Steve Azzolin

ELLIS PhD student @ UNITN/UniCambridge || Prev. Visiting Research Student at UniCambridge || Prev. Research intern at SISLab

[Steve Azzolin @steveazzolin]

🚨 New ICLR paper🚨

We found that:
😨Self-Explainable graph models can produce fabricated explanations
😱These can go undetected

Why is this important?

Explanations can conceal malicious behavior: handle them with care!

[Steve Azzolin @steveazzolin]

@looselycorrect Let me read that again

[Steve Azzolin @steveazzolin]

Heading to @imperialcollege to present our latest work on🌟graph explanations that do not explain🌟

Seminar starts at 16 BST
Link: https://t.co/wgnvv3UuD7
Paper: https://t.co/iVPe0LKjVv

[Steve Azzolin @steveazzolin]

@DonatoCrisosto1 It must be, vecio

[Andrea Pugnana @ingenuobayes]

📢The submission link for the fourth edition of the Hybrid Human-Machine Learning and Decision Making (HLDM’26) - co-hosted by @toniwuest, @cc_jalvarez, @andrea_whatever, @annamonreale and @KatyaTentori - is up and running! 📢

Deadline: 7 June 2026

1/2

[Hannah Rose Kirk @hannahrosekirk]

🪧 In a follow-up paper, we extended this work from attitudes to actions.

Using two pre-registered experiments (~15k participants), we showed that conversational AI can persuade people to sign real petitions and donate money to charity.

[Andrea Pugnana @ingenuobayes]

Finally, on Tuesday @LachiVeronica will also present another ongoing project studying whether GNNs are truly calibrated. This is a joint work with @franciferrini_ @AntonioLonga94 @andrea_whatever @barbera_cesare and Manfred Jaeger. :)

https://t.co/1cOmRPLjUn

[Steve Azzolin @steveazzolin]

If you're into calibration, check this out

[Cesare Barbera @barbera_cesare]

Hi all! Today I will be presenting our work on Local Multiclass Calibration at #AISTATS26 !

Come check our poster #107!

Paper link: https://t.co/UDp474pVme

[Steve Azzolin @steveazzolin]

@iclr_conf

[Steve Azzolin @steveazzolin]

🚨 New ICLR paper🚨

We found that:
😨Self-Explainable graph models can produce fabricated explanations
😱These can go undetected

Why is this important?

Explanations can conceal malicious behavior: handle them with care!

[Steve Azzolin @steveazzolin]

To be presented at poster session 4 (P4-#4012), Fri 24, 3:15 pm 🇧🇷.

Full paper: https://t.co/NhM8CKuLPH

w @looselycorrect @andrea_whatever @SagarMa98786413 @brulepri

[Steve Azzolin @steveazzolin]

To address the issue:
👉We introduce a robust detection metric that reliably spots these fake explanations

[Andrea Pugnana @ingenuobayes]

Excited to share that I will co-organize with @toniwuest, @cc_jalvarez, @andrea_whatever, @annamonreale and @KatyaTentori the fourth edition of the Hybrid Human-Machine Learning and Decision Making (HLDM’26) @ECMLPKDD in Naples. 👇

1/2

[Learning on Graphs Conference 2026 @LogConference]

📢 LoG 2026 Organizer Recruitment

We are recruiting new organizers to help support the Learning on Graphs 2026 conference and community.

Apply here:
🔗 https://t.co/hKlywZRPUH
🗓 Deadline: March 16

Please share with anyone in the graph learning community who may be interested!

[Gianni Franchi @ ICLR @GianniFranchi10]

My PhD student & I are studying if XAI heatmaps really help humans understand what AI "sees" in image classification. It takes max 10 min, fully anonymous, no login. Help us improve explainable AI — just a few clicks! https://t.co/MHRywPSnny Thanks! #XAI #AI #Computervision

[Learning on Graphs Conference 2026 @LogConference]

The conference may be over, but the LOG community never slows down 🧑‍💼🌍

Join us at upcoming meetups worldwide:

🇳🇴 Tromsø 🇮🇳 Gandhinagar 🇮🇹 Pisa 🇫🇷 Paris 🇧🇷 São arlos 🇮🇳 New Delhi

[Steve Azzolin @steveazzolin]

@barbera_cesare felicitaciones, bien hecho

[Steve Azzolin @steveazzolin]

I’m not crying, I just got a concept in my eye

[Stefano Teso ✈️ AAAI26 @looselycorrect]

Good news everyone! This year we will be organizing a workshop on Unifying Concept Representation Learning at ICLR'26!

The workshop is about unifying ideas and techniques from #NeSy AI, #XAI and #Causal representation learning. Have a look at the CfP!

[Learning on Graphs Conference 2026 @LogConference]

Join us at LoG 2025 (Dec 10–12)!

🗓️Keynotes on graph learning & geometry from Aditi Krishnapriyan (Day1⃣), Haggai Maron (Day2⃣), Michael Galkin & Melanie Weber (Day3⃣) — from molecular ML to neural artifacts & graph foundation models.

Register now: https://t.co/xwPMjSkOGI

[Learning on Graphs Conference 2026 @LogConference]

📢 Now that we have an in-person LoG, why are we still supporting meetups?

👉 Meetups keep the community alive all year round, beyond the main event
👉 They give the community more chances to connect locally and inclusively
👉 They complement—not compete with—the in-person LoG