Hi there, I’m Ben 👋

I’m interested in NLP&IR, with a special focus on developing cool, smaller models to power LLM-based applications. You may know me from my frequent twitter rants about using late-interaction over dense vectors, or my love for encoders.

I do R&D at Answer.AI 👀, where every quest is a side-quest and scope-creep is encouraged. Currently, I’ve helped gather a really cool crowd of people to bring BERT to 2024 👀!

I’ve also built 🪤RAGatouille, a python library whose grand aim is to bridge the gap between state-of-the-art research code and commonly used practices and Rerankers, a library whose aim is to make it really easy to use just about any common reranking method, and swap them in and out painlessly.

Feel free to reach out if you’d like to chat!

[Answer.AI] Small but Mighty: Introducing answerai-colbert-small

Say hello to answerai-colbert-small-v1, a tiny ColBERT model that punches well above its weight.

[Answer.AI] JaColBERTv2.5🇯🇵: Optimising Retrieval Training for Lower-Resources Languages

Introducing JaColBERTv2.5🇯🇵, the new best Japanese retrieval model. Through this release, we present a thorough analysis to better understand what helps in training a good multi-vector retrieval model.

[Answer.AI] A little pooling goes a long way for multi-vector representations

Blog post offering a quick overview of ColBERT and how it works, and introducing an efficient pooling trick to alleviate the issues it faces.

Questions & Answer(s): thoughts and joining Answer.AI

If you’re old school, you can find a raw HTML version of this post here This is a fairly long, stream-of-thought post about how I currently (Sunday Feb 4, 2024) feel about the broader ML/NLP/IR ecosystem and its future. Everything here’s on-the-fly opinion and can & will change. In summary: I think ML developments are incredibly exciting, and we need to continue to work on bridging the gap between ML-as-a-commodity-for-ML-practitioners to ML-as-a-commodity-for-everyone....

Fio

日本語版は近日公開予定です（日本語を勉強中なので、間違いはご容赦ください！） Welcome to this stream-of-thoughts report on fio-base-japanese-v0.1, the first version of Fio, a family of Japanese sentence embeddings. These are all notes I took while training the models, vaguely ordered in relevant categories! I hope that they can be useful to anyone interested in Japanese embeddings. In short: Fio-v0.1 is currently (18/12/2023) the best similarity sentence embeddings model for Japanese, as well as the best overall monolingual model (similarity + retrieval)....