DenseOn with the LateOn: Fully Open Dense and Late-Interaction Models for Multilingual, Long-Context, and Code Search

By Raphaël Sourty, Antoine Chaffin, Paulo Roberto Moura, A. Chatelain. 2026

Subfields: Information Retrieval & Search, Representation & Self-Supervised Learning

Citations in the corpus

• Embedding Models Measure in Peculiar Ways (2026, unknown): “These are granite-embedding-english-r2 (Awasthy et al., 2025), the multilingual granite-embedding-97m-multilingual-r2 and granite-embedding-311m-multilingual-r2 (Awasthy et al., 2026), and DenseOn (Sourty et al., 2026), an English retrieval model trained on openly released data.”
• Q2D-Web: A Large-Scale Benchmark for Retrieval in Agentic RAG Systems (2026, unknown): “Specifically, (Sourty et al., 2026) dense 0.3B 768 single 8k 2026-07 Nemotron-3-Embed-1B (Babakhin et al., 2026) dense 1.1B 2048 single 32k 2026-07 Nemotron-3-Embed-8B dense 8B 4096 single 32k 2026-07 pplx-embed-v1-0.”