🏗️ Building on HF

Tom Aarsen

tomaarsen

2302 700 2043

https://linkedin.com/in/tomaarsen

AI & ML interests

NLP: text embeddings, information retrieval, named entity recognition, few-shot text classification

Recent Activity

updated a collection 3 minutes ago

Encoder comparison after equivalent retrieval fine-tuning

updated a collection 17 minutes ago

Encoder comparison after equivalent retrieval fine-tuning

updated a collection 18 minutes ago

Encoder comparison after equivalent retrieval fine-tuning

View all activity

Organizations

reacted to tardellirs's post with 👍😎🔥 1 day ago

Post

4,815

Crivo-150M is now on the Pareto frontier of the Decision Index 🎯

Scored by the maintainers on the private tests: Full 22.21, the best of all submissions at or below 150M and higher than every model below 0.75B, including models up to 4× its size. Median latency on the board's GPU: 13 ms.

It is a 149M typed-decision encoder (Ettin / ModernBERT): a state plus choice, noul or score questions in, a probability for every option out, in one forward pass. Trained on open data only.

Next: tasks from new domains, the weak spot of every small encoder on the board.

tardellirs/crivo-150m
multimodalart/jev-decision-index

Thanks to @multimodalart for the Decision Index and its open kit, @orionweller and the Ettin team, @tomaarsen for the Ettin rerankers, @hotchpotch for the Bekko System One data and @sileod for tasksource.

posted an update about 2 months ago

Post

4,291

🚨 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

reacted to BramVanroy's post with 👍 2 months ago

Post

2,814

**I benchmarked HF buckets against https access for Common Crawl.**

Took me a while to get round to do this but I benchmarked access to Common Crawl via https vs hf buckets. Both experiments were run at night in Europe. I do not think other hardware problems were impacting the speeds since CPU processing time of the non-download pipeline components were highly similar (within 2% identical) and below only the WarcReader speeds of datatrove are used.

Experiment: selected 5 disjoint samples of 64 files each (randomly from the latest crawl; 20,499 docs/file). Those five batches were then processed by 32 single-core tasks with 4GB/core (five batches to calculate CIs). Paired experiment between using https and hf bucket.

- https: 40.0 [39.3-40.6] (seconds per WARC file)
- hf bucket: 172.0 [122.7-221.2]

That is a difference of about 4x in streaming speed. You'll see that https is also more stable (smaller CI).

I also ran raw throughput tests to the endpoints to measure rate limiting (64MiB transfer at 8/32/128/256 concurrent readers) and rate limiting seems not an issue for either: at any of those parallel reader numbers, their respective speeds stay about the same.

Note that, given CC scale, this is still a small test. Rate limiting may become more obvious when processing a full crawl. I do not know whether the https endpoint vs HF bucket will shut you out earlier with which limits.

posted an update 5 months ago

Post

1,746

🤗 Announcing the Ettin Reranker family: six new state-of-the-art CrossEncoder rerankers for search from 17M to 1B parameters, plus the full training data and the ~150-line recipe. Built on the Ettin ModernBERT encoders, Apache 2.0. Details:

All six were trained with the same single-stage pointwise MSE distillation recipe, with mixedbread-ai/mxbai-rerank-large-v2 (1.54B) as the teacher. Only the learning rate and per-device batch size change between sizes. The 1B student matches the teacher within 0.0001 NDCG@10 on MTEB(eng, v2) Retrieval, the 150M is the strongest reranker I tested in the under-600M range, and the 17M beats the 33M ms-marco-MiniLM-L12-v2 by +0.051 NDCG@10 at roughly half the parameter count.

Speed matters as much as quality for a reranker, since it determines whether the model fits the latency budget between retrieval and showing results. Our 17M is the fastest reranker in the whole comparison at 7517 pairs/sec on an H100. Our 150M runs 2.3x faster than the two other 150M ModernBERT-base rerankers (gte-reranker-modernbert-base and granite-embedding-reranker-english-r2) because the modular Transformer module propagates unpadded inputs through every layer rather than just the FA2 attention kernel. And our 1B is 2.4x faster than its 1.5B teacher while matching it on quality.

I bootstrapped the training recipe with the new train-sentence-transformers Agent Skill shipped in Sentence Transformers v5.5.0. Install it with hf skills add train-sentence-transformers --claude and ask Claude Code (or Codex / Cursor / Gemini CLI) to fine-tune a SentenceTransformer, CrossEncoder, or SparseEncoder model on your data.

I wrote a blog post walking through usage, results across six embedder pairings, the speed story, and the complete training script. Check it out, or just point your Agent to the URL:

https://huggingface.co/blog/ettin-reranker

Collection: https://huggingface.co/collections/cross-encoder/ettin-rerankers

replied to their post 5 months ago

I haven't tried it with a lot of custom data yet, I've mostly just let it decide which datasets to use for e.g. "German retrieval" or "Korean clustering" or "Code search" etc., but I'm quite confident it'll work nicely if you ask it to look at your own data. It'll likely try and pitch some data preprocessing options to turn it into e.g. pairs or triplets for training.

posted an update 5 months ago

Post

658

🤖 I've just published Sentence Transformers v5.5.0, headlined by a new train-sentence-transformers Agent Skill that lets your AI coding agent (Claude Code, Codex, Cursor, Gemini CLI, ...) train and finetune embedding, reranker, and sparse encoder models for you. Plus training losses & fixes. Details:

The skill bundles curated guidance for the whole training workflow across all three model types: base model selection, loss and evaluator choice, hard-negative mining, distillation, LoRA, Matryoshka, multilingual training, static embeddings, etc. It also ships production-ready training template scripts the agent can adapt. Install it with hf skills add train-sentence-transformers, then just describe what you want, e.g. "finetune a reranker on my (question, answer) pairs, mine hard negatives, and push it to the Hub".

On the loss side: EmbedDistillLoss is a new embedding-level distillation loss for SentenceTransformer. Instead of distilling teacher scores like MarginMSELoss, it aligns the student's embeddings directly with pre-computed teacher embeddings, wtih an optional learnable projection for when the student and teacher dimensions differ. Second, ADRMSELoss is a new listwise learning-to-rank loss for CrossEncoder from the Rank-DistilLLM paper, aimed at the LLM-distillation reranking setting.

encode() and predict() also gained a per-call processing_kwargs override, so you can change processor settings like max_length, a vision-language model's image resolution, or a video's fps, for a single call without rebuilding the model.

The Agent Skill is the part of this release I'm most keen for people to try. Curious to hear how it works for you. I've been using it myself a lot to quickly set up some training runs that immediately use a bunch of best practices.

> pip install sentence-transformers==5.5.0
> hf skills add train-sentence-transformers

The full release notes: https://github.com/huggingface/sentence-transformers/releases/tag/v5.5.0

3 replies

posted an update 6 months ago

Post

1,148

🌐 I've just published Sentence Transformers v5.4 to make the project fully multimodal for embeddings and reranking. The release also includes a modular CrossEncoder, and automatic Flash Attention 2 input flattening. Details:

You can now use SentenceTransformer and CrossEncoder with text, images, audio, and video, with the same familiar API. That means you can compute embeddings for an image and a text query using model.encode(), compare them with model.similarity(), and it just works. Models like Qwen3-VL-Embedding-2B and jinaai/jina-reranker-m0 are supported out of the box.

Beyond multimodal, I also fully modularized the CrossEncoder class. It's now a torch.nn.Sequential of composable modules, just like SentenceTransformer has been. This unlocked support for generative rerankers (CausalLM-based models like mxbai-rerank-v2 and the Qwen3 rerankers) via a new LogitScore module, which wasn't possible before without custom code.

Also, Flash Attention 2 now automatically skips padding for text-only inputs. If your batch has a mix of short and long texts, this gives you a nice speedup and lower VRAM usage for free.

I wrote a blog post walking through the multimodal features with practical examples. Check it out if you want to get started, or just point your Agent to the URL: https://huggingface.co/blog/multimodal-sentence-transformers

This release has set up the groundwork for more easily introducing late-interaction models (both text-only and multimodal) into Sentence Transformers in the next major release. I'm looking forward to it!

posted an update 10 months ago

Post

4,522

🐦‍🔥 I've just published Sentence Transformers v5.2.0! It introduces multi-processing for CrossEncoder (rerankers), multilingual NanoBEIR evaluators, similarity score outputs in mine_hard_negatives, Transformers v5 support and more. Details:

- CrossEncoder multi-processing: Similar to SentenceTransformer and SparseEncoder, you can now use multi-processing with CrossEncoder rerankers. Useful for multi-GPU and CPU settings, and simple to configure: just device=["cuda:0", "cuda:1"] or device=["cpu"]*4 on the model.predict or model.rank calls.

- Multilingual NanoBEIR Support: You can now use community translations of the tiny NanoBEIR retrieval benchmark instead of only the English one, by passing dataset_id, e.g. dataset_id="lightonai/NanoBEIR-de" for the German benchmark.

- Similarity scores in Hard Negatives Mining: When mining for hard negatives to create a strong training dataset, you can now pass output_scores=True to get similarity scores returned. This can be useful for some distillation losses!

- Transformers v5: This release works with both Transformers v4 and the upcoming v5. In the future, Sentence Transformers will only work with Transformers v5, but not yet!

- Python 3.9 deprecation: Now that Python 3.9 has lost security support, Sentence Transformers no longer supports it.

Check out the full changelog for more details: https://github.com/huggingface/sentence-transformers/releases/tag/v5.2.0

I'm quite excited about what's coming. There's a huge draft PR with a notable refactor in the works that should bring some exciting support. Specifically, better multimodality, rerankers, and perhaps some late interaction in the future!

posted an update 12 months ago

Post

4,745

🤗 Sentence Transformers is joining Hugging Face! 🤗 This formalizes the existing maintenance structure, as I've personally led the project for the past two years on behalf of Hugging Face! Details:

Today, the Ubiquitous Knowledge Processing (UKP) Lab is transferring the project to Hugging Face. Sentence Transformers will remain a community-driven, open-source project, with the same open-source license (Apache 2.0) as before. Contributions from researchers, developers, and enthusiasts are welcome and encouraged. The project will continue to prioritize transparency, collaboration, and broad accessibility.

Read our full announcement for more details and quotes from UKP and Hugging Face leadership: https://huggingface.co/blog/sentence-transformers-joins-hf

We see an increasing wish from companies to move from large LLM APIs to local models for better control and privacy, reflected in the library's growth: in just the last 30 days, Sentence Transformer models have been downloaded >270 million times, second only to transformers.

I would like to thank the UKP Lab, and especially Nils Reimers and Iryna Gurevych, both for their dedication to the project and for their trust in myself, both now and two years ago. Back then, neither of you knew me well, yet you trusted me to take the project to new heights. That choice ended up being very valuable for the embedding & Information Retrieval community, and I think this choice of granting Hugging Face stewardship will be similarly successful.

I'm very excited about the future of the project, and for the world of embeddings and retrieval at large!

1 reply

posted an update about 1 year ago

Post

5,936

ModernBERT goes MULTILINGUAL! One of the most requested models I've seen, The Johns Hopkins University's CLSP has trained state-of-the-art massively multilingual encoders using the ModernBERT architecture: mmBERT.

Model details:
- 2 model sizes:
- jhu-clsp/mmBERT-small
- jhu-clsp/mmBERT-base
- Uses the ModernBERT architecture, but with the Gemma2 multilingual tokenizer (so: flash attention, alternating global/local attention, unpadding/sequence packing, etc.)
- Maximum sequence length of 8192 tokens, on the high end for encoders
- Trained on 1833 languages using DCLM, FineWeb2, and many more sources
- 3 training phases: 2.3T tokens pretraining on 60 languages, 600B tokens mid-training on 110 languages, and 100B tokens decay training on all 1833 languages.
- Both models are MIT Licensed, and the full datasets and intermediary checkpoints are also publicly released

Evaluation details:
- Very competitive with ModernBERT at equivalent sizes on English (GLUE, MTEB v2 English after finetuning)
- Consistently outperforms equivalently sized models on all Multilingual tasks (XTREME, classification, MTEB v2 Multilingual after finetuning)
- In short: beats commonly used multilingual base models like mDistilBERT, XLM-R (multilingual RoBERTa), multilingual MiniLM, etc.
- Additionally: the ModernBERT-based mmBERT is much faster than the alternatives due to its architectural benefits. Easily up to 2x throughput in common scenarios.

Check out the full blogpost with more details. It's super dense & gets straight to the point: https://huggingface.co/blog/mmbert

Based on these results, mmBERT should be the new go-to multilingual encoder base models at 300M and below. Do note that the mmBERT models are "base" models, i.e. they're currently only trained to perform Mask Filling. They'll need to be finetuned for downstream tasks like semantic search, classification, clustering, etc.

replied to s-emanuilov's post about 1 year ago

Indeed, prompts are done automatically with MTEB! The transformers version does likely mean that you'll get better results with an update

replied to s-emanuilov's post about 1 year ago

Quick question, were you on the recommended transformers version (https://huggingface.co/blog/embeddinggemma#sentence-transformers), and did you use the prompts (i.e. either call the model with model.encode_query/model.encode_document/model.encode(..., prompt_name="query") in Sentence Transformers?

There's a few ways for it to quietly go wrong!

posted an update about 1 year ago

Post

4,594

😎 I just published Sentence Transformers v5.1.0, and it's a big one. 2x-3x speedups of SparseEncoder models via ONNX and/or OpenVINO backends, easier distillation data preparation with hard negatives mining, and more:

1️⃣ Faster ONNX and OpenVINO backends for SparseEncoder models
Usage is as simple as backend="onnx" or backend="openvino" when initializing a SparseEncoder to get started, but I also included utility functions for optimization, dynamic quantization, and static quantization, plus benchmarks.

2️⃣ New n-tuple-scores output format from mine_hard_negatives
This new output format is immediately compatible with the MarginMSELoss and SparseMarginMSELoss for training SentenceTransformer, CrossEncoder, and SparseEncoder losses.

3️⃣ Gathering across devices
When doing multi-GPU training using a loss that has in-batch negatives (e.g. MultipleNegativesRankingLoss), you can now use gather_across_devices=True to load in-batch negatives from the other devices too! Essentially a free lunch, pretty big impact potential in my evals.

4️⃣ Trackio support
If you also upgrade transformers, and you install trackio with pip install trackio, then your experiments will also automatically be tracked locally with trackio. Just open up localhost and have a look at your losses/evals, no logins, no metric uploading.

5️⃣ MTEB Documentation
We've added some documentation on evaluating SentenceTransformer models properly with MTEB. It's rudimentary as the documentation on the MTEB side is already great, but it should get you started.

Plus many more smaller features & fixes (crash fixes, compatibility with datasets v4, FIPS compatibility, etc.).

See the full release notes here: https://github.com/UKPLab/sentence-transformers/releases/tag/v5.1.0

Big thanks to all of the contributors for helping with the release, many of the features from this release were proposed by others. I have a big list of future potential features that I'd love to add, but I'm

posted an update over 1 year ago

Post

3,233

‼️Sentence Transformers v5.0 is out! The biggest update yet introduces Sparse Embedding models, encode methods improvements, Router module for asymmetric models & much more. Sparse + Dense = 🔥 hybrid search performance! Details:

1️⃣ Sparse Encoder Models
Brand new support for sparse embedding models that generate high-dimensional embeddings (30,000+ dims) where <1% are non-zero:

- Full SPLADE, Inference-free SPLADE, and CSR architecture support
- 4 new modules, 12 new losses, 9 new evaluators
- Integration with @elastic-co , @opensearch-project , @NAVER LABS Europe, @qdrant , @IBM , etc.
- Decode interpretable embeddings to understand token importance
- Hybrid search integration to get the best of both worlds

2️⃣ Enhanced Encode Methods & Multi-Processing
- Introduce encode_query & encode_document automatically use predefined prompts
- No more manual pool management - just pass device list directly to encode()
- Much cleaner and easier to use than the old multi-process approach

3️⃣ Router Module & Advanced Training
- Router module with different processing paths for queries vs documents
- Custom learning rates for different parameter groups
- Composite loss logging - see individual loss components
- Perfect for two-tower architectures

4️⃣ Comprehensive Documentation & Training
- New Training Overview, Loss Overview, API Reference docs
- 6 new training example documentation pages
- Full integration examples with major search engines
- Extensive blogpost on training sparse models

Read the comprehensive blogpost about training sparse embedding models: https://huggingface.co/blog/train-sparse-encoder

See the full release notes here: https://github.com/UKPLab/sentence-transformers/releases/v5.0.0

What's next? We would love to hear from the community! What sparse encoder models would you like to see? And what new capabilities should Sentence Transformers handle - multimodal embeddings, late interaction models, or something else? Your feedback shapes our roadmap!

reacted to giadap's post with 🔥 over 1 year ago

Post

1,873

🤗 Just published: "Consent by Design" - exploring how we're building better consent mechanisms across the HF ecosystem!

Our research shows open AI development enables:
- Community-driven ethical standards
- Transparent accountability
- Context-specific implementations
- Privacy as core infrastructure

Check out our Space Privacy Analyzer tool that automatically generates privacy summaries of applications!

Effective consent isn't about perfect policies; it's about architectures that empower users while enabling innovation. 🚀

Read more: https://huggingface.co/blog/giadap/consent-by-design

3 replies

posted an update over 1 year ago

Post

4,283

I just released Sentence Transformers v4.1; featuring ONNX and OpenVINO backends for rerankers offering 2-3x speedups and improved hard negatives mining which helps prepare stronger training datasets. Details:

🏎️ ONNX, OpenVINO, Optimization, Quantization
- I've added ONNX and OpenVINO support with just one extra argument: "backend" when loading the CrossEncoder reranker, e.g.: CrossEncoder("cross-encoder/ms-marco-MiniLM-L6-v2", backend="onnx")
- The export_optimized_onnx_model, export_dynamic_quantized_onnx_model, and export_static_quantized_openvino_model functions now work with CrossEncoder rerankers, allowing you to optimize (e.g. fusions, gelu approximations, etc.) or quantize (int8 weights) rerankers.
- I've uploaded ~340 ONNX & OpenVINO models for all existing models under the cross-encoder Hugging Face organization. You can use these without having to export when loading.

⛏ Improved Hard Negatives Mining
- Added 'absolute_margin' and 'relative_margin' arguments to mine_hard_negatives.
- absolute_margin ensures that sim(query, negative) < sim(query, positive) - absolute_margin, i.e. an absolute margin between the negative & positive similarities.
- relative_margin ensures that sim(query, negative) < sim(query, positive) * (1 - relative_margin), i.e. a relative margin between the negative & positive similarities.
- Inspired by the excellent NV-Retriever paper from NVIDIA.

And several other small improvements. Check out the full release notes here: https://github.com/UKPLab/sentence-transformers/releases/tag/v4.1.0

With this release, I introduce near-feature parity between the SentenceTransformer embedding & CrossEncoder reranker models, which I've wanted to do for quite some time! With rerankers very strongly supported now, it's time to look forward to other useful architectures!

reacted to AdinaY's post with 🔥 over 1 year ago

Post

2,584

Let's check out the latest releases from the Chinese community in March!

👉 https://huggingface.co/collections/zh-ai-community/march-2025-releases-from-the-chinese-community-67c6b479ebb87abbdf8e2e76

✨MLLM
> R1 Omni by Alibaba Tongyi - 0.5B
> Qwen2.5 Omni by Alibaba Qwen - 7B with apache2.0

🖼️Video
> CogView-4 by ZhipuAI - Apacha2.0
> HunyuanVideo-I2V by TencentHunyuan
> Open Sora2.0 - 11B with Apache2.0
> Stepvideo TI2V by StepFun AI - 30B with MIT license

🎵Audio
> DiffDiffRhythm - Apache2.0
> Spark TTS by SparkAudio - 0.5B

⚡️Image/3D
> Hunyuan3D 2mv/2mini (0.6B) by @TencentHunyuan
> FlexWorld by ByteDance - MIT license
> Qwen2.5-VL-32B-Instruct by Alibaba Qwen - Apache2.0
> Tripo SG (1.5B)/SF by VastAIResearch - MIT license
> InfiniteYou by ByteDance

> LHM by Alibaba AIGC team - Apache2.0
> Spatial LM by ManyCore

🧠Reasoning
> QwQ-32B by Alibaba Qwen - Apache2.0
> Skywork R1V - 38B with MIT license
> RWKV G1 by RWKV AI - 0.1B pure RNN reasoning model with Apache2.0
> Fin R1 by SUFE AIFLM Lab - financial reasoning

🔠LLM
> DeepSeek v3 0324 by DeepSeek -MIT license
> Babel by Alibaba DAMO - 9B/83B/25 languages

4 replies

posted an update over 1 year ago

Post

2,805

‼️Sentence Transformers v4.0 is out! You can now train and finetune reranker models with multi-GPU training, bf16 support, loss logging, callbacks & much more. I also prove that finetuning on your domain helps much more than you might think.

1️⃣ Reranker Training Refactor
Reranker models can now be trained using an extensive trainer with a lot of powerful features:
- MultiGPU Training (Data Parallelism (DP) and Distributed Data Parallelism (DDP))
- bf16 training support; loss logging
- Evaluation datasets + evaluation loss
- Improved callback support + an excellent Weights & Biases integration
- Gradient checkpointing, gradient accumulation
- Model card generation
- Resuming from a training checkpoint without performance loss
- Hyperparameter Optimization
and much more!

Read my detailed blogpost to learn about the components that make up this new training approach: https://huggingface.co/blog/train-reranker
Notably, the release is fully backwards compatible: all deprecations are soft, meaning that they still work but emit a warning informing you how to upgrade.

2️⃣ New Reranker Losses
- 11 new losses:
- 2 traditional losses: BinaryCrossEntropy and CrossEntropy
- 2 distillation losses: MSE and MarginMSE
- 2 in-batch negatives losses: MNRL (a.k.a. InfoNCE) and CMNRL
- 5 learning to rank losses: Lambda, p-ListMLE, ListNet, RankNet, ListMLE

3️⃣ New Reranker Documentation
- New Training Overview, Loss Overview, API Reference docs
- 5 new, 1 refactored training examples docs pages
- 13 new, 6 refactored training scripts
- Migration guides (2.x -> 3.x, 3.x -> 4.x)

4️⃣ Blogpost
Alongside the release, I've written a blogpost where I finetune ModernBERT on a generic question-answer dataset. My finetunes easily outperform all general-purpose reranker models, even models 4x as big. Finetuning on your domain is definitely worth it: https://huggingface.co/blog/train-reranker

See the full release notes here: https://github.com/UKPLab/sentence-transformers/releases/v4.0.1

Tom Aarsen

AI & ML interests

Recent Activity

Organizations

tomaarsen's activity