arXiv benchmark finds int8 quantization nearly matches float32 on text embeddings
int8 quantization retains 94.4% baseline effectiveness at 32x storage reduction across 38 MTEB tasks.
Stories per day · 30 days
Relatedmteb2efficiency1hierarchical-questions1hierarchical-representations1multilingual1quantization1training-free1
3 stories, newest first · one card per story
Stories tagged #text-embeddings
Hierarchical questions improve interpretability of text embeddings
Hierarchical questions reduce the number humans must examine while maintaining task performance in QA-based embeddings.
Paired-residual removal improves multilingual LLM text embeddings
The inference-time method was tested on 18 languages and English bitext mining against more than 200 languages.