Benjamin Clavié
Bio
Benjamin Clavié is a member of technical staff at Mixedbread based in Tokyo, Japan. He's been working on information retrieval, with a focus on late-interaction (“ColBERT”) models and generally how we can design retrieval to match modern needs. In the past, Ben co-led the ModernBERT project at Answer.AI.
Session (1)
Benjamin Clavié is a researcher at Mixedbread specializing in encoder models and late-interaction retrieval, known for co-leading ModernBERT and creating RAGatouille and rerankers — tools that made state-of-the-art retrieval methods accessible to practitioners. His session is worth attending for deep, research-grounded perspectives on multimodal and multi-vector retrieval, encoder model design, and where information retrieval is heading beyond single-vector approaches.
GitHub
@bclavie- AnswerDotAI/RAGatouilleorg★ 3,937Easily use and train state of the art late-interaction retrieval methods (ColBERT) in any RAG pipeline.
- AnswerDotAI/rerankersorg★ 1,621A lightweight, low-dependency, unified API to use all common reranking and cross-encoder models.
- AnswerDotAI/byaldiorg★ 850Use late-interaction multi-modal models such as ColPali in just a few lines of code.
Recent writing (3)
- Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference · paper · Jul 2025
- Simple Projection Variants Improve ColBERT Performance · paper · Oct 2025
- Fantastic (small) Retrievers and How to Train Them: mxbai-edge-colbert-v0 Tech Report · paper · Oct 2025
Public activity researched automatically · as of Jun 2026