Model Overview
bylevlabs/stella_en_v5 is a 1.5 billion parameter model developed by blevlabs, showcasing solid performance across a diverse set of MTEB (Massive Text Embedding Benchmark) tasks. It features a substantial context length of 131072 tokens, enabling it to handle and process very long sequences of text effectively.
Key Capabilities
- Classification: Achieves high accuracy on various classification benchmarks, including 92.87% on MTEB AmazonCounterfactualClassification (en), 97.16% on MTEB AmazonPolarityClassification, and 89.79% on MTEB Banking77Classification.
- Retrieval: Demonstrates strong retrieval performance with notable scores such as 94.83% main score on MTEB FEVER and 76.67% on MTEB HotpotQA, indicating its ability to accurately retrieve relevant information from large datasets.
- Clustering: Shows competitive results in clustering tasks, with v-measure scores of 55.44% on MTEB ArxivClusteringP2P and 50.66% on MTEB ArxivClusteringS2S.
Good for
- Text Classification: Ideal for applications requiring accurate categorization of text, such as sentiment analysis, intent recognition, and spam detection.
- Information Retrieval: Well-suited for search engines, question-answering systems, and document retrieval where identifying relevant passages from extensive text is crucial.
- Long Context Processing: Its large context window makes it effective for tasks involving detailed document analysis, summarization of long articles, or understanding complex conversations.