Research map / Large language models / Language, safety and interpretability
Towards Coherent and Consistent Use of Entities in Narrative Generation
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 |
|---|---|---|---|
| Add an entity memory | Vanilla-LM · WikiPlots generated stories | long-range entity coherence (C): 3.29 -> 5.18 | Helped |
| Add an entity memory | Vanilla-LM · WritingPrompts generated stories | long-range entity coherence (C): 5.22 -> 5.45 | Helped |
| Add an entity memory | Vanilla-LM · WikiPlots human evaluation | coherence rating: 2.36 -> 2.59 | Helped |
| Use a dynamic rather than static entity memory | Mneme-LM · WritingPrompts generated stories | normalized attribute consistency (U): 30.81 -> 32.49 | Helped |
| Use a dynamic rather than static entity memory | Mneme-LM · WikiPlots generated stories | normalized attribute consistency (U): 32.97 -> 33.47 | Helped |
| Add a dynamic entity memory | Vanilla-LM · WikiPlots test | perplexity: 16.06 -> 15.97 | Helped |
| Add a dynamic entity memory | Vanilla-LM · WritingPrompts test | perplexity: 17.59 -> 17.44 | Helped |
| Remove memory initialization from entity prompts (ablation) | D-Mneme-LM · WikiPlots test | perplexity: 15.97 -> 16.61 | Hurt |
| Remove memory initialization from entity prompts (ablation) | D-Mneme-LM · WritingPrompts test | perplexity: 17.44 -> 18.22 | Hurt |
| Remove the entity-specific auxiliary loss (ablation) | D-Mneme-LM · WikiPlots test | perplexity: 15.97 -> 17.76 | Hurt |
| Remove the entity-specific auxiliary loss (ablation) | D-Mneme-LM · WritingPrompts test | perplexity: 17.44 -> 19.09 | Hurt |
| Add an entity memory | Vanilla-LM · WikiPlots human evaluation | consistency rating: 3.07 vs. 3.05; 3.07 vs. 3.00 | No clear effect |
Closest papers by meaning
- Agents' Room: Narrative Generation through Multi-step Collaboration (ICLR 2025 · similarity 0.86)
- Precise Information Control in Long-Form Text Generation (NeurIPS 2025 · similarity 0.86)
- Episodic Memories Generation and Evaluation Benchmark for Large Language Models (ICLR 2025 · similarity 0.85)
- One-Prompt-One-Story: Free-Lunch Consistent Text-to-Image Generation Using a Single Prompt (ICLR 2025 · similarity 0.84)
- Parameterized Synthetic Text Generation with SimpleStories (NeurIPS 2025 · similarity 0.84)
- Attributing Culture-Conditioned Generations to Pretraining Corpora (ICLR 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.