Llmware

Models by this creator

🤿

dragon-mistral-7b-v0

llmware

Total Score

109

The dragon-mistral-7b-v0 is part of the dRAGon ("Delivering RAG On ...") model series, which is a RAG-instruct trained model built on top of a Mistral-7B base model. The DRAGON models have been fine-tuned with the specific objective of fact-based question-answering over complex business and legal documents, with an emphasis on reducing hallucinations and providing short, clear answers for workflow automation. The Mistral-7B is a transformer model with architectural choices like Grouped-Query Attention, Sliding-Window Attention, and a Byte-fallback BPE tokenizer. The Mistral-7B-Instruct-v0.1 and Mistral-7B-Instruct-v0.2 are instruct fine-tuned versions of the Mistral-7B. Model Inputs and Outputs Inputs Text passages or documents Questions or prompts related to the provided context Outputs Short, fact-based answers to questions Extractions of key information from documents Concise summaries of complex content Capabilities The dragon-mistral-7b-v0 model is designed for enterprise automation use cases, especially in knowledge-intensive industries like financial services, legal, and regulatory. It excels at common RAG scenarios like question-answering, key-value extraction, and basic summarization without the need for complex instruction verbiage. What Can I Use It For? The dragon-mistral-7b-v0 model can be leveraged for a variety of business process automation tasks, such as contract review, risk analysis, customer service, and regulatory compliance. Its ability to provide clear, fact-based responses makes it well-suited for workflow optimization in industries that rely on comprehensive document understanding. Things to Try Some ideas for trying out the dragon-mistral-7b-v0 model: Integrate it into a document management system to automate the extraction of key information from legal or financial documents. Use it to build a virtual assistant that can answer questions and summarize content for employees, reducing the need for manual research. Explore its performance on specialized datasets or benchmarks related to your industry to assess its fit for your specific use cases.

Read more

Updated 5/17/2024

🔍

dragon-yi-6b-v0

llmware

Total Score

58

The dragon-yi-6b-v0 is part of the dRAGon ("Delivering RAG On ...") model series, which is a RAG-instruct trained model built on top of a Yi-6B base model. According to the maintainer llmware, DRAGON models have been fine-tuned with the specific objective of fact-based question-answering over complex business and legal documents, with an emphasis on reducing hallucinations and providing short, clear answers for workflow automation. The model has been evaluated against the RAG-Instruct-Benchmark-Tester and achieved an accuracy score of 99.5 out of 100, with high performance on tasks like boolean questions, math/logic, and complex questions. Importantly, no hallucinations were observed during the test runs. Similar models in the dRAGon series include the dragon-mistral-7b-v0, which is built on top of a Mistral-7B base model. Model inputs and outputs Inputs Text passage context Specific questions or instructions based on the text passage Outputs Fact-based, short and clear responses to the provided questions or instructions Capabilities include question-answering, key-value extraction, and basic summarization Capabilities The dragon-yi-6b-v0 model demonstrates strong performance on fact-based question-answering tasks, especially over complex business and legal documents. It is able to provide concise and accurate responses while avoiding hallucinations, making it well-suited for enterprise automation use cases in knowledge-intensive industries like finance, legal, and regulatory. What can I use it for? The dragon-yi-6b-v0 model is designed for enterprise automation use cases, particularly in industries that deal with complex information sources, such as financial services, legal, and regulatory. The model's capabilities in question-answering, key-value extraction, and basic summarization make it a valuable tool for tasks like document analysis, information retrieval, and workflow automation. Things to try One interesting aspect of the dragon-yi-6b-v0 model is its emphasis on reducing hallucinations, which is a common challenge in large language models. Developers and researchers could explore the techniques used by the maintainer, llmware, to achieve this, and potentially apply similar approaches to other models or domains. Another area to explore would be the model's performance on different types of business and legal documents, as well as its ability to handle complex queries and instructions. Developers could test the model's limits and identify areas for further improvement or refinement.

Read more

Updated 5/17/2024