Linkup Research has released SPARSEUP, an open-source sparse embedding model built on a 149 million-parameter ModernBERT backbone and distributed under the Apache 2.0 license. According to the team, the model posted an average nDCG@10 score of 56.4 on the BEIR-13 benchmark. It also described SPARSEUP as the strongest known sparse encoder under 150 million parameters that uses a public vocabulary. The model weights are hosted on Hugging Face and can be loaded through either Transformers or Sentence Transformers. Linkup Research said the release is meant to fill the sparse retrieval gap left after LightOn recently introduced DenseOn and LateOn, while keeping the same backbone family and fine-tuning data so the three retrieval styles can be compared side by side. On MS MARCO, SPARSEUP averages 47 non-zero terms per query and 190 per document. Using the Seismic inverted index, it reaches more than 97% recall in about 380 microseconds per query on a single thread.
Linkup Research has released SPARSEUP, an open-source sparse embedding model built on a 149 million-parameter ModernBERT backbone. The model is distributed under the Apache 2.0 license.
According to the team, SPARSEUP recorded an average nDCG@10 score of 56.4 on the BEIR-13 benchmark. It said the model is currently the strongest known sparse encoder under 150 million parameters that is based on a public vocabulary.
The model weights are hosted on Hugging Face and can be loaded through Transformers or Sentence Transformers. Linkup Research said SPARSEUP is intended to fill the missing sparse retrieval option after LightOn recently released DenseOn and LateOn, using the same backbone family and the same fine-tuning data so the three retrieval styles can be compared on a like-for-like basis.
On speed and sparsity, SPARSEUP averages 47 non-zero terms per query and 190 per document on the MS MARCO dataset. With the Seismic inverted index, it can reach more than 97% recall in about 380 microseconds per query on a single thread.
The item cited MarkTechPost as the source.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
