Research map / Reinforcement learning / Policy gradient and actor-critic
TaylorGAN: Neighbor-Augmented Policy Update Towards Sample-Efficient Natural Language Generation
Chun-Hsing Lin, Siang-Ruei Wu, Hung-yi Lee, Yun-Nung Chen
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 |
|---|---|---|---|
| Use Taylor neighbor augmentation instead of the Gumbel-Softmax estimator | TaylorGAN on EMNLP 2017 News validation · EMNLP 2017 News val | FED: 0.0218 -> 0.0149 | Improved |
| Use Taylor neighbor augmentation instead of the REINFORCE estimator | TaylorGAN on EMNLP 2017 News validation · EMNLP 2017 News val | FED: 0.0181 -> 0.0149 | Improved |
| Use Taylor neighbor augmentation instead of the Straight-Through estimator | TaylorGAN on EMNLP 2017 News validation · EMNLP 2017 News val | FED: 0.0449 -> 0.0149 | Improved |
| Increase bandwidth Λ from 0.50 to 1.00 (ablation) | TaylorGAN · EMNLP 2017 News val | Perplexity: 72 -> 60 | Improved |
| Increase bandwidth Λ from 0.50 to 1.00 (ablation) | TaylorGAN · EMNLP 2017 News val | FED: 0.0149 -> 0.0199 | Hurt |
| Increase discriminator constraint λSN from 0.07 to 0.10 (ablation) | TaylorGAN · EMNLP 2017 News val | Perplexity: 72 -> 64 | Improved |
| Increase discriminator constraint λSN from 0.07 to 0.10 (ablation) | TaylorGAN · EMNLP 2017 News val | FED: 0.0149 -> 0.0178 | Hurt |
| Reduce discriminator constraint λSN from 0.07 to 0.01 (ablation) | TaylorGAN · EMNLP 2017 News val | Perplexity: 72 -> 99 | Hurt |
| Remove entropy maximization (ablation) | TaylorGAN · EMNLP 2017 News val | Perplexity: 72 -> 310 | Hurt |
| Increase bandwidth Λ from 0.50 to 0.25 (ablation) | TaylorGAN · EMNLP 2017 News val | FED: 0.0149 -> 0.0179 | Hurt |
Closest papers by meaning
- Is Reinforcement Learning (Not) for Natural Language Processing: Benchmarks, Baselines, and Building Blocks for Natural Language Policy Optimization (ICLR 2023 · similarity 0.85)
- Language GANs Falling Short (ICLR 2020 · similarity 0.85)
- MaskGAN: Better Text Generation via Filling in the _______ (ICLR 2018 · similarity 0.85)
- Dialogue for Prompting: A Policy-Gradient-Based Discrete Prompt Generation for Few-Shot Learning (AAAI 2024 · similarity 0.84)
- Degeneration-free Policy Optimization: RL Fine-Tuning for Language Models without Degeneration (ICML 2024 · similarity 0.84)
- SimpleStrat: Diversifying Language Model Generation with Stratification (NeurIPS 2025 · similarity 0.83)
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.
