Maxence Lasbordes's picture

In a Training Loop 🔄

Maxence Lasbordes

MaxLSB

·

AI & ML interests

Large Language Models

Organizations

upvoted an article 25 days ago

Article

Ulysses Sequence Parallelism: Training with Million-Token Contexts

Mar 9

•

25

upvoted a collection about 2 months ago

OriOn 💫

Visual long document VLMs based on Mistral-Small-3.1-24B-Instruct-2503 and Qwen3-VL-32B-Instruct • 5 items • Updated 1 day ago • 4

upvoted an article about 2 months ago

Article

LateOn-Code & ColGrep: LightOn unveils state-of-the-art code retrieval models and code search tooling

Feb 12

•

53

upvoted an article 2 months ago

Article

LightOnOCR-2-1B: a lightweight high-performance end-to-end OCR model family

Jan 19

•

89

upvoted a collection 4 months ago

Olmo 3 Post-training

All artifacts for post-training Olmo 3. Datasets follow the model that resulted from training on them. • 32 items • Updated Dec 23, 2025 • 51

upvoted a paper 6 months ago

Luth: Efficient French Specialization for Small Language Models and Cross-Lingual Transfer

Paper • 2510.05846 • Published Oct 7, 2025 • 3

upvoted 2 collections 7 months ago

🎯 Liquid Nanos

Library of task-specific models: https://www.liquid.ai/blog/introducing-liquid-nanos-frontier-grade-performance-on-everyday-devices • 26 items • Updated 2 days ago • 112

Luth Models

5 items • Updated Sep 26, 2025 • 2

upvoted a collection 8 months ago

Luth x LFM2

6 items • Updated Sep 24, 2025 • 15

upvoted an article 8 months ago

Article

Luth: Efficient French Specialization for Small Language Models

Aug 11, 2025

•

19

upvoted a collection 8 months ago

Luth x Qwen3

4 items • Updated Sep 24, 2025 • 7

upvoted 2 collections 9 months ago

Luth Datasets

2 items • Updated Mar 2 • 5

🧠 SmolLM3

Smol, multilingual, long-context reasoner • 14 items • Updated Oct 9, 2025 • 99

upvoted a paper 9 months ago

Seq vs Seq: An Open Suite of Paired Encoders and Decoders

Paper • 2507.11412 • Published Jul 15, 2025 • 31

upvoted 3 collections 9 months ago

Tulu 3 Datasets

All datasets released with Tulu 3 -- state of the art open post-training recipes. • 32 items • Updated Mar 2 • 96

💧 LFM2

LFM2 is a new generation of hybrid models, designed for on-device deployment. • 28 items • Updated 2 days ago • 153

Falcon-H1

Falcon-H1 Family of Hybrid-Head Language Models (Transformer-SSM), including 0.5B, 1.5B, 1.5B-Deep, 3B, 7B, and 34B (pretrained & instruction-tuned). • 33 items • Updated Mar 2 • 59

upvoted 2 articles 9 months ago

Article

Efficient MultiModal Data Pipeline

+3

Jul 8, 2025

•

70

Article

SmolLM3: smol, multilingual, long-context reasoner

+21

Jul 8, 2025

•

768

upvoted a paper 10 months ago

Magistral

Paper • 2506.10910 • Published Jun 12, 2025 • 68