Federico Barbero

Petar Veličković, Federico Barbero, Christos Perivolaropoulos et al. · deepmind

Perplexity -- a function measuring a model's overall level of "surprise" when encountering a particular output -- has gained significant traction in recent years, both as a loss function and as a simple-to-compute metric of model quality. Prior studies have pointed out several limitations of perplexity, often from an empirical manner. Here we leverage recent results on Transformer continuity to show in a rigorous manner how perplexity may be an unsuitable metric for model selection. Specifically, we prove that, if there is any sequence that a compact decoder-only Transformer model predicts accurately and confidently -- a necessary pre-requisite for strong generalisation -- it must imply existence of another sequence with very low perplexity, but not predicted correctly by that same model. Further, by analytically studying iso-perplexity plots, we find that perplexity will not always select for the more accurate model -- rather, any increase in model confidence must be accompanied by a commensurate rise in accuracy for the new model to be selected.

Federico Barbero, Cristian Bodnar, Haitz Sáez de Ocáriz Borde et al. · cambridge

A Sheaf Neural Network (SNN) is a type of Graph Neural Network (GNN) that operates on a sheaf, an object that equips a graph with vector spaces over its nodes and edges and linear maps between these spaces. SNNs have been shown to have useful theoretical properties that help tackle issues arising from heterophily and over-smoothing. One complication intrinsic to these models is finding a good sheaf for the task to be solved. Previous works proposed two diametrically opposed approaches: manually constructing the sheaf based on domain knowledge and learning the sheaf end-to-end using gradient-based methods. However, domain knowledge is often insufficient, while learning a sheaf could lead to overfitting and significant computational overhead. In this work, we propose a novel way of computing sheaves drawing inspiration from Riemannian geometry: we leverage the manifold assumption to compute manifold-and-graph-aware orthogonal maps, which optimally align the tangent spaces of neighbouring data points. We show that this approach achieves promising results with less computational overhead when compared to previous SNN models. Overall, this work provides an interesting connection between algebraic topology and differential geometry, and we hope that it will spark future research in this direction.

Francesco Di Giovanni, Lorenzo Giusti, Federico Barbero et al.

Message Passing Neural Networks (MPNNs) are instances of Graph Neural Networks that leverage the graph to send messages over the edges. This inductive bias leads to a phenomenon known as over-squashing, where a node feature is insensitive to information contained at distant nodes. Despite recent methods introduced to mitigate this issue, an understanding of the causes for over-squashing and of possible solutions are lacking. In this theoretical work, we prove that: (i) Neural network width can mitigate over-squashing, but at the cost of making the whole network more sensitive; (ii) Conversely, depth cannot help mitigate over-squashing: increasing the number of layers leads to over-squashing being dominated by vanishing gradients; (iii) The graph topology plays the greatest role, since over-squashing occurs between nodes at high commute (access) time. Our analysis provides a unified framework to study different recent methods introduced to cope with over-squashing and serves as a justification for a class of methods that fall under graph rewiring.

Dharshan Kumaran, Arthur Conmy, Federico Barbero et al.

Verbal confidence -- prompting LLMs to state their confidence as a number or category -- is widely used to extract uncertainty estimates from black-box models. However, how LLMs internally generate such scores remains unknown. We address two questions: first, when confidence is computed - just-in-time when requested, or automatically during answer generation and cached for later retrieval; and second, what verbal confidence represents - token log-probabilities, or a richer evaluation of answer quality? Focusing on Gemma 3 27B and Qwen 2.5 7B, we provide convergent evidence for cached retrieval. Activation steering, patching, noising, and swap experiments reveal that confidence representations emerge at answer-adjacent positions before appearing at the verbalization site. Attention blocking pinpoints the information flow: confidence is gathered from answer tokens, cached at the first post-answer position, then retrieved for output. Critically, linear probing and variance partitioning reveal that these cached representations explain substantial variance in verbal confidence beyond token log-probabilities, suggesting a richer answer-quality evaluation rather than a simple fluency readout. These findings demonstrate that verbal confidence reflects automatic, sophisticated self-evaluation -- not post-hoc reconstruction -- with implications for understanding metacognition in LLMs and improving calibration.

Federico Barbero, Ameya Velingker, Amin Saberi et al.

Graph Neural Networks (GNNs) are popular models for machine learning on graphs that typically follow the message-passing paradigm, whereby the feature of a node is updated recursively upon aggregating information over its neighbors. While exchanging messages over the input graph endows GNNs with a strong inductive bias, it can also make GNNs susceptible to over-squashing, thereby preventing them from capturing long-range interactions in the given graph. To rectify this issue, graph rewiring techniques have been proposed as a means of improving information flow by altering the graph connectivity. In this work, we identify three desiderata for graph-rewiring: (i) reduce over-squashing, (ii) respect the locality of the graph, and (iii) preserve the sparsity of the graph. We highlight fundamental trade-offs that occur between spatial and spectral rewiring techniques; while the former often satisfy (i) and (ii) but not (iii), the latter generally satisfy (i) and (iii) at the expense of (ii). We propose a novel rewiring framework that satisfies all of (i)--(iii) through a locality-aware sequence of rewiring operations. We then discuss a specific instance of such rewiring framework and validate its effectiveness on several real-world benchmarks, showing that it either matches or significantly outperforms existing rewiring approaches.

Haitz Sáez de Ocáriz Borde, Anees Kazi, Federico Barbero et al.

Graph Neural Networks usually rely on the assumption that the graph topology is available to the network as well as optimal for the downstream task. Latent graph inference allows models to dynamically learn the intrinsic graph structure of problems where the connectivity patterns of data may not be directly accessible. In this work, we generalize the discrete Differentiable Graph Module (dDGM) for latent graph learning. The original dDGM architecture used the Euclidean plane to encode latent features based on which the latent graphs were generated. By incorporating Riemannian geometry into the model and generating more complex embedding spaces, we can improve the performance of the latent graph inference system. In particular, we propose a computationally tractable approach to produce product manifolds of constant curvature model spaces that can encode latent features of varying structure. The latent representations mapped onto the inferred product manifold are used to compute richer similarity measures that are leveraged by the latent graph learning model to obtain optimized latent graphs. Moreover, the curvature of the product manifold is learned during training alongside the rest of the network parameters and based on the downstream task, rather than it being a static embedding space. Our novel approach is tested on a wide range of datasets, and outperforms the original dDGM model.

Haitz Sáez de Ocáriz Borde, Federico Barbero

We demonstrate the applicability of model-agnostic algorithms for meta-learning, specifically Reptile, to GNN models in molecular regression tasks. Using meta-learning we are able to learn new chemical prediction tasks with only a few model updates, as compared to using randomly initialized GNNs which require learning each regression task from scratch. We experimentally show that GNN layer expressivity is correlated to improved meta-learning. Additionally, we also experiment with GNN emsembles which yield best performance and rapid convergence for k-shot learning.

Federico Barbero, Feargus Pendlebury, Fabio Pierazzi et al.

Machine learning for malware classification shows encouraging results, but real deployments suffer from performance degradation as malware authors adapt their techniques to evade detection. This phenomenon, known as concept drift, occurs as new malware examples evolve and become less and less like the original training examples. One promising method to cope with concept drift is classification with rejection in which examples that are likely to be misclassified are instead quarantined until they can be expertly analyzed. We propose TRANSCENDENT, a rejection framework built on Transcend, a recently proposed strategy based on conformal prediction theory. In particular, we provide a formal treatment of Transcend, enabling us to refine conformal evaluation theory -- its underlying statistical engine -- and gain a better understanding of the theoretical reasons for its effectiveness. In the process, we develop two additional conformal evaluators that match or surpass the performance of the original while significantly decreasing the computational overhead. We evaluate TRANSCENDENT on a malware dataset spanning 5 years that removes sources of experimental bias present in the original evaluation. TRANSCENDENT outperforms state-of-the-art approaches while generalizing across different malware domains and classifiers. To further assist practitioners, we determine the optimal operational settings for a TRANSCENDENT deployment and show how it can be applied to many popular learning algorithms. These insights support both old and new empirical findings, making Transcend a sound and practical solution for the first time. To this end, we release TRANSCENDENT as open source, to aid the adoption of rejection strategies by the security community.

Federico Barbero, Álvaro Arroyo, Xiangming Gu et al. · deepmind

Large Language Models (LLMs) tend to attend heavily to the first token in the sequence -- creating a so-called attention sink. Many works have studied this phenomenon in detail, proposing various ways to either leverage or alleviate it. Attention sinks have been connected to quantisation difficulties, security issues, and streaming attention. Yet, while many works have provided conditions in which they occur or not, a critical question remains shallowly answered: Why do LLMs learn such patterns and how are they being used? In this work, we argue theoretically and empirically that this mechanism provides a method for LLMs to avoid over-mixing, connecting this to existing lines of work that study mathematically how information propagates in Transformers. We conduct experiments to validate our theoretical intuitions and show how choices such as context length, depth, and data packing influence the sink behaviour. We hope that this study provides a new practical perspective on why attention sinks are useful in LLMs, leading to a better understanding of the attention patterns that form during training.

Álvaro Arroyo, Alessio Gravina, Benjamin Gutteridge et al.

Graph Neural Networks (GNNs) are models that leverage the graph structure to transmit information between nodes, typically through the message-passing operation. While widely successful, this approach is well known to suffer from the over-smoothing and over-squashing phenomena, which result in representational collapse as the number of layers increases and insensitivity to the information contained at distant and poorly connected nodes, respectively. In this paper, we present a unified view of these problems through the lens of vanishing gradients, using ideas from linear control theory for our analysis. We propose an interpretation of GNNs as recurrent models and empirically demonstrate that a simple state-space formulation of a GNN effectively alleviates over-smoothing and over-squashing at no extra trainable parameter cost. Further, we show theoretically and empirically that (i) GNNs are by design prone to extreme gradient vanishing even after a few layers; (ii) Over-smoothing is directly related to the mechanism causing vanishing gradients; (iii) Over-squashing is most easily alleviated by a combination of graph rewiring and vanishing gradient mitigation. We believe our work will help bridge the gap between the recurrent and graph neural network literature and will unlock the design of new deep and performant GNNs.

Jacob Bamberger, Federico Barbero, Xiaowen Dong et al.

The dominant paradigm for learning on graph-structured data is message passing. Despite being a strong inductive bias, the local message passing mechanism suffers from pathological issues such as over-smoothing, over-squashing, and limited node-level expressivity. To address these limitations we propose Bundle Neural Networks (BuNN), a new type of GNN that operates via message diffusion over flat vector bundles - structures analogous to connections on Riemannian manifolds that augment the graph by assigning to each node a vector space and an orthogonal map. A BuNN layer evolves the features according to a diffusion-type partial differential equation. When discretized, BuNNs are a special case of Sheaf Neural Networks (SNNs), a recently proposed MPNN capable of mitigating over-smoothing. The continuous nature of message diffusion enables BuNNs to operate on larger scales of the graph and, therefore, to mitigate over-squashing. Finally, we prove that BuNN can approximate any feature transformation over nodes on any (potentially infinite) family of graphs given injective positional encodings, resulting in universal node-level expressivity. We support our theory via synthetic experiments and showcase the strong empirical performance of BuNNs over a range of real-world tasks, achieving state-of-the-art results on several standard benchmarks in transductive and inductive settings.

Benedict Aaron Tjandra, Federico Barbero, Michael Bronstein

Despite the successful application of Temporal Graph Networks (TGNs) for tasks such as dynamic node classification and link prediction, they still perform poorly on the task of dynamic node affinity prediction -- where the goal is to predict 'how much' two nodes will interact in the future. In fact, simple heuristic approaches such as persistent forecasts and moving averages over ground-truth labels significantly and consistently outperform TGNs. Building on this observation, we find that computing heuristics over messages is an equally competitive approach, outperforming TGN and all current temporal graph (TG) models on dynamic node affinity prediction. In this paper, we prove that no formulation of TGN can represent persistent forecasting or moving averages over messages, and propose to enhance the expressivity of TGNs by adding source-target identification to each interaction event message. We show that this modification is required to represent persistent forecasting, moving averages, and the broader class of autoregressive models over messages. Our proposed method, TGNv2, significantly outperforms TGN and all current TG models on all Temporal Graph Benchmark (TGB) dynamic node affinity prediction datasets.

Itay Yona, Ilia Shumailov, Jamie Hayes et al. · deepmind

Large Language Models (LLMs), despite their impressive capabilities, often fail to accurately repeat a single word when prompted to, and instead output unrelated text. This unexplained failure mode represents a vulnerability, allowing even end-users to diverge models away from their intended behavior. We aim to explain the causes for this phenomenon and link it to the concept of ``attention sinks'', an emergent LLM behavior crucial for fluency, in which the initial token receives disproportionately high attention scores. Our investigation identifies the neural circuit responsible for attention sinks and shows how long repetitions disrupt this circuit. We extend this finding to other non-repeating sequences that exhibit similar circuit disruptions. To address this, we propose a targeted patch that effectively resolves the issue without negatively impacting the model's overall performance. This study provides a mechanistic explanation for an LLM vulnerability, demonstrating how interpretability can diagnose and address issues, and offering insights that pave the way for more secure and reliable models.

Xidong Feng, Vivek Veeriah, Marcus Chiam et al.

While Generative AI rapidly advances in various domains, generating truly creative, aesthetic, and counter-intuitive outputs remains a challenge. This paper presents an approach to tackle these difficulties in the domain of chess puzzles. We start by benchmarking Generative AI architectures, and then introduce an RL framework with novel rewards based on chess engine search statistics to overcome some of those shortcomings. The rewards are designed to enhance a puzzle's uniqueness, counter-intuitiveness, diversity, and realism. Our RL approach dramatically increases counter-intuitive puzzle generation by 10x, from 0.22\% (supervised) to 2.5\%, surpassing existing dataset rates (2.1\%) and the best Lichess-trained model (0.4\%). Our puzzles meet novelty and diversity benchmarks, retain aesthetic themes, and are rated by human experts as more creative, enjoyable, and counter-intuitive than composed book puzzles, even approaching classic compositions. Our final outcome is a curated booklet of these AI-generated puzzles, which is acknowledged for creativity by three world-renowned experts.

Vivek Veeriah, Federico Barbero, Marcus Chiam et al.

The rapid advancement of Generative AI has raised significant questions regarding its ability to produce creative and novel outputs. Our recent work investigates this question within the domain of chess puzzles and presents an AI system designed to generate puzzles characterized by aesthetic appeal, novelty, counter-intuitive and unique solutions. We briefly discuss our method below and refer the reader to the technical paper for more details. To assess our system's creativity, we presented a curated booklet of AI-generated puzzles to three world-renowned experts: International Master for chess compositions Amatzia Avni, Grandmaster Jonathan Levitt, and Grandmaster Matthew Sadler. All three are noted authors on chess aesthetics and the evolving role of computers in the game. They were asked to select their favorites and explain what made them appealing, considering qualities such as their creativity, level of challenge, or aesthetic design.

Federico Barbero, Xiangming Gu, Christopher A. Choquette-Choo et al.

In this work, we show that it is possible to extract significant amounts of alignment training data from a post-trained model -- useful to steer the model to improve certain capabilities such as long-context reasoning, safety, instruction following, and maths. While the majority of related work on memorisation has focused on measuring success of training data extraction through string matching, we argue that embedding models are better suited for our specific goals. Distances measured through a high quality embedding model can identify semantic similarities between strings that a different metric such as edit distance will struggle to capture. In fact, in our investigation, approximate string matching would have severely undercounted (by a conservative estimate of $10\times$) the amount of data that can be extracted due to trivial artifacts that deflate the metric. Interestingly, we find that models readily regurgitate training data that was used in post-training phases such as SFT or RL. We show that this data can be then used to train a base model, recovering a meaningful amount of the original performance. We believe our work exposes a possibly overlooked risk towards extracting alignment data. Finally, our work opens up an interesting discussion on the downstream effects of distillation practices: since models seem to be regurgitating aspects of their training set, distillation can therefore be thought of as indirectly training on the model's original dataset.

Enrique Queipo-de-Llano, Álvaro Arroyo, Federico Barbero et al.

Attention sinks and compression valleys have attracted significant attention as two puzzling phenomena in large language models, but have been studied in isolation. In this work, we present a surprising connection between attention sinks and compression valleys, tracing both to the formation of massive activations in the residual stream. We prove theoretically that massive activations necessarily produce representational compression and establish bounds on the resulting entropy reduction. Through experiments across several models (410M-120B parameters), we confirm that when the beginning-of-sequence token develops extreme activation norms in the middle layers, both compression valleys and attention sinks emerge simultaneously. Targeted ablation studies validate our theoretical predictions. This unified view motivates us to propose the Mix-Compress-Refine theory of information flow, as an attempt to explain how LLMs organize their computation in depth by controlling attention and representational compression via massive activations. Specifically, we posit that Transformer-based LLMs process tokens in three distinct phases: (1) broad mixing in the early layers, (2) compressed computation with limited mixing in the middle layers, and (3) selective refinement in the late layers. Our framework helps explain why embedding tasks perform best at intermediate layers, whereas generation tasks benefit from full-depth processing, clarifying differences in task-dependent representations.

Isaac Reid, Arijit Sehanobish, Cederik Höfs et al.

We introduce WIRE: Wavelet-Induced Rotary Encodings. WIRE extends Rotary Position Encodings (RoPE), a popular algorithm in LLMs and ViTs, to graph-structured data. We demonstrate that WIRE is more general than RoPE, recovering the latter in the special case of grid graphs. WIRE also enjoys a host of desirable theoretical properties, including equivariance under node ordering permutation, compatibility with linear attention, and (under select assumptions) asymptotic dependence on graph resistive distance. We test WIRE on a range of synthetic and real-world tasks, including identifying monochromatic subgraphs, semantic segmentation of point clouds, and more standard graph benchmarks. We find it to be effective in settings where the underlying graph structure is important.

Federico Barbero, Andrea Banino, Steven Kapturowski et al.

We study how information propagates in decoder-only Transformers, which are the architectural backbone of most existing frontier large language models (LLMs). We rely on a theoretical signal propagation analysis -- specifically, we analyse the representations of the last token in the final layer of the Transformer, as this is the representation used for next-token prediction. Our analysis reveals a representational collapse phenomenon: we prove that certain distinct sequences of inputs to the Transformer can yield arbitrarily close representations in the final token. This effect is exacerbated by the low-precision floating-point formats frequently used in modern LLMs. As a result, the model is provably unable to respond to these sequences in different ways -- leading to errors in, e.g., tasks involving counting or copying. Further, we show that decoder-only Transformer language models can lose sensitivity to specific tokens in the input, which relates to the well-known phenomenon of over-squashing in graph neural networks. We provide empirical evidence supporting our claims on contemporary LLMs. Our theory also points to simple solutions towards ameliorating these issues.