Federico Bianchi

Postdoctoral Researcher at Stanford University

Stanford University

Biography

Hello! I am Federico Bianchi, I am a post-doctoral researcher at Stanford University with Prof. Dan Jurafsky.

You can find my work in the main track of different AI venues (e.g., NAACL, EACL, AAAI, ACL, ICLR, RecSys) and Journals (e.g., Cognitive Science, Nature PJQI, SWJ) and my work as also been featured in press or in company media outlets (see, 1, 2, 3, 4).

I have experience in building and deploying large language models (e.g., FashionCLIP), model understanding (e.g., VLMs Analysis) and fairness in recommender systems (e.g., EvalRS).

I recently joined the Giving What We Can initiative and donate monthly; You should consider joining, if you can. While I haven’t formally pledged yet, I hope to be able to do so soon.

Interests

• Artificial Intelligence
• Natural Language Processing
• Capybaras

Education

• PhD in Computer Science, 2020

  University of Milano-Bicocca

• MSc in Computer Science, 2016

  University of Milano-Bicocca

• BSc in Computer Science, 2014

  University of Milano-Bicocca

Featured Publications

Easily Accessible Text-to-Image Generation Amplifies Demographic Stereotypes at Large Scale

We show that text-to-image generation models can amplify stereotypes at large scale.

Federico Bianchi, Pratyusha Kalluri, Esin Durmus, Faisal Ladhak, Myra Cheng, Debora Nozza, Tatsunori Hashimoto, Dan Jurafsky, James Zou, Aylin Caliskan

PDF Cite DOI Thread

Words with Consistent Diachronic Usage Patterns are Learned Earlier: A Computational Analysis Using Temporally Aligned Word Embeddings.

In this study, we use temporally aligned word embeddings and a large diachronic corpus of English to quantify language change in a data‐driven, scalable way, which is grounded in language use.

Giovanni Cassani, Federico Bianchi, Marco Marelli

PDF Cite Code Tweet

Training Temporal Word Embeddings with a Compass

We introduce a novel model for word embedding alignment and test it on temporal word embeddings obtaining SOTA results.

Valerio Di Carlo, Federico Bianchi, Matteo Palmonari

PDF Cite Code Blog Post

Blog Posts

Teaching CLIP Some Fashion

This is a short blog post describing FashionCLIP. If you are a data scientist you probably have to deal with both images and text. However, your data will be very specific to your domain, and standard models might not work well. This post explains how domain-specific vision and language models can be used in a domain-specific setting and why using them can be a promising way to create a search engine or a (zero-shot) classifier.

Federico Bianchi

Mar 11, 2023 1 min read Vision and Language Models

Read on Medium

A Rounded Evaluation of Recommender Systems

This article describes a novel data and code challenge that is currently running: EvalRS. We decided to organize EvalRS with friends from Coveo, Microsoft, and NVIDIA, to better understand evaluation in Recommender Systems.

Federico Bianchi

Aug 11, 2022 1 min read Recommender Systems

Read on Medium

Research Philosophy

I think scientific research should be easy to access and to replicate. You can find most of my papers on ArXiv and on open access platforms - unfortunately, this is not always possible. The code I write is available online and I try to make it easy to use so that everyone can benefit from it.

I also try to actively participate in scientific communication. I have many blog posts online, explaining different research topics to the public and I am also part of the ComIntArt (Comitato Intelligenza Artificiale) where I do videos for the Italian public (see for example, this video on GPT-3 limits).

Recent & Upcoming Talks

Talk: Domain-specific multi-modal ML with CLIP

Building Domain-specific contrastive models for language and image.

Sep 29, 2022 1:00 PM — 3:00 PM Pinecone

Federico Bianchi

Video Follow Paper

Talk: Twitter-Demographer

In this talk we describe our newest package for enriching Twitter data

Jul 19, 2022 1:00 PM Text as Data PoliSci Reading Group

Federico Bianchi

Code Video Follow

Talk: Fantastic CLIPs and How To Tune Them

How to fine-tune Contrastive Language-Image Pre-training models for different usecases

Jul 5, 2022 1:00 PM Stanford University

Federico Bianchi

Follow

Talk: How to Train Your CLIP

How to train a Contrastive Language-Image Pre-training for the Italian Language

Oct 7, 2021 1:00 PM — 3:00 PM Tilburg University

Federico Bianchi

Follow

Talk: Natural Language Processing to Better Understand Language – and Vice Versa

We show how NLP can help us understand language aspects like grounding, but also discuss several cognitively-informed methods that can be directly applied to more practical tasks.

Oct 3, 2021 1:00 PM Edinburgh University

Federico Bianchi

Slides Follow

See all events

Recent Publications

Quickly discover relevant content by filtering publications. You can also look at my Google Scholar profile for an updated list of my publications.

Federico Bianchi, Debora Nozza, Dirk Hovy (2022). Language Invariant Properties in Natural Language Processing. In NLP Power!.

PDF Cite Code

Debora Nozza, Federico Bianchi, Dirk Hovy, Anne Lauscher (2022). Measuring Harmful Sentence Completion in Language Models for LGBTQIA+ Individuals. Second Workshop on Language Technology for Equality, Diversity and Inclusion at ACL 2022.

Cite Code

Federico Bianchi, Debora Nozza, Dirk Hovy (2022). XLM-EMO: Multilingual Emotion Prediction in Social Media Text. 12th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis at EACL 2021.

Cite Code

Debora Nozza, Federico Bianchi, Dirk Hovy (2022). Pipelines for Social Bias Testing of Large Language Models. First Workshop on Challenges & Perspectives in Creating Large Language Models at ACL 2022.

PDF Cite Poster

See all publications