topk-io/topk-embed-v1-small
Feature Extraction • 2B • Updated • 2.22k • 10
Troy Baker
jtroybaker
AI & ML interests
Predictive Maintenance, Reinforcement Learning, Natural Language Processing
Recent Activity
liked a model 5 days ago
topk-io/topk-embed-v1-small liked a model 15 days ago
XiaomiMiMo/MiMo-V2.6-Flash-RL liked a model 16 days ago
convaiinnovations/layaOrganizations
liked a
model 5 days ago
liked a
model 15 days ago
liked a
model 16 days ago
reacted to tomaarsen's post with 🔥 about 2 months ago
Post
4182
🚨 I've just published Sentence Transformers v6.0, introducing MultiVectorEncoder: ColBERT-style late interaction models are now a fourth model type, for training, inference, and interpretation, alongside the dense, sparse, and reranker models! Details:
Where a regular embedding model compresses a whole text into one vector, a multi-vector model keeps one vector per token and scores query against document with the MaxSim operator. That preserves token-level matching information that a single vector has to average away. It is also the state of the art for visual document retrieval, where a text query is matched against page images directly, charts and tables included, with no OCR step in between.
Any PyLate, Stanford ColBERT, or ColPali checkpoint loads straight into the same familiar API: model.encode_query(), model.encode_document(), and model.similarity() just work, whether the documents are texts or page images.
Does it help? LightOn trained LateOn (multi-vector) and DenseOn (dense) on the same data with the same 149M ModernBERT backbone, and the multi-vector model wins on 9 of the 13 NanoBEIR datasets: 0.6868 vs 0.6764 mean NDCG@10. The price is a bigger index, and the new HierarchicalTokenPooling module halves it at roughly no retrieval cost.
Antoine Chaffin, Raphaël Sourty, and I wrote a blog post walking through multi-vector models in practice: loading the various checkpoint formats, encoding and scoring, plugging them into a search stack, running them on page images, and keeping the index affordable. Check it out if you want to get started, or just point your Agent to the URL: https://huggingface.co/blog/multi-vector-encoder
pip install sentence-transformers==6.0.0
Release notes: https://github.com/huggingface/sentence-transformers/releases/tag/v6.0.0
Where a regular embedding model compresses a whole text into one vector, a multi-vector model keeps one vector per token and scores query against document with the MaxSim operator. That preserves token-level matching information that a single vector has to average away. It is also the state of the art for visual document retrieval, where a text query is matched against page images directly, charts and tables included, with no OCR step in between.
Any PyLate, Stanford ColBERT, or ColPali checkpoint loads straight into the same familiar API: model.encode_query(), model.encode_document(), and model.similarity() just work, whether the documents are texts or page images.
Does it help? LightOn trained LateOn (multi-vector) and DenseOn (dense) on the same data with the same 149M ModernBERT backbone, and the multi-vector model wins on 9 of the 13 NanoBEIR datasets: 0.6868 vs 0.6764 mean NDCG@10. The price is a bigger index, and the new HierarchicalTokenPooling module halves it at roughly no retrieval cost.
Antoine Chaffin, Raphaël Sourty, and I wrote a blog post walking through multi-vector models in practice: loading the various checkpoint formats, encoding and scoring, plugging them into a search stack, running them on page images, and keeping the index affordable. Check it out if you want to get started, or just point your Agent to the URL: https://huggingface.co/blog/multi-vector-encoder
pip install sentence-transformers==6.0.0
Release notes: https://github.com/huggingface/sentence-transformers/releases/tag/v6.0.0
liked 2
models about 2 months ago
liked 5
models 2 months ago
inclusionAI/Ling-3.0-flash
Text Generation • 127B • Updated • 13.3k • • 423
AtomicChat/Ling-3.0-flash-GGUF
Text Generation • 124B • Updated • 264k • 65
MiniMaxAI/MiniMax-H3
Image-Text-to-Video • 33B • Updated • 3.48M • • 5.93k
poolside/Laguna-S-2.1
Text Generation • 118B • Updated • 204k • • 1.04k
BlueBackup/Qwen3.6-35B-A3B-uncensored-heretic-IQ2_M
Text Generation • 35B • Updated • 827 • 9
liked 4
models 3 months ago
unsloth/Laguna-S-2.1-GGUF
Text Generation • 118B • Updated • 14.9k • 309
vmlinux/Qwen3.5-122B-A10B-ROCmFP4-iMatrix-GGUF
Text Generation • 122B • Updated • 112 • 6
thinkingmachines/Inkling
Image-Text-to-Text • 952B • Updated • 923k • • 1.81k
nvidia/Nemotron-Labs-Audex-30B-A3B
Text Generation • Updated • 394 • 179
liked 3
models 5 months ago
liked 2
models 6 months ago