fabriziof (at) campus.technion.ac.il
Fabrizio Frasca
Postdoc · Technion | Incoming Faculty · CISPA Helmholtz Center for Information Security (Jan 2027)
Hiring a PhD student and a postdoc
I am building a research group at CISPA on machine learning for structured data and reliable foundation models.
How to join →Hi! I am Fabrizio. I develop principled machine learning methods for structured data: sets, graphs, sequences, and their combinations. I care about models that respect the symmetries of their data, are provably expressive, and scale. Increasingly, I apply these ideas to the computational traces of foundation models, such as activations and attention maps, which I treat as structured data in their own right. The goal is to make these systems more reliable, for example by detecting hallucinations from a model’s internals.
I am a Postdoctoral Fellow at Technion with Prof. Haggai Maron, and will join CISPA as Faculty in January 2027. Before that, I obtained my PhD at Imperial College London with Prof. Michael Bronstein, working on expressive Graph Neural Networks, and was a Machine Learning Researcher at Twitter Cortex (2019–2023), which I joined through the acquisition of Fabula AI. I also co-founded and co-organise GLOW, a monthly reading group on graph learning.
News
Selected Publications
(* indicates equal contribution, † equal supervision; see the complete list)
Learning Cross-Model Activation Alignments with Explicit Many-to-Many Layer Maps
Alina Sudakov*, Guy Bar-Shalom*, Fabrizio Frasca†, Haggai Maron†
Abstract
LLMs are released at a rapid pace, raising a natural question: how do two independently trained models relate, both in which layers correspond and in how features transform between them? We study this by learning an activation alignment, a map from a source model's layerwise activations to a target's. Our method, MATCHA, factors this map into a layer map, whose output is an explicit target-by-source matrix that can be extracted and inspected, and a layer-shared feature map between hidden spaces. Most of prior work fixes the layer correspondence in advance, pairing layers at roughly the same relative depth; in contrast, we learn both factors jointly from prompts. Across 42 pairs of seven models spanning three different families, MATCHA reconstructs the target's activations more faithfully and improves retrieval-based metrics substantially, w.r.t. previous approaches. The recovered maps are broadly monotone in depth but, in contrast with most previous approaches, are consistently many-to-many: each target layer draws on a band of source layers. Our alignments also enable transfer of activation-space interventions, allowing steering vectors and probes developed for one model to transfer to another.
Neural Message-Passing on Attention Graphs for Hallucination Detection
Fabrizio Frasca*, Guy Bar-Shalom*, Yftah Ziser, Haggai Maron
ICLR 2026
Abstract
Large Language Models (LLMs) often generate incorrect or unsupported content, known as hallucinations. Existing detection methods rely on heuristics or simple models over isolated computational traces such as activations, or attention maps. We unify these signals by representing them as attributed graphs, where tokens are nodes, edges follow attentional flows, and both carry features from attention scores and activations. Our approach, CHARM, casts hallucination detection as a graph learning task and tackles it by applying GNNs over the above attributed graphs. We show that CHARM provably subsumes prior attention-based heuristics and, experimentally, it consistently outperforms other leading approaches across diverse benchmarks. Our results shed light on the relevant role played by the graph structure and on the benefits of combining computational traces, whilst showing CHARM exhibits promising zero-shot performance on cross-dataset transfer.
Lost in Serialization: Invariance and Generalization of LLM Graph Reasoners
Daniel Herbst, Lea Karbevska, Divyanshu Kumar, Akanksha Ahuja, Fatemeh Gholamzadeh Nasrabadi, Fabrizio Frasca
ICML 2026 Workshop on Graph Foundation Models
Abstract
While promising, graph reasoners based on Large Language Models (LLMs) lack built-in invariance to symmetries in graph representations. Operating on sequential graph serializations, LLMs can produce different outputs under node reindexing, edge reordering, or formatting changes, raising robustness concerns. We systemati- cally analyze these effects, studying how fine-tuning impacts encoding sensitivity as well generalization on unseen tasks. We propose a principled decomposition of graph serializations into node labeling, computational structure, and surface encoding, and evaluate LLM robustness to variations of each of these factors on a comprehensive benchmarking suite. We also contribute a novel set of spectral tasks to further assess generalization abilities of fine-tuned reasoners. Results show that larger (non-fine-tuned) models are more robust, and fine-tuning reduces sensitivity to node relabeling but may increase it to variations in structure and format, while it does not consistently improve performance on unseen tasks.
Beyond Token Probes: Hallucination Detection via Activation Tensors with ACT-ViT
Guy Bar-Shalom*, Fabrizio Frasca*, Yaniv Galron, Yftah Ziser, Haggai Maron
NeurIPS 2025
Abstract
Detecting hallucinations in Large Language Model-generated text is crucial for their safe deployment. While probing classifiers show promise, they operate on isolated layer–token pairs and are LLM-specific, limiting their effectiveness and hindering cross-LLM applications. In this paper, we introduce a novel approach to address these shortcomings. We build on the natural sequential structure of activation data in both axes (layers × tokens) and advocate treating full activation tensors akin to images. We design ACT-ViT, a Vision Transformer-inspired model that can be effectively and efficiently applied to activation tensors and supports training on data from multiple LLMs simultaneously. Through comprehensive experiments encompassing diverse LLMs and datasets, we demonstrate that ACT-ViT consistently outperforms traditional probing techniques while remaining extremely efficient for deployment. In particular, we show that our architecture benefits substantially from multi-LLM training, achieves strong zero-shot performance on unseen datasets, and can be transferred effectively to new LLMs through fine-tuning. Full code is available at https://github.com/BarSGuy/ACT-ViT.
Beyond Next Token Probabilities: Learnable, Fast Detection of Hallucinations and Data Contamination on LLM Output Distributions
Guy Bar-Shalom*, Fabrizio Frasca*, Derek Lim, Yoav Gelberg, Yftah Ziser, Ran El-Yaniv, Gal Chechik, Haggai Maron
AAAI 2026
Abstract
The automated detection of hallucinations and training data contamination is pivotal to the safe deployment of Large Language Models (LLMs). These tasks are particularly challenging in settings where no access to model internals is available. Current approaches in this setup typically leverage only the probabilities of actual tokens in the text, relying on simple task-specific heuristics. Crucially, they overlook the information contained in the full sequence of next-token probability distributions. We propose to go beyond hand-crafted decision rules by learning directly from the complete observable output of LLMs -- consisting not only of next-token probabilities, but also the full sequence of next-token distributions. We refer to this as the LLM Output Signature (LOS), and treat it as a reference data type for detecting hallucinations and data contamination. To that end, we introduce LOS-Net, a lightweight attention-based architecture trained on an efficient encoding of the LOS, which can provably approximate a broad class of existing techniques for both tasks. Empirically, LOS-Net achieves superior performance across diverse benchmarks and LLMs, while maintaining extremely low detection latency. Furthermore, it demonstrates promising transfer capabilities across datasets and LLMs.
Balancing Efficiency and Expressiveness: Subgraph GNNs with Walk-Based Centrality
Joshua Southern*, Yam Eitan, Guy Bar-Shalom, Michael Bronstein, Haggai Maron, Fabrizio Frasca*
ICML 2025
Abstract
We propose an expressive and efficient approach that combines the strengths of two prominent extensions of Graph Neural Networks (GNNs): Subgraph GNNs and Structural Encodings (SEs). Our approach leverages walk-based centrality measures, both as a powerful form of SE and also as a subgraph selection strategy for Subgraph GNNs. By drawing a connection to perturbation analysis, we highlight the effectiveness of centrality-based sampling, and show it significantly reduces the computational burden associated with Subgraph GNNs. Further, we combine our efficient Subgraph GNN with SEs derived from the calculated centrality and demonstrate this hybrid approach, dubbed HyMN, gains in discriminative power. HyMN effectively addresses the expressiveness limitations of Message Passing Neural Networks (MPNNs) while mitigating the computational costs of Subgraph GNNs. Through a series of experiments on synthetic and real-world tasks, we show it outperforms other subgraph sampling approaches while being competitive with full-bag Subgraph GNNs and other state-of-the-art approaches with a notably reduced runtime.
Understanding and Extending Subgraph GNNs by Rethinking Their Symmetries
Fabrizio Frasca*, Beatrice Bevilacqua*, Michael M. Bronstein, Haggai Maron
NeurIPS 2022 · Oral (1.7% acceptance rate)
Abstract
Subgraph GNNs are a recent class of expressive Graph Neural Networks (GNNs) which model graphs as collections of subgraphs. So far, the design space of possible Subgraph GNN architectures as well as their basic theoretical properties are still largely unexplored. In this paper, we study the most prominent form of subgraph methods, which employs node-based subgraph selection policies such as ego-networks or node marking and deletion. We address two central questions: (1) What is the upper-bound of the expressive power of these methods? and (2) What is the family of equivariant message passing layers on these sets of subgraphs?. Our first step in answering these questions is a novel symmetry analysis which shows that modelling the symmetries of node-based subgraph collections requires a significantly smaller symmetry group than the one adopted in previous works. This analysis is then used to establish a link between Subgraph GNNs and Invariant Graph Networks (IGNs). We answer the questions above by first bounding the expressive power of subgraph methods by 3-WL, and then proposing a general family of message-passing layers for subgraph methods that generalises all previous node-based Subgraph GNNs. Finally, we design a novel Subgraph GNN dubbed SUN, which theoretically unifies previous architectures while providing better empirical performance on multiple benchmarks.
Equivariant Subgraph Aggregation Networks
Beatrice Bevilacqua*, Fabrizio Frasca*, Derek Lim*, Balasubramaniam Srinivasan, Chen Cai, Gopinath Balamurugan, Michael M. Bronstein, Haggai Maron
ICLR 2022 · Spotlight (5% acceptance rate)
Abstract
Message-passing neural networks (MPNNs) are the leading architecture for deep learning on graph-structured data, in large part due to their simplicity and scalability. Unfortunately, it was shown that these architectures are limited in their expressive power. This paper proposes a novel framework called Equivariant Subgraph Aggregation Networks (ESAN) to address this issue. Our main observation is that while two graphs may not be distinguishable by an MPNN, they often contain distinguishable subgraphs. Thus, we propose to represent each graph as a set of subgraphs derived by some predefined policy, and to process it using a suitable equivariant architecture. We develop novel variants of the 1-dimensional Weisfeiler-Leman (1-WL) test for graph isomorphism, and prove lower bounds on the expressiveness of ESAN in terms of these new WL variants. We further prove that our approach increases the expressive power of both MPNNs and more expressive architectures. Moreover, we provide theoretical results that describe how design choices such as the subgraph selection policy and equivariant neural architecture affect our architecture's expressive power. To deal with the increased computational cost, we propose a subgraph sampling scheme, which can be viewed as a stochastic version of our framework. A comprehensive set of experiments on real and synthetic datasets demonstrates that our framework improves the expressive power and overall performance of popular GNN architectures.
My personal website has been realised through Jekyll and Github Pages. Original theme is by orderedlist, but it has been customised quite a bit.