Research map / Image generation / GANs
Adversarial Ranking for Language 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 |
|---|---|---|---|
| Replace the binary discriminator with relative ranking rewards | RankGAN · Synthetic data | NLL: 8.736 -> 8.247 | Helped |
| Replace the binary discriminator with relative ranking rewards | RankGAN · Chinese poems | BLEU-2: 0.738 -> 0.812 | Helped |
| Use relative ranking rewards rather than SeqGAN's discriminator | RankGAN · Chinese poems | human score: 3.58 -> 4.52 | Helped |
| Replace the binary discriminator with relative ranking rewards | RankGAN · COCO captions | BLEU-2: 0.815 -> 0.845 | Helped |
| Replace the binary discriminator with relative ranking rewards | RankGAN · COCO captions | BLEU-3: 0.636 -> 0.668 | Helped |
| Replace the binary discriminator with relative ranking rewards | RankGAN · COCO captions | BLEU-4: 0.587 -> 0.614 | Helped |
| Use relative ranking rewards rather than SeqGAN's discriminator | RankGAN · COCO captions | human score: 3.44 -> 4.61 | Helped |
| Replace the binary discriminator with relative ranking rewards | RankGAN · Romeo and Juliet | BLEU-2: 0.887 -> 0.914 | Helped |
| Replace the binary discriminator with relative ranking rewards | RankGAN · Romeo and Juliet | BLEU-3: 0.842 -> 0.878 | Helped |
| Replace the binary discriminator with relative ranking rewards | RankGAN · Romeo and Juliet | BLEU-4: 0.815 -> 0.856 | Helped |
| Use ranking information rather than token-level BLEU similarity as the policy-gradient reward | RankGAN · Synthetic data | NLL: 8.946 -> 8.247 | Helped |
Closest papers by meaning
- Training Language GANs from Scratch (NeurIPS 2019 · similarity 0.88)
- MaskGAN: Better Text Generation via Filling in the _______ (ICLR 2018 · similarity 0.87)
- Branch-GAN: Improving Text Generation with (not so) Large Language Models (ICLR 2024 · similarity 0.87)
- Adversarial Feature Matching for Text Generation (ICML 2017 · similarity 0.87)
- Generative Cooperative Networks for Natural Language Generation (ICML 2022 · similarity 0.87)
- Improving Neural Language Modeling via Adversarial Training (ICML 2019 · 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.