Sparse Weight Decomposition for Efficient Circuit Extraction

• URL: http://arxiv.org/abs/2608.03913v1
• Date: Tue, 04 Aug 2026 16:40:48 GMT
• Title: Sparse Weight Decomposition for Efficient Circuit Extraction
• Abstract summary: Dense pretrained transformers do not naturally expose interpretable units for circuit extraction.<n>We propose Sparse Weight Decomposition, which factorizes pretrained projections by factorizing each weight matrix into two sparse factors.<n>We show that SWD remains effective for full-model replacement of all attention and weight matrices after fine-tuning the nonzero factor values.
• Score: 36.87892448834478
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Dense pretrained transformers do not naturally expose interpretable units for circuit extraction. Existing approaches obtain such units by learning auxiliary sparse representations or training sparse models, incurring substantial additional computation while potentially introducing a fidelity gap between the representation being analyzed and the original pretrained model. We propose Sparse Weight Decomposition (SWD), which reparameterizes pretrained linear projections by factorizing each weight matrix into two sparse factors whose shared intermediate coordinates serve as individually addressable circuit units. Without training a separate replacement network, this parametric representation supports the same scoring, selection, and ablation circuit extraction workflow used for methods that learn sparse features. Across single-matrix replacements, SWD matches the held-out fidelity achieved by Transcoder and other strong baselines while using less than 1% of the data that those baselines use to train their replacements. For matched replacement fidelity, SWD reaches the same circuit sufficiency and necessity targets with fewer active read/write edges and selected units across tasks on GPT-2, Qwen2.5, and Qwen3.5-27B. We further show that SWD remains effective for full-model replacement of all attention and MLP weight matrices after fine-tuning the nonzero factor values. Finally, SWD also features a zero-data variant, allowing broader use of mechanistic interpretability analysis (e.g., per-step analysis).

Related papers

• Equation Recast for Canonical Operator Learning Across Parametric PDEs [59.44573033543284]
  We introduce equation recast, which reformulates parametric operator learning as the learning of a single canonical operator.<n>In high-fidelity tokamak simulations for nuclear fusion, the framework unifies electron-temperature data across four device geometries.
  arXiv  Detail & Related papers  (2026-09-02T13:15:50Z)
• One-Sided Quantile Coupling for Flow Matching [8.295726579539016]
  Flow Matching trains continuous-time generative models by regressing the velocity field of a probability path between a simple source distribution and a target data distribution.<n>We propose Quantile Coupling Flow Matching (QC-FM), a lightweight one-sided coupling that samples only the data batch and constructs each paired source directly.<n>QC-FM improves over the Baseline under matched training budgets, reducing FID by up to 12.9%, and outperforms OT-CFM on all four datasets.
  arXiv  Detail & Related papers  (2026-08-02T04:03:59Z)
• Action-Inspired Generative Models [0.0]
  Action-Inspired Generative Models (AGMs)<n>We introduce a dual-network generative framework motivated by the observation that existing bridge-matching methods assign uniform regression weight to every transition in the transport landscape.<n>We demonstrate that selectively penalising uninformative transport paths through the learned potential yields consistent improvements in generation quality across fidelity and coverage metrics.
  arXiv  Detail & Related papers  (2026-05-14T09:43:32Z)
• In-Context Symbolic Regression for Robustness-Improved Kolmogorov-Arnold Networks [3.7163623763519733]
  Symbolic regression aims to replace blackbox predictors with analytical expressions that can be inspected and validated in machine learning.<n> Kolmogorov-Arnold Pursuit Networks (KANs) are well suited to this goal because each connection between adjacent units (an "edge") ismetd by a univariate function.<n>In practice, however, symbolic extraction is a bottleneck: the standard KAN-to-symbol approach fits operators to each learned edge function in fine isolation.
  arXiv  Detail & Related papers  (2026-03-16T13:21:26Z)
• WUSH: Near-Optimal Adaptive Transforms for LLM Quantization [52.77441224845925]
  Quantization to low bitwidth is a standard approach for deploying large language models.<n>A few extreme weights and activations stretch the dynamic range and reduce the effective resolution of the quantizer.<n>We derive, for the first time, closed-form optimal linear blockwise transforms for joint weight-activation quantization.
  arXiv  Detail & Related papers  (2025-11-30T16:17:34Z)
• MLPMoE: Zero-Shot Architectural Metamorphosis of Dense LLM MLPs into Static Mixture-of-Experts [0.0]
  Large Language Models (LLMs) are predominantly deployed as dense transformers, where every parameter in every feed-forward block is activated for every token.<n>Recent upcycling methods such as MoEfication, CMoE, ToMoE, and MoORE reveal that much of the useful computation lives in sparse, semi-modular substructures inside dense feed-forward networks.<n>This paper introducesMoE (MLP-Experts), a training-free transformation that restructures the dense in transformer blocks into a static, high-cardinality mixture of experts.
  arXiv  Detail & Related papers  (2025-11-26T06:14:26Z)
• Incorporating Arbitrary Matrix Group Equivariance into KANs [69.30866522377694]
  Kolmogorov-Arnold Networks (KANs) have seen great success in scientific domains.<n>We propose Equivariant Kolmogorov-Arnold Networks (EKAN) to broaden their applicability to more fields.
  arXiv  Detail & Related papers  (2024-10-01T06:34:58Z)
• Improving the Sample-Complexity of Deep Classification Networks with Invariant Integration [77.99182201815763]
  Leveraging prior knowledge on intraclass variance due to transformations is a powerful method to improve the sample complexity of deep neural networks. We propose a novel monomial selection algorithm based on pruning methods to allow an application to more complex problems. We demonstrate the improved sample complexity on the Rotated-MNIST, SVHN and CIFAR-10 datasets.
  arXiv  Detail & Related papers  (2022-02-08T16:16:11Z)
• Unfolding Projection-free SDP Relaxation of Binary Graph Classifier via GDPA Linearization [59.87663954467815]
  Algorithm unfolding creates an interpretable and parsimonious neural network architecture by implementing each iteration of a model-based algorithm as a neural layer. In this paper, leveraging a recent linear algebraic theorem called Gershgorin disc perfect alignment (GDPA), we unroll a projection-free algorithm for semi-definite programming relaxation (SDR) of a binary graph. Experimental results show that our unrolled network outperformed pure model-based graph classifiers, and achieved comparable performance to pure data-driven networks but using far fewer parameters.
  arXiv  Detail & Related papers  (2021-09-10T07:01:15Z)
• Exact Backpropagation in Binary Weighted Networks with Group Weight Transformations [0.0]
  Quantization based model compression serves as high performing and fast approach for inference. Models that constrain the weights to binary values enable efficient implementation of the ubiquitous dot product.
  arXiv  Detail & Related papers  (2021-07-03T10:29:34Z)
• Bayesian Transformer Language Models for Speech Recognition [59.235405107295655]
  State-of-the-art neural language models (LMs) represented by Transformers are highly complex. This paper proposes a full Bayesian learning framework for Transformer LM estimation.
  arXiv  Detail & Related papers  (2021-02-09T10:55:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.