aisingapore/Llama-SEA-LION-v3-70B-IT
Llama-SEA-LION-v3-70B-IT is a 70 billion parameter instruction-tuned decoder-only language model developed by AI Singapore, built upon the Llama 3.1 architecture. It is specifically designed and optimized for Southeast Asian languages, supporting Burmese, Chinese, English, Filipino, Indonesian, Javanese, Khmer, Lao, Malay, Sundanese, Tamil, Thai, and Vietnamese. With a 128k context length, this model excels in instruction-following and general language tasks across this diverse linguistic region.
Loading preview...
Llama-SEA-LION-v3-70B-IT: Southeast Asian Language Model
Llama-SEA-LION-v3-70B-IT is a 70 billion parameter instruction-tuned large language model developed by AI Singapore, funded by the Singapore NRF. It is based on the Llama 3.1 architecture and is specifically designed to serve the Southeast Asian (SEA) region, supporting a wide array of languages including Burmese, Chinese, English, Filipino, Indonesian, Javanese, Khmer, Lao, Malay, Sundanese, Tamil, Thai, and Vietnamese.
Key Capabilities & Features
- Multilingual Support: Extensive coverage of 13 Southeast Asian languages, with instruction tuning performed in English and several SEA languages.
- Llama 3.1 Architecture: Utilizes the robust Llama 3.1 decoder model with its default tokenizer.
- Extended Context Length: Features a substantial context window of 128k tokens, enabling processing of longer inputs and generating more coherent responses.
- Instruction Following: Evaluated on SEA-IFEval and SEA-MTBench, demonstrating capabilities in adhering to prompt constraints and engaging in multi-turn conversations.
- General Language Tasks: Performance assessed across various tasks via the SEA-HELM benchmark, including Question Answering, Sentiment Analysis, Translation, and Summarization.
Intended Use Cases
This model is particularly well-suited for applications requiring strong language understanding and generation in a multilingual Southeast Asian context. It can be used for:
- Developing chatbots and virtual assistants for SEA markets.
- Content creation and summarization in supported languages.
- Translation services between English and SEA languages.
- Research and development in multilingual NLP for the region.
Limitations
Users should be aware that, like many LLMs, Llama-SEA-LION-v3-70B-IT may exhibit hallucinations and occasional inconsistencies in reasoning. The model has not been aligned for safety, and developers are advised to perform their own safety fine-tuning and implement necessary security measures.