Language Large Models (LLMs) have taken the world by storm with their human-like capabilities and features. The latest addition to the long list of LLMs, the GPT-4 model, has greatly increased the usefulness of ChatGPT due to its multimodal nature. This latest version takes input in the form of text and images and is already used to build quality websites and chatbots. Recently, a new paradigm has been introduced to democratize ChatGPT, i.e. to make it more accessible and accessible to a wider audience, regardless of language or geographical restrictions.
This last model, called Phoenix, aims for competitive performance not only in English and Chinese but also in languages with limited resources, such as Latin and non-Latin languages. Phoenix, a multilingual LLM that performs great between English and Chinese open source models, has just been released to make ChatGPT available in places with restrictions imposed by OpenAI or local governments.
The author described the significance of the Phoenix as follows –
🚀 Join the fastest ML Subreddit community
- Phoenix is introduced as the first open source, multilingual, and democratized ChatGPT template. This was achieved using rich multilingual data at the initial training and final education stages.
- The team performed an adaptation following instructions in multiple languages, with an emphasis on non-Latin languages. Both instructions and conversation data were used to train the model. This approach allows Phoenix to take advantage of both, enabling it to generate context-appropriate, coherent responses in different language settings.
- Phoenix is a large Chinese first-class language paradigm that has achieved performance close to ChatGPT. Its Latin version, Chimera, is competitive in English.
- The authors claim that Phoenix is an open source SOTA large language model for many languages other than Chinese and English.
- Phoenix is among the first to systematically assess large-scale LLM, using both automatic and human assessments and evaluating multiple aspects of language generations.
Phoenix has demonstrated superior performance compared to currently open source Chinese-language LLMs, including models such as BELLE and Chinese-LLaMA-Alpaca. In other non-Latin languages such as Arabic, Japanese, and Korean, Phoenix is vastly superior to existing models. Phoenix did SOTA results for Vicuna, an open source chatbot with 13B parameters that was trained by tuning LLaMA to user-shared conversations.
This is because Phoenix had to pay a polyglot tax when dealing with non-Latin or non-Cyrillic languages. Multilingual tax refers to the performance degradation that a multilingual model may experience when text is generated in languages other than its base language. The team considered it worthwhile to pay the tax to democratization as a way to cater to small groups speaking relatively low-resource languages. The team proposed a tax-free Phoenix solution: Chimera for multilingual tax relief in Latin and Cyrillic. This includes replacing Phoenix’s backbone with LLaMA. In English, Chimera GPT-4 impressed with ChatGPT quality of 96.6%.
Phoenix looks promising because of its multilingual potential and its ability to enable people from diverse linguistic backgrounds to harness the power of language models to meet their specific needs.
scan the paper And github. Don’t forget to join 19k+ML Sub RedditAnd discord channelAnd Email newsletter, where we share the latest AI research news, cool AI projects, and more. If you have any questions regarding the above article or if we’ve missed anything, feel free to email us at Asif@marktechpost.com
🚀 Check out 100’s AI Tools in the AI Tools Club
Tania Malhotra is a final year from University of Petroleum and Energy Studies, Dehradun, pursuing a BTech in Computer Science Engineering with a specialization in Artificial Intelligence and Machine Learning.
She is passionate about data science and has good analytical and critical thinking, along with a keen interest in acquiring new skills, leading groups, and managing work in an organized manner.
🚀 Join the fastest ML Subreddit community