ByteDanceSeed-X
Seed-X PPO 7B
Seed-X PPO 7B is a translation model for 28 languages, trained with reinforcement learning.
Model design
The Mistral-architecture checkpoint needs a target-language tag in its prompt. The card advises against multi-turn chat formatting. Official GPTQ INT8 and AWQ INT4 versions are also published.Seed-X PPO 7B official model card
Practical use
Practical guidance. Keep terminology and source text explicit. Evaluate each translation direction separately, including low-frequency names and domain terms. Do not use a translation-only checkpoint as a general assistant.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | Seed-X-PPO-7BSeed-X-PPO-7B-GPTQ-Int8Seed-X-PPO-7B-AWQ-Int4 |
| License | OpenMDWSeed-X PPO 7B official model card |
| Access | Weights are published in the official ByteDance Seed repository.Seed-X PPO 7B official model card |
Related models
- Seed-X Instruct 7B
ByteDance · Seed-X
Seed-X Instruct 7B is the instruction-tuned translation checkpoint in the Seed-X family.
Launch date not confirmedDownloadable - Seed-X RM 7B
ByteDance · Seed-X
Seed-X RM 7B scores translation quality using a model trained on human preferences.
Launch date not confirmedDownloadable