Sign in

Andrea Pugnana

@andrepugni.bsky.social
54 followers 90 following 21 posts

Research Fellow @ University of Trento. Studying ML models that know what they do not know. He/Him.

PostsRepliesMedia
Andrea Pugnana @andrepugni.bsky.social · 18/09/2025
For other believers, link to the old arxiv, arxiv.org/abs/2503.161.... updated version to be out soon.
arxiv.org
Deferring Concept Bottleneck Models: Learning to Defer Interventions to Inaccurate Experts
Concept Bottleneck Models (CBMs) are machine learning models that improve interpretability by grounding their predictions on human-understandable concepts, allowing for targeted interventions in their...
000 0
Andrea Pugnana @andrepugni.bsky.social · 18/09/2025
Deferring Concept Bottleneck Models go to @neuripsconf.bsky.social 2025! Thanks to all my amazing coauthors (especially @rmassidda.it who is the only one on 🦋 as far as I know) who believed in the marriage between CBMs and Learning to Defer! See u in San Diego! 😄🌊🏖️
131 0
Reposted by Andrea Pugnana
Emanuele Marconato @ema-ridopoco.bsky.social · 03/05/2025
Now in Thailand to present our paper at #AISTATS2025 🇹🇭🌴 📍Today at 3:00-6:00 pm, poster number 118! More details here: openreview.net/forum?id=XCm...
openreview.net
All or None: Identifiable Linear Properties of Next-Token...
We analyze identifiability as a possible explanation for the ubiquity of linear properties across language models, such as the vector difference between the representations of “easy” and “easiest”...
031 0
Andrea Pugnana @andrepugni.bsky.social · 29/04/2025
👀 Want to dive deeper? Curious how ML + humans really perform together? 📍Come find us at our poster: Hall A-E 160, May 3rd. 5/5
000 0
Andrea Pugnana @andrepugni.bsky.social · 29/04/2025
🔍 We define wo scenarios: 1️⃣ Both human and ML predictions for deferred cases? We can estimate individual & aggregate causal effects. 2️⃣ Only human predictions for deferred cases? We can use Regression Discontinuity Design for local causal effects. 4/5
100 0
Andrea Pugnana @andrepugni.bsky.social · 29/04/2025
In our work, we bring causal inference to the arena. Using the potential outcomes framework, we ask: What’s the actual impact of deferring to humans on system performance? 🎯 3/5
110 0
Andrea Pugnana @andrepugni.bsky.social · 29/04/2025
🧠🤖 Deferring systems let ML models pass the mic to humans when they are better at predicting. But what is the actual impact on predictive performance of such systems? Surprisingly, such a question is still overlooked. 2/5
100 0
Andrea Pugnana @andrepugni.bsky.social · 29/04/2025
On my way to Mai Khao🏝️ for #AISTATS2025, where I will present A Causal Framework for Evaluating Deferring Systems, a joint work with Filippo Palomba, Jose Alvarez and Salvatore Ruggieri. 1/5
121 0
Andrea Pugnana @andrepugni.bsky.social · 29/04/2025
👀 Want to dive deeper? Curious how ML + humans really perform together? 📍Come find us at our poster: Hall A-E 160, May 3rd. 5/5
000 0
Andrea Pugnana @andrepugni.bsky.social · 29/04/2025
🔍 We define wo scenarios: 1️⃣ Both human and ML predictions for deferred cases? We can estimate individual & aggregate causal effects. 2️⃣ Only human predictions for deferred cases? We can use Regression Discontinuity Design for local causal effects. 4/5
100 0
Andrea Pugnana @andrepugni.bsky.social · 29/04/2025
In our work, we bring causal inference to the arena. Using the potential outcomes framework, we ask: What’s the actual impact of deferring to humans on system performance? 🎯 3/5
110 0
Andrea Pugnana @andrepugni.bsky.social · 29/04/2025
🧠🤖 Deferring systems let ML models pass the mic to humans when they are better at predicting. But what is the actual impact on predictive performance of such systems? Surprisingly, such a question is still overlooked. 2/5
100 0
Reposted by Andrea Pugnana
Steve Azzolin @steveazzolin.bsky.social · 25/04/2025
Happening tomorrow! Poster number 508 Saturday's session 10-12:30
001 0
Reposted by Andrea Pugnana
Steve Azzolin @steveazzolin.bsky.social · 17/04/2025
Faithfulness of GNN explanations isn’t one-size-fits-all🧢 Our last @iclr-conf.bsky.social paper breaks it down across: 1. Evaluation metrics 2. Model implementations 3. OOD generalisation w: Antonio L. @looselycorrect.bsky.social @andreapasserini.bsky.social 1/5
163 0
Reposted by Andrea Pugnana
Causal Club @handle.invalid · 10/04/2025
If you are into Structural Causal Models, you're going to love Deep SCMs! ❤️ On the next (and last for now!) talk, Audrey Poinsot (INRIA) will review DSCMs, their applications and open challenges. 🤌 See you tomorrow April 11th at 16:00 (CET) at meet.google.com/tng-aibh-yuz 🗓️ tinyurl.com/deepscm
032 0
Andrea Pugnana @andrepugni.bsky.social · 05/04/2025
❗️Deadline extended 18 April❗️ Working on Human-AI interaction and trustworthy AI? Consider submitting your ongoing work to our workshop tcai2025.isti.cnr.it @hhaiconference.bsky.social Co-hosted by Apoorva Singh, Bruno Lepri, @burcusayin.bsky.social and Roberto Pellungrini in Pisa. #AI #HHAI
tcai2025.isti.cnr.it
TCAI2025
TCAI2025
022 0
Andrea Pugnana @andrepugni.bsky.social · 27/03/2025
Great talk tomorrow at the @causalclub.di.unipi.it by Nina Corvelo Benz on counterfactual token generation.
000 0
Reposted by Andrea Pugnana
Causal Club @handle.invalid · 20/03/2025
Time for the next Causal Club speaker! ⏳ Mouad El Bouchattaoui (Paris-Saclay University) will present a novel method for conterfactual regression over time (openreview.net/pdf?id=bKOZY...). 🤌 See you in Pisa on March 21 at 16:00 (CET) or live at meet.google.com/tng-aibh-yuz
042 0
Andrea Pugnana @andrepugni.bsky.social · 12/03/2025
Also quite curious to see what new technical issue will appear on Friday.
010 0
Andrea Pugnana @andrepugni.bsky.social · 12/03/2025
Happy to share our second episode with @giovannidetoni.bsky.social ready to rock at @causalclub.di.unipi.it on Friday at 16:00.
000 0
Reposted by Andrea Pugnana
Causal Club @handle.invalid · 07/03/2025
⏰ Don't miss Mátyás talk today at 3PM! 🎥 See you online meet.google.com/cqt-ufji-xfz 🤌 ...or live in the CS Department of the University of Pisa!
073 0
Andrea Pugnana @andrepugni.bsky.social · 28/02/2025
The Causal Club is back for Season 3! Do not miss it!
000 0
Andrea Pugnana @andrepugni.bsky.social · 25/02/2025
Are you in Philly for @realaaai.bsky.social 2025, and still do not know what oral to attend? Do not worry. Also, ML models do not know many things. Check what they know they don't know with Salvatore Ruggieri, who will present our joint work in the senior track morning session on Friday!
042 0
Andrea Pugnana @andrepugni.bsky.social · 22/01/2025
Finally, we thank the reviewers for their valuable suggestions and will add the link to the complete version as soon as possible. See you in Mai Khao! 4/4
000 0
Andrea Pugnana @andrepugni.bsky.social · 22/01/2025
Moreover, our experiments showcase how one can reason about these causal effects. We believe that shifting the paradigm from pure performance to the causal effect of introducing a human-AI team is a necessary building block to make AI truly human-centred and aligned. 3/4
100 0
Andrea Pugnana @andrepugni.bsky.social · 22/01/2025
The paper shows how to retrieve the causal effect on predictive performance by introducing a deferring system in a decision-making pipeline and framing two scenarios. We then show which causal effect can be retrieved under the potential outcome framework for each scenario. 2/4
100 0
Andrea Pugnana @andrepugni.bsky.social · 22/01/2025
I am thrilled to share that our work, "A Causal Framework for Evaluating Deferring Systems," has been accepted at #AISTATS2025! I am very attached to this work (joint with F. Palomba, J. Alvarez, and S. Ruggieri—who are not on 🦋) as it bridges causal inference with deferring systems. 1/4
130 0
Reposted by Andrea Pugnana
Luca Pappalardo @lucapappalardo.bsky.social · 12/12/2024
The KDD Lab Workshop 2024 just started in San Miniato, Italy. Two days for sharing the last achievements and new horizons in research...
042 0