🔄 In a Training Loop

Maxence Lasbordes

MaxLSB

lightonai

·

AI & ML interests

Large Language Models

Recent Activity

updated a dataset about 19 hours ago

kurakurai/Dolci-Instruct-SFT-Tool-Use-FR

published a dataset about 19 hours ago

kurakurai/Dolci-Instruct-SFT-Tool-Use-FR

updated a dataset 3 days ago

kurakurai/instruction-following-RL-FR-broadened

View all activity

Organizations

upvoted an article 8 days ago

Article

One Adapter, Both Modalities: Field Notes from Building and Serving a Multimodal Reranker

lightonai

•

12 days ago

• 17

upvoted a paper 2 months ago

Rethinking the Multilingual Reasoning Gap with Layer Swap

Paper • 2605.26735 • Published May 26 • 3

upvoted an article 3 months ago

Article

DenseOn with the LateOn: Open State-of-the-Art Single and Multi-Vector Models

lightonai

•

Apr 21

• 44

upvoted an article 4 months ago

Article

Ulysses Sequence Parallelism: Training with Million-Token Contexts

kashif, stas

•

Mar 9

• 32

upvoted a collection 5 months ago

OriOn 💫

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

upvoted 2 articles 6 months ago

Article

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

lightonai

•

Feb 12

• 58

Article

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

lightonai

•

Jan 19

• 101

upvoted a collection 8 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 • 59

upvoted a paper 10 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 10 months ago

🎯 Liquid Nanos

Library of task-specific models: https://www.liquid.ai/blog/introducing-liquid-nanos-frontier-grade-performance-on-everyday-devices • 34 items • Updated Jun 25 • 124

Luth Models

10 items • Updated 8 days ago • 2

upvoted a collection 11 months ago

Luth x LFM2

6 items • Updated 8 days ago • 15

upvoted an article 12 months ago

Article

Luth: Efficient French Specialization for Small Language Models

MaxLSB

•

Aug 11, 2025

• 21

upvoted 2 collections about 1 year ago

Luth Datasets

2 items • Updated 8 days ago • 5

🧠 SmolLM3

Smol, multilingual, long-context reasoner • 14 items • Updated Oct 9, 2025 • 106

upvoted a paper about 1 year ago

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

Paper • 2507.11412 • Published Jul 15, 2025 • 33

upvoted 3 collections about 1 year ago

Tulu 3 Datasets

All datasets released with Tulu 3 -- state of the art open post-training recipes. • 32 items • Updated Mar 2 • 102

💧 LFM2

LFM2 is a new generation of hybrid models, designed for on-device deployment. • 28 items • Updated Jun 25 • 155

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 • 60

upvoted an article about 1 year ago

Article

Efficient MultiModal Data Pipeline

+3

ariG23498, lusxvr, andito, sergiopaniego, pcuenq

•

Jul 8, 2025

• 73