Research map / Reinforcement learning / RL from feedback
Neural Machine Translation with Adequacy-Oriented Learning
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 |
|---|---|---|---|
| Use the adequacy-oriented discriminator and CDR orientator | RNNSearch · Zh⇒En NIST test sets (All) | BLEU: 36.76 -> 38.42 | Helped |
| Add discriminator and CDR orientator to the coverage model | RNNSearch-Coverage · Zh⇒En NIST test sets (All) | BLEU: 39.13 -> 39.81 | Helped |
| Use the RNN discriminator and CDR orientator | RNNSearch · IWSLT 2014 De⇒En | BLEU: 26.51 -> 27.79 | Helped |
| Use the RNN discriminator and CDR orientator | Transformer-Base · WMT 2014 En⇒De | BLEU: 27.30 -> 28.01 | Helped |
| Use the RNN discriminator and CDR orientator | Transformer-Big · WMT 2014 En⇒De | BLEU: 28.35 -> 28.99 | Helped |
| Replace the CNN discriminator with the RNN discriminator (ablation) | RNNSearch adversarial setup · Zh⇒En NIST test sets (All) | BLEU: 37.54 vs. 37.59 | No clear effect |
| Use CDR reward instead of BLEU or chrF3 reward (ablation) | RNNSearch orientator · Zh⇒En NIST test sets (All) | BLEU: 37.44; 37.75; 38.02 | Helped |
| Add the discriminator to the CDR orientator (ablation) | RNNSearch · Zh⇒En NIST test sets (All) | BLEU: 38.02 -> 38.42 | Helped |
| Add the CDR orientator to the discriminator (ablation) | RNNSearch · Zh⇒En NIST test sets (All) | BLEU: 37.59 -> 38.42 | Helped |
| Add the CDR orientator without the discriminator (ablation) | Transformer-Base · WMT 2014 En⇒De | BLEU: 27.30 -> 27.80 | Helped |
| Add the RNN discriminator to the CDR orientator (ablation) | Transformer-Base · WMT 2014 En⇒De | BLEU: 27.80 -> 28.01 | Helped |
| Add the CDR orientator without the discriminator (ablation) | Transformer-Big · WMT 2014 En⇒De | BLEU: 28.35 -> 28.63 | Helped |
Closest papers by meaning
- On the Weaknesses of Reinforcement Learning for Neural Machine Translation (ICLR 2020 · similarity 0.90)
- Dual Learning for Machine Translation (NeurIPS 2016 · similarity 0.88)
- Reducing the Probability of Undesirable Outputs in Language Models Using Probabilistic Inference (NeurIPS 2025 · similarity 0.86)
- Calibrating Translation Decoding with Quality Estimation on LLMs (NeurIPS 2025 · similarity 0.85)
- A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models (ICLR 2024 · similarity 0.85)
- SimulPL: Aligning Human Preferences in Simultaneous Machine Translation (ICLR 2025 · similarity 0.85)
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.