Mechanistic Interpretability for Neural Networks: Circuits, Sparse Features and Symbolic Reasoning

• URL: http://arxiv.org/abs/2607.07316v1
• Date: Wed, 08 Jul 2026 12:04:38 GMT
• Title: Mechanistic Interpretability for Neural Networks: Circuits, Sparse Features and Symbolic Reasoning
• Abstract summary: mechanistic interpretability is an emerging field that seeks to reverse-engineer the internal algorithms of modern neural networks.<n>The paper provides a detailed examination of Transformer circuit analysis, exploring how internal components drive complex tasks and in-context learning.<n>It then tackles the core challenge of superposition and polysemanticity, demonstrating how tools like Sparse Autoencoders (SAEs) and transcoders can decompose tangled network activations into distinct, human-interpretable features.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: This article offers a comprehensive overview of mechanistic interpretability, an emerging field that seeks to reverse-engineer the internal algorithms of modern neural networks. While traditional explainable AI methods often stop at surface-level input-output correlations, this approach directly addresses the opaque "black box" nature of machine learning models, which is essential for ensuring safety and auditability in high-stakes deployments. The paper provides a detailed examination of Transformer circuit analysis, exploring how internal components like the residual stream, attention mechanisms, and induction heads drive complex tasks and in-context learning. It subsequently tackles the core challenge of superposition and polysemanticity, demonstrating how tools like Sparse Autoencoders (SAEs) and transcoders can decompose tangled network activations into distinct, human-interpretable features. Furthermore, the paper explores methods for actively controlling and modifying model behavior through steering vectors and causal interventions. Finally, it connects these mechanistic insights with neurosymbolic AI frameworks designed to translate neural representations into explicit, executable logical rules.

Related papers

• Transcoders for Investigating Deception in Language Models [0.0]
  We investigate the use of transcoders to analyse deceptive behaviour in language models.<n>We identify a dictionary of deception-related features that exert a stronger influence on deceptive outputs.
  arXiv  Detail & Related papers  (2026-07-16T10:10:10Z)
• Seeing Through Circuits: Faithful Mechanistic Interpretability for Vision Transformers [54.2202951543048]
  We investigate whether useful mechanistic circuits can be identified through computational graphs in vision transformers.<n>We propose an effective method for Automatic Visual Circuit Discovery (Vi-CD) that recovers class-specific circuits for classification.<n>We find that insightful and actionable edge-based circuits can be recovered from vision transformers, adding transparency to the internal computations of these models.
  arXiv  Detail & Related papers  (2026-04-15T23:25:46Z)
• Mechanistic Interpretability for Large Language Model Alignment: Progress, Challenges, and Future Directions [16.821238326410324]
  Large language models (LLMs) have achieved remarkable capabilities across diverse tasks, yet their internal decision-making processes remain largely opaque.<n>Mechanistic interpretability has emerged as a critical research direction for understanding and aligning these models.<n>We analyze how interpretability insights have informed alignment strategies including reinforcement learning from human feedback, constitutional AI, and scalable oversight.
  arXiv  Detail & Related papers  (2026-01-21T11:43:57Z)
• Circuit Insights: Towards Interpretability Beyond Activations [20.178085579725472]
  We propose WeightLens and CircuitLens, two complementary methods for mechanistic interpretability.<n>WeightLens interprets features directly from their learned weights, removing the need for explainer models or datasets.<n> CircuitLens captures how feature activations arise from interactions between components, revealing circuit-level dynamics.
  arXiv  Detail & Related papers  (2025-10-16T17:49:41Z)
• Unlocking Out-of-Distribution Generalization in Transformers via Recursive Latent Space Reasoning [50.99796659680724]
  This work investigates out-of-distribution (OOD) generalization in Transformer networks using a GSM8K-style modular arithmetic on computational graphs task as a testbed.<n>We introduce and explore a set of four architectural mechanisms aimed at enhancing OOD generalization.<n>We complement these empirical results with a detailed mechanistic interpretability analysis that reveals how these mechanisms give rise to robust OOD generalization abilities.
  arXiv  Detail & Related papers  (2025-10-15T21:03:59Z)
• A General Framework for Property-Driven Machine Learning [1.0485739694839669]
  We show that adversarial training can be used to improve robustness to small perturbations within $epsilon$-cubes.<n> domains other than computer vision may require more flexible input region specifications via generalised hyper-rectangles.<n>In this paper, we investigate how these two complementary approaches can be unified within a single framework for property-driven machine learning.
  arXiv  Detail & Related papers  (2025-05-01T11:33:38Z)
• Meta-Representational Predictive Coding: Biomimetic Self-Supervised Learning [51.22185316175418]
  We present a new form of predictive coding that we call meta-representational predictive coding (MPC)<n>MPC sidesteps the need for learning a generative model of sensory input by learning to predict representations of sensory input across parallel streams.
  arXiv  Detail & Related papers  (2025-03-22T22:13:14Z)
• Exploring Latent Pathways: Enhancing the Interpretability of Autonomous Driving with a Variational Autoencoder [79.70947339175572]
  A bio-inspired neural circuit policy model has emerged as an innovative control module. We take a leap forward by integrating a variational autoencoder with the neural circuit policy controller. In addition to the architectural shift toward a variational autoencoder, this study introduces the automatic latent perturbation tool.
  arXiv  Detail & Related papers  (2024-04-02T09:05:47Z)
• Mechanistic Neural Networks for Scientific Machine Learning [58.99592521721158]
  We present Mechanistic Neural Networks, a neural network design for machine learning applications in the sciences. It incorporates a new Mechanistic Block in standard architectures to explicitly learn governing differential equations as representations. Central to our approach is a novel Relaxed Linear Programming solver (NeuRLP) inspired by a technique that reduces solving linear ODEs to solving linear programs.
  arXiv  Detail & Related papers  (2024-02-20T15:23:24Z)
• Tensor Networks for Explainable Machine Learning in Cybersecurity [0.0]
  We develop an unsupervised clustering algorithm based on Matrix Product States (MPS)<n>Our investigation proves that MPS rival traditional deep learning models such as autoencoders and GANs in terms of performance.<n>Our approach naturally facilitates the extraction of feature-wise probabilities, Von Neumann Entropy, and mutual information.
  arXiv  Detail & Related papers  (2023-12-29T22:35:45Z)
• Data-driven emergence of convolutional structure in neural networks [83.4920717252233]
  We show how fully-connected neural networks solving a discrimination task can learn a convolutional structure directly from their inputs. By carefully designing data models, we show that the emergence of this pattern is triggered by the non-Gaussian, higher-order local structure of the inputs.
  arXiv  Detail & Related papers  (2022-02-01T17:11:13Z)

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.