for Computer-Use Agents on PowerPoint Tasks (ICML 2026 · similarity 0.84)
- Paper2Poster: Towards Multimodal Poster Automation from Scientific Papers (NeurIPS 2025 · similarity 0.84)
- StoryBox: Collaborative Multi-Agent Simulation for Hybrid Bottom-Up Long-Form Story Generation Using Large Language Models (AAAI 2026 · similarity 0.84)
- From Words to Structured Visuals: A Benchmark and Framework for Text-to-Diagram Generation and Editing (CVPR 2025 · similarity 0.84)
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.
