Research map / Image generation / Text-to-image diffusion
Non-Monotonic Sequential Text 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 |
|---|---|---|---|
| Train the non-monotonic decoder with an annealed oracle instead of a uniform oracle | Non-monotonic generation · IWSLT'16 German-English tst-2014 | BLEU: 21.40 -> 23.30 | Helped |
| Add tree-encoding to the annealed model (ablation) | Non-monotonic machine translation · IWSLT'16 German-English tst-2014 | BLEU: 23.30 -> 24.30 | Helped |
| Add tree-encoding to the annealed model (ablation) | Non-monotonic machine translation · IWSLT'16 German-English tst-2014 | Ribes: 80.91 -> 81.64 | Helped |
| Tune the end-token prediction threshold on top of tree-encoding (ablation) | Non-monotonic machine translation · IWSLT'16 German-English tst-2014 | BLEU: 24.30 -> 24.60 | Helped |
| Tune the end-token prediction threshold on top of tree-encoding (ablation) | Non-monotonic machine translation · IWSLT'16 German-English tst-2014 | Ribes: 81.64 -> 80.53 | Hurt |
| Train word reordering with a non-monotonic uniform oracle rather than left-right order | Persona-Chat word reordering · Persona-Chat word reordering test | F1: 0.903 -> 0.960 | Helped |
| Train word reordering with an annealed oracle rather than left-right order | Persona-Chat word reordering · Persona-Chat word reordering test | BLEU: 46.3 -> 46.0 | No clear effect |
| Use annealed rather than uniform oracle training | Persona-Chat sentence completion · Persona-Chat sentence completion | BLEU: 38.9 -> 44.7 | Helped |
| Use an annealed non-monotonic policy rather than left-right generation | Non-monotonic language model · Persona-Chat unconditional generation | BLEU: 47.0 -> 56.2 | Helped |
| Use non-monotonic generation rather than left-right generation | Persona-Chat unconditional generation · Persona-Chat samples | %Novel: 17.8 -> 93.1 | Helped |
Closest papers by meaning
- Copy is All You Need (ICLR 2023 · similarity 0.87)
- Guaranteed Generation from Large Language Models (ICLR 2025 · similarity 0.86)
- Successor Features for Efficient Multi-Subject Controlled Text Generation (ICML 2024 · similarity 0.85)
- Syntactic and Semantic Control of Large Language Models via Sequential Monte Carlo (ICLR 2025 · similarity 0.84)
- Retrieval is Accurate Generation (ICLR 2024 · similarity 0.84)
- LongGenBench: Benchmarking Long-Form Generation in Long Context LLMs (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.