Fabrizio Frasca
@ffabffrasca
Followers
2K
Following
2K
Media
36
Statuses
517
Postdoctoral Fellow @TechnionLive — Geometric Deep Learning in some of its various forms — PhD @imperialcollege — Previously @twitter, @fabula_ai and @polimi
Joined June 2019
“Balancing Efficiency and Expressiveness: Subgraph GNNs with Walk-Based Centrality” is at #ICML2025!. Drop by our poster on Tuesday 11:00–13:30, East Exhib. Hall A-B # E-3005. Joint work w/ @JoshSouthern13 @ytn_ym @GuyBarSh @mmbronstein @HaggaiMaron . 🧵
2
13
51
6/. @JoshSouthern13 and I be at #ICML2025, poster session Tuesday — stop by and chat if you're around!. I would also be happy to meet up and chat about graphs, (graphs and) LLMs, and how to detect their hallucinations 😳. Feel free to reach out!.
0
1
4
5/. We seamlessly scaled HyMN to graphs with thousands of nodes, far beyond what full-bag Subgraph GNNs can handle. HyMN is competitive with — and can outperform — sparse Graph Transformers at a fraction of their practical runtime.
1
0
1
4/. Maximising these lets the model operate closest to the expressive (but costly) full-bag regime. On top of marking, we use Structural Encodings derived from centrality computations for provably larger discriminative power. 👉 * Hybrid* Marking Network = marking + encodings.
1
0
1
3/. We mark only those nodes with highest walk-based centrality, like the Katz Index or Subgraph Centrality. We show these node importance measures recapitulate how much marking perturbs graph representations, connecting to works on GNN stability.
1
0
1
2/. We propose HyMN, a scalable and expressive graph learning model. Like Subgraph GNNs, HyMN runs message passing on node-marked variants of the input graph — but it only marks a small constant number of nodes, making it way more efficient.
1
0
1
RT @mciccone_AI: Heading to #ICML2025 in Vancouver 🇨🇦 July 13–19!. 🚀Ping me if you want to discuss research and collabs on how to build bet….
0
6
0
🌟 Tune in for GLOW's summer session this Wednesday!. 🗓️ July 2nd, 5pm CEST on Zoom!. 🌴@Rythian47 will present Tropical Attention for algo reasoning (. 🧪 @jeremy_wayland will tell us about the intricacies of benchmarking GNNs (
0
4
12
RT @chrsmrrs: Nice read!. Also, have a look at our two ICML position papers:. (Practice) and .
arxiv.org
Machine learning on graphs, especially using graph neural networks (GNNs), has seen a surge in interest due to the wide availability of graph data across a broad spectrum of disciplines, from life...
0
6
0
In the post, we write and elaborate on what emerged from one of our open discussion sessions. We touched upon where the community stands and where it may be headed next. Hope you enjoy the read 🤗. This is a joint effort by the GLOW organisers with inputs from many researchers!.
Check out this blogpost from @ffabffrasca and the GLOW reading group on the future of graph learning!. I’ve also contributed and my main take is - its actually working and its an exciting moment to work on applications!
0
4
24
Resources:.- On Incorporating Scale into Graph Networks (.- On the Expressive Power of Sparse Geometric MPNNs (.
0
0
2
GLOW is coming back this Wednesday! 🌟 . We will hear from – and interact with – @ChristianKoke (incorporating scale in GNNs) and @YSbrdlwb (sparse geometric MPNNs and their expressive power). 🗓️When .May 28th, 5pm CEST on Zoom. 🌐 Details & sign-up:
1
8
24
Excited our paper got accepted and ICML 2025 — and looking forward to discussing this position with other researchers in July. ☺️. We are at a point where these kinds of interactions are extremely relevant in our community! Hope our contribution will foster them even more.
📣 Our spicy ICML 2025 position paper: “Graph Learning Will Lose Relevance Due To Poor Benchmarks”. Graph learning is less trendy in the ML world than it was in 2020-2022. We believe the problem is in poor benchmarks that hold the field back - and suggest ways to fix it!.🧵1/10
1
6
53
Come and see our new work at the “QUESTION” workshop today at #ICLR2025 !. We tackle Data Contamination and Hallucination Detection in LLMs with a learnable approach on their (structured) output signatures ☺️.
📢 Introducing:. Learning on LLM Output Signatures for Gray-box LLM Behavior Analysis [.A joint work with @ffabffrasca (co-first author) and our amazing collaborators:. @dereklim_lzh @yoav_gelberg @YftahZ @el_yaniv @GalChechik @HaggaiMaron .🧵Thread.
0
1
11
Make sure to tune in next week!. and stay up-to-date by subscribing to our mailing list:
0
0
8
GLOW is returning on 𝗠𝗮𝗿𝗰𝗵 𝟮𝟲𝘁𝗵, 𝟱𝗽𝗺 𝗖𝗘𝗧 with a special guest: @PetarV_93 🌟. He will lecture on LLMs as GNNs – a topic which received quite some attention at our last session. Specifically, we will learn how Graph ML tools can help understand LLM generalisation👓.
3
8
49
Apply if you are working on Geometric Deep Learning!. LOGML is an amazing summer school. Attended twice as a student — got exposed to many new ideas and had the chance to connect with outstanding researchers. This year I will be joining as a mentor — hope to see you in London ☺️.
🌟Applications open- LOGML 2025🌟. 👥Mentor-led projects, expert talks, tutorials, socials, and a networking night.✍️Application form: 📅Apply by 6th April 2025 .✉️Questions? logml.committee@gmail.com . #MachineLearning #SummerSchool #LOGML #Geometry.
0
6
35
. Stay tuned by checking out our website ( and by registering to our mailing list ( [3/3].
sites.google.com
GLOW is a new reading group designed to foster discussions on the foundations and latest developments in Graph Machine Learning.
0
0
5
No slides, no formal talks — just a casual conversation with PhDs, postdocs, and all graph enthusiasts. Come share your thoughts — whether you're excited, skeptical, or just curious! [2/3].
1
0
3
GLOW is back 𝐭𝐡𝐢𝐬 Wednesday, 🗓️ Februrary 19th, 5pm CET 🌟. Before our returning to our regular format in March, we’re hosting an open discussion on where Graph Learning is headed and where early-career researchers can make the most impact. [1/3].
1
8
21