Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Zero-Resource Neural Machine Translation with Multi-Agent Communication Game
Yun Chen, Yang Liu, Victor Li
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 |
|---|---|---|---|
| Jointly train the captioner and translator in the communication game rather than use the fixed-captioner TS teacher-student method | Zero-resource NMT · IAPR-TC12 De-En | BLEU: 14.1 -> 18.6 | Improved |
| Jointly train the captioner and translator in the communication game rather than use the fixed-captioner TS teacher-student method | Zero-resource NMT · IAPR-TC12 En-De | BLEU: 9.2 -> 14.2 | Improved |
| Jointly train the captioner and translator in the communication game rather than use the fixed-captioner TS teacher-student method | Zero-resource NMT · Multi30K De-En | BLEU: 16.4 -> 19.6 | Improved |
| Jointly train the captioner and translator in the communication game rather than use the fixed-captioner TS teacher-student method | Zero-resource NMT · Multi30K En-De | BLEU: 10.3 -> 16.6 | Improved |
| Jointly optimize the captioner along with the translator instead of keeping the captioner fixed (ablation) | Zero-resource NMT · IAPR-TC12 De-En | BLEU: 17.6 -> 18.6 | Improved |
| Jointly optimize the captioner along with the translator instead of keeping the captioner fixed (ablation) | Zero-resource NMT · IAPR-TC12 En-De | BLEU: 11.9 -> 14.2 | Improved |
| Jointly optimize the captioner along with the translator instead of keeping the captioner fixed (ablation) | Zero-resource NMT · Multi30K De-En | BLEU: 16.0 -> 19.6 | Improved |
| Jointly optimize the captioner along with the translator instead of keeping the captioner fixed (ablation) | Zero-resource NMT · Multi30K En-De | BLEU: 12.1 -> 16.6 | Improved |
| Jointly train rather than keep the captioner fixed (ablation) | Image captioner · IAPR-TC12 image-to-German | BLEU: 10.7 -> 10.5 | Hurt |
| Jointly train rather than keep the captioner fixed (ablation) | Image captioner · IAPR-TC12 image-to-English | BLEU: 18.0 -> 17.3 | Hurt |
| Jointly train rather than keep the captioner fixed (ablation) | Image captioner · Multi30K image-to-German | BLEU: 10.9 -> 12.3 | Improved |
| Jointly train rather than keep the captioner fixed (ablation) | Image captioner · Multi30K image-to-English | BLEU: 22.7 -> 21.3 | Hurt |
Closest papers by meaning
- Emergent Translation in Multi-Agent Communication (ICLR 2018 · similarity 0.87)
- Cross-Lingual Pre-Training Based Transfer for Zero-Shot Neural Machine Translation (AAAI 2020 · similarity 0.86)
- Dual Learning for Machine Translation (NeurIPS 2016 · similarity 0.85)
- Unsupervised Neural Machine Translation (ICLR 2018 · similarity 0.85)
- NExT-GPT: Any-to-Any Multimodal LLM (ICML 2024 · similarity 0.84)
- Scalable Multilingual Multimodal Machine Translation with Speech-Text Fusion (ICLR 2026 · 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.
