Aurélien Lac (@Aurelien_L_) header
Aurélien Lac (@Aurelien_L_) profile picture

AI Research Engineer @liquidai

Joined May 2012
Photo by Aurélien Lac (@Aurelien_L_)Photo by Aurélien Lac (@Aurelien_L_)Photo by Aurélien Lac (@Aurelien_L_)Photo by Aurélien Lac (@Aurelien_L_)Photo by Aurélien Lac (@Aurelien_L_)Photo by Aurélien Lac (@Aurelien_L_)Photo by Aurélien Lac (@Aurelien_L_)Photo by Aurélien Lac (@Aurelien_L_)Photo by Aurélien Lac (@Aurelien_L_)Photo by Aurélien Lac (@Aurelien_L_)
⚡ A lot of representation models still start from a massive generative VLM. We trained an encoder from scratch instead: native on text and images, multilingual, and built to be fast from the beginning. I’m super proud to share NeoMME with my friend @tonywu_71. 260M and 800M, no vision tower. We also fine-tuned a visual document retriever on top: NeoMME-Retriever.
1
12
4
75
13,680
GTA VI got some people banned over nudity on stream. With our decision models like d1.2 or d1.2-ultra, you could run a NSFW check on every image before it goes to the stream, for about $1 an hour with our API, in real time. Just as a family-friendly example, here I use one d1.2 API call per image to detect and blur people in swimsuits, but it works with any detection pattern 👀 (You could also do the same for sound since all our models are Omni!)
32
25
14
588
67,104
I’m so happy to see that d1-2B is widely used, and still the #1 trending decision model on Hugging Face 3 weeks after release 🤗
lately @liquidai is killing it with their small models out there, they release VLMs, Omni models as well as their decision models I use anything and everything they release through @Llama_app very grateful there's a player releasing tiny models everyone can run 🤗
1
4
43
1,889
total liquid cultural victory btw 🫡 huggingface.co/models?other=…
testing d1-Omni on my vision harness side project, ran 100 receipts for parsing misc info got dates with 98% acc, full address with 52%, company names with 85% I'd still opt for tiny OCR models by lighton etc for now, but this is promising direction
3
5
1
73
4,402
Aurélien Lac retweeted
Best API decision model: Liquid AI d1.2-ultra Most popular open decision model: Liquid AI d1-2B
1
10
1
93
3,846
Aurélien Lac retweeted
it's a skill issue to use frontier LLMs to detect objects or do vision instead, if you use LLMs as vision interns you can train a detector for $6 that runs at little cost. but your agent is a bad vision intern, it only knows old models that are proprietary and makes rookie mistakes so I built something for it, I basically distilled myself for you to use🤝 I put my favorite models from keypoint detection to OCR as a tool for your agent to call and I also made an e2e object detection trainer (from labelling to VLM as a judge to training and eval) if you're interested in it and the results, my talk at @aiDotEngineer WF is finally out, you can watch it here youtube.com/watch?v=dKcTBQzR… many thanks @swyx on the next post you can find the github repository, I also shipped decision model support and updated VLMs (LFM3, Qwen3.6/8)
30
59
4
586
30,440
Aurélien Lac retweeted
d1-Omni is a strict judge that is out to get false positives and negatives I used it in a VLM as a judge pipeline to judge labels from Qwen, it rejected (p<0.5) bboxes previously accepted by LFM2.5-VL and Gemma-4 it also has higher agreement rate with human annotations
5
5
1
92
4,820
Aurélien Lac retweeted
all modalities supported with many new output functions. enjoy d1.2 🚀 blog: liquid.ai/blog/d1-2
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
8
8
2
130
6,712
Big release today: d1.2-ultra and d1.2 are now live on the Liquid API ⚡️ They’re of course stronger than our previous iterations and come with super useful new capabilities: > Same typed answers from our open d1-2B and d1-Omni-2B: extract a number, a date or a name out of a document with the exact place it came from, bounding boxes around what you describe in a photo, estimates with a range, and select several options at once. > Audio 🎧: send a clip and ask about it, or just ask your question out loud!
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
10
6
1
73
3,260
We track hundreds of public and private evals internally, and d1.2-ultra is ahead of every decision model we’ve tested on them. But most of the work went into making it more reliable on real tasks, and able to handle more kinds of input, like audio.
1
5
191
Aurélien Lac retweeted
Introducing 𝙻𝚒𝚚𝚞𝚒𝚍𝙲𝚘𝚍𝚎𝚌, our ~74M-parameter streaming audio codec powering LFM3-Audio-2B and LFM3-Omni-2B. An audio codec turns speech into tokens a language model can predict, then converts those tokens back into audio. We built ours around two requirements: tokens that are easy to predict and fast conversion back into speech. On a single CPU thread on an Apple M4 Max, LiquidCodec runs at approximately: > 361× real time for encoding > 87× real time for decoding Both measurements use FP32 precision. 1/ 🧵
75
127
33
1,544
5,069,714
Aurélien Lac retweeted
What is an 𝗮𝘂𝗱𝗶𝗼 𝗰𝗼𝗱𝗲𝗰? (And why our audio team built their own) 𝗪𝗵𝗮𝘁 𝗶𝘀 𝗮𝗻 𝗮𝘂𝗱𝗶𝗼 𝗰𝗼𝗱𝗲𝗰? 1. Turns speech into tokens 2. Turns the tokens back into audio 𝗛𝗼𝘄 𝗱𝗼𝗲𝘀 𝗮𝗻 𝗮𝘂𝗱𝗶𝗼 𝗰𝗼𝗱𝗲𝗰 𝘄𝗼𝗿𝗸? → Encode: turn raw audio into feature vectors → Quantize: map those vectors into discrete tokens → Decode: reconstruct the waveform from the tokens 𝗪𝗵𝘆 𝗯𝘂𝗶𝗹𝗱 𝗼𝘂𝗿 𝗼𝘄𝗻? Our earlier audio models used a third-party tokenizer. But over time, its token quality and design choices started to limit us. 𝗟𝗶𝗾𝘂𝗶𝗱𝗖𝗼𝗱𝗲𝗰 is a ~74M-parameter streaming codec with: • 𝗛𝗶𝗴𝗵-𝗳𝗶𝗱𝗲𝗹𝗶𝘁𝘆: reconstructed audio is closer to the original. • 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗲𝗱 𝗳𝗼𝗿 𝘀𝗽𝗲𝗲𝗱: It's fast, even on-device. Learn more about how our audio team built it in our newest engineering blog: liquid.ai/blog/liquidcodec
Introducing 𝙻𝚒𝚚𝚞𝚒𝚍𝙲𝚘𝚍𝚎𝚌, our ~74M-parameter streaming audio codec powering LFM3-Audio-2B and LFM3-Omni-2B. An audio codec turns speech into tokens a language model can predict, then converts those tokens back into audio. We built ours around two requirements: tokens that are easy to predict and fast conversion back into speech. On a single CPU thread on an Apple M4 Max, LiquidCodec runs at approximately: > 361× real time for encoding > 87× real time for decoding Both measurements use FP32 precision. 1/ 🧵
2
14
1
137
4,944
Very nice update! While working on multimodal embeddings with NeoMME, @tonywu_71 and I got 255x smaller embeddings while keeping >95% of nDCG@10, using token pooling + asymmetric quantization, with no pooling-aware training at all. Training for it could push this much further, especially for images, where so many patches are near-empty. Bonus point: v6.3 now supports our late-interaction-kernels 🙌 github.com/hcompai/late-inte…
🤗 I've released Sentence Transformers v6.3 and wrote a blogpost to highlight its key new feature: token pooling-aware training! Train multi-vector models to preserve retrieval quality with fewer document vectors. Here's how 🧵 huggingface.co/blog/token-po…
3
2
38
1,013
Aurélien Lac retweeted
As a reminder, these were the compression results @Aurelien_L_ and I got by combining token pooling with asymmetric quantization with NeoMME. nitter.gq/tonywu_71/status/20955…
Replying to @tonywu_71
For NeoMME-Retriever-260M on ViDoRe v3, high-resolution late-interaction embeddings average about 1.5 MB per page. Token pooling and asymmetric quantization reduce them to 6 kB, 255x smaller, while retaining more than 95% of baseline nDCG@10. (9/N)
1
2
6
219
Aurélien Lac retweeted
I got a 1.9B model to watch a restaurant camera and tell staff which tables need cleaning 🍽️ It runs on my laptop. No cloud, and it generates zero tokens. Built over a weekend with @liquidai d1-2B
309
402
71
7,123
407,657
The anomaly of the LightOn R&D alumnis need to be studied honestly @iacopo_poli really created something special IFYKYK
6
8
1
57
3,472