Graph Circuit Learning is introduced, a supervised, amortized framework that trains a GNN across multiple model--task pairs and applies it to unseen cases and preliminary results suggest that graph machine learning offers a natural and potentially powerful perspective on circuit localization.
Abstract
Circuit localization is a mechanistic interpretability task whose goal is to identify a sparse subgraph of a transformer's computation graph sufficient to reproduce a particular behavior. Most established methods localize circuits independently for each model--task pair. We instead frame circuit localization as a graph machine learning problem in which the edges of a computation graph represent computational pathways, and graph neural networks (GNNs) model interactions among these pathways. We introduce Graph Circuit Learning (GCL), a supervised, amortized framework that trains a GNN across multiple model--task pairs and applies it to unseen cases. To provide sufficient data, we augment the InterpBench benchmark with additional cases derived from the TracrBench programs. Of the 14 evaluated GCL configurations, the highest scored a median edge AUROC of $0.902$ (interquartile interval $[0.861, 0.942]$) on the 16 original held-out InterpBench cases. This is close to the published InterpBench median of $0.910$ for EAP-IG while remaining below ACDC's $0.959$. Removing all message-passing edges reduces the median to $0.825$. We also adapt PGExplainer, a GNN explainability method, to circuit localization, obtaining a median edge AUROC of $0.858$ on the same cases. These preliminary results suggest that graph machine learning offers a natural and potentially powerful perspective on circuit localization, and we hope this perspective encourages closer exchange between the two communities.
A variety of approaches have been proposed for subgraph-level representation learning. However, these approaches have primarily been developed and evaluated under node-induced subgraph settings, where each subgraph is defined by a selected set of nodes. In contrast, subgraph prediction tasks in which subgraphs are induced by selected sets of edges remain largely unexplored, despite arising naturally in domains such as knowledge graph reasoning, scene graph understanding, and functional connectivity analysis in network neuroscience. Edge-induced subgraph prediction introduces two technical requirements beyond those of the node-induced setting: (1) sensitivity to subgraph-internal edge structure and (2) isolation of subgraph-specific information within a mini-batch. To address these requirements, we introduce the segregated graph, a construction that represents the internal structure of each subgraph via subgraph-specific copies of base-graph nodes connected only by the edges selected for that subgraph. We perform message passing in parallel on the segregated graph and the base graph, and fuse the resulting representations at each layer through identity-based mixing, thereby combining internal-structure awareness with boundary and global contextual information. Experiments on three benchmarks derived from DocRED, Visual Genome, and the Human Connectome demonstrate that our method consistently outperforms existing subgraph prediction approaches, confirming the effectiveness of jointly modeling the segregated graph and the base graph.
S. Baek, Hogun Park· Proceedings of the 32nd ACM...· 0 citations
Topological neural networks (TNNs) enable leveraging high-order structures on graphs (e.g., cycles and cliques) to boost the expressive power of message-passing neural networks. In turn, however, these structures are typically identified a priori through an unsupervised graph lifting operation. Notwithstanding, this choice is crucial and may have a drastic impact on a TNN's performance on downstream tasks. To circumvent this issue, we propose $\partial$lift (DiffLift), a general framework for learning graph liftings to hypergraphs and cellular- and simplicial complexes in an end-to-end fashion. In particular, our approach leverages learned vertex-level latent representations to identify and parameterize distributions over candidate higher-order cells for inclusion. This results in a scalable model which can be readily integrated into any TNN. Our experiments show that $\partial$lift outperforms existing lifting methods on multiple benchmarks for graph and node classification across different TNN architectures. Notably, our approach leads to gains of up to 45% over static liftings, including both connectivity- and feature-based ones.
J. L. Franco, Gabriel Duarte, Alexander Nikitin et al.· 2 citations
We study the topology of learned representations in predictive coding networks (PCNs), a neuro-inspired bidirectional architecture, using a quantitative layer-wise persistent homology analysis. We train well-performing PCNs on a synthetic classification dataset ($\geq 99.9\%$ test accuracy) and on MNIST ($\geq 95\%$ test accuracy), and measure how topological features change across layers for different architectures and activation functions. We find that smaller PCNs collapse connected components across layers earlier than larger models (Spearman $\unicode{x1D70C} \in [0.72, 0.79]$ across activations), with model size measured as the sum of hidden-layer widths. We also observe a strong negative correlation ($\unicode{x1D70C} = -0.58$) between the depth at which simplification occurs and reconstruction error; i.e., architectures that simplify later reconstruct better. Finally, a seed-level bootstrap comparison across architectures and activations shows that PCNs consistently collapse connected components later than matched MLPs, with an average difference of $3.6$ layers. These results suggest that persistent homology offers a useful quantitative lens on the compression--reconstruction tradeoff in PCNs, and that both model capacity and the recurrent, bidirectional dynamics of predictive coding inference shape when this tradeoff is resolved across layers.
One approach to mechanistic interpretability explains behavior through circuits: the components and connections that carry it. Frozen discovery often returns hundreds of edges, making them hard to inspect, compare, or verify exhaustively. We introduce Circuit Condensation, which post-trains models to concentrate behaviors into smaller causal graphs. Each round prunes low-attribution edges and trains a low-rank adapter to match the original through what remains, retaining the cut only if task performance and general capability survive. Across four behaviors and eight models, condensed circuits are smaller than the strongest frozen baseline in 30 of 32 settings, by $8.1\times$ on average and up to $316\times$. Repeating the search without weight updates produces larger circuits in 29 of 32 settings, showing that weight updates, rather than search alone, drive the reduction. Testing every subset of 19 circuits finds 11 that cannot be reduced and reveals removable edges in the rest. Pair ablations expose dependencies between edges, showing that their effects cannot be understood independently. On indirect object identification, condensation isolates 24 heads, 17 of them with documented roles, against 61 heads and 36 undocumented ones for the matched frozen circuit: a sufficient sub-circuit of the published mechanism rather than a reconstruction of it. The resulting circuit tracks the original model's next-token distribution and predicts its errors.
Finding a representative description of graph entities that captures their structural roles and homophily is a challenging goal for graph embedding techniques due to the non-Euclidean nature of graphs. Traditionally, Graph embeddings achieve top performance via random-walk methods and graph neural networks. However, these methods are transductive and utilize an expensive global optimization via softmax or a dense representation trained in an end-to-end pipeline with gradient descent. Nonetheless, other variants of GNNs can map to unseen nodes; they still rely on iterative message passing and backpropagation, incurring high computational and memory costs. Conversely, the mammalian cortex solves structurally similar problems by learning to map its input stream of patterns into a compact representation for downstream regions. We present the biologically inspired Columnar-Embedder architecture for learning binary Sparse Distributed Representations (SDRs) of graph nodes. The learning is driven by a local Bienenstock-Cooper-Munro (BCM) Hebbian rule modulated by positive pointwise mutual information (PPMI) computed from online streams of random walks. Continuous learning from streaming random-walk pairs without labels, backpropagation, or supervision enables the architecture to exhibit natural resistance to catastrophic forgetting. Across five graph benchmarks, the performance of SDRs is competitive with that of real-valued dense embeddings on node classification and link prediction, while the architecture exhibits portability, resilience to noise, and robustness to data corruption.
Sheaf Neural Networks (SNNs) generalize message passing by replacing scalar edge weights of standard Graph Neural Networks (GNNs) with learnable, edge-dependent restriction maps between node stalks. Despite their strong theoretical foundations and promising transductive results, SNNs have been evaluated almost exclusively on transductive node classification, leaving their behaviour under inductive protocols unknown. We address this gap through the first systematic benchmark of the sheaf design space, evaluating three diffusion mechanisms (neural sheaf diffusion, sheaf attention, and sheaf attention with Graph Attention Network v2), three restriction-map parameterizations, three stalk dimensions, and six modern GNN architectural components, within a message-passing reformulation that never assembles the heavy sheaf Laplacian, making the full design space trainable under cross-graph batching. Across $1{,}890$ controlled experiments on 14 inductive datasets, multiple insights emerge: restriction maps are the dominant design choice and general maps are preferable, larger stalks add capacity but not long-range reach, architectural components explain more performance variation than the entire sheaf-specific design space itself. Under a matched protocol, SNNs transfer to inductive settings but do not reach the strongest baselines, with gaps being dataset-dependent. Practically, a single sheaf configuration can generalize across datasets, so effort is better spent tuning the surrounding architectural recipe than the sheaf operator itself.
Stefano Fiorini, Edoardo Coppola, Pietro Liò· 0 citations