Structural Instability of Feature Composition

• URL: http://arxiv.org/abs/2605.05223v1
• Date: Sat, 18 Apr 2026 05:53:31 GMT
• Title: Structural Instability of Feature Composition
• Abstract summary: We study the compositional-collapse threshold under a spherical dictionary model.<n>We show that, in the high-bias regime, ReLU converts microscopic correlation-induced variance fluctuations into a systematic drift.<n>We validate the predicted scaling trends on structured semantic features extracted from CLEVR.
• Score: 2.5754366051855837
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Sparse Autoencoders (SAEs) have emerged as a powerful paradigm for disentangling feature superposition in transformer-based architectures, enabling precise control via activation steering. However, the theoretical foundations of compositional steering -- the simultaneous activation of distinct semantic latents -- remain under-explored. The prevailing Linear Representation Hypothesis often abstracts away non-linear interference effects that arise in overcomplete dictionaries. We present a geometric framework for analyzing the instability of feature unions. Modeling the activation space as a high-dimensional sparse cone manifold, we derive an asymptotic compositional-collapse threshold under a spherical dictionary model, characterized by the Gaussian mean width (statistical dimension) of the signal cone. We further show that, in the high-bias regime, ReLU rectification converts microscopic correlation-induced variance fluctuations into a systematic drift that accumulates under composition, yielding interference growth consistent with a ratchet effect. We validate the predicted scaling trends on structured semantic features extracted from CLEVR, where hierarchical correlations accelerate the transition relative to random baselines. Together, our results highlight geometric constraints on the scalability of union-based steering and motivate composition mechanisms that explicitly manage interference beyond naive linear superposition.

Related papers

• Prototype-Anchored Generalized Manifold Regression for Unknown-Domain Object Detection [58.25418970608328]
  We study Single-Domain Generalized Object Detection (Single-DGOD), which aims to transfer a detector trained on a single source domain to multiple unseen domains.<n>Existing methods mainly rely on simulation-driven strategies, such as data augmentation or textual prompts, to enlarge the training distribution.<n>We propose Manifold Regression with Visual-Text Dual Chain-of-Thought (MR-DCoT), which formulates unknown-domain generalization as a manifold regression problem.
  arXiv  Detail & Related papers  (2026-07-08T09:27:15Z)
• Effective Covariance Dynamics in Solvable High-Dimensional GANs [2.639628765667673]
  We study a solvable high-dimensional model of generative adversarial network (GAN) training in which a linear generator learns a low-dimensional subspace from data.<n>We prove that the microscopic training process converges, in the high-dimensional limit, to deterministic ordinary differential equations governed by this effective covariance.
  arXiv  Detail & Related papers  (2026-06-25T16:33:54Z)
• Unified Geometry-Guided ML-FTLE for Tracking Transient Chaos from Scalar Time Series [0.0]
  We propose a geometry-guided machine learning framework that unifies predictive trajectory divergence with macroscopic attractor morphology to track abrupt regime shifts.<n>This equation-free framework provides a highly accurate, noise-resilient diagnostic for monitoring structural transitions in complex non-stationary systems.
  arXiv  Detail & Related papers  (2026-06-05T15:22:57Z)
• Hierarchical Consistency Learning for Test-time Adaptation in Camouflage Perception [50.278200968044665]
  Camouflaged object detection (COD) aims to localize targets that exhibit minimal perceptual differences from backgrounds through physical attributes.<n>Existing methods, constrained by the static train-then-freeze paradigm, suffer from domain rigidity and dependency annotation.<n>We propose the hierarchical consistency learning framework, which integrates test-time adaptation for dynamic representation recalibration.
  arXiv  Detail & Related papers  (2026-05-25T09:57:46Z)
• Beyond Point-Wise Matching: Structural Representation Alignment for Accelerating Diffusion Transformers [93.3976834364707]
  We propose sREPA, a structural REPresentation Alignment framework to enforce consistency in the relational geometry of feature maps.<n>By encouraging the model to internalize holistic spatial layouts and structural correlations from pre-trained features, sREPA achieves faster and more stable convergence.
  arXiv  Detail & Related papers  (2026-05-16T12:01:04Z)
• Hyperbolic Enhanced Representation Learning for Incomplete Multi-view Clustering [57.38215918201251]
  We propose HERL, a Hyperbolic Enhanced Representation Learning framework for Incomplete Multi-View Clustering.<n> operating within the Poincaré ball, HERL constructs a structure-aware latent space to enhance representation learning.<n>We show that HERL consistently outperforms state-of-the-art approaches.
  arXiv  Detail & Related papers  (2026-04-18T10:50:46Z)
• Non-monotonic causal discovery with Kolmogorov-Arnold Fuzzy Cognitive Maps [0.40611352512781856]
  Fuzzy Cognitive Maps constitute a neuro-symbolic paradigm for modeling complex dynamic systems.<n>This research proposes the Kolmogorov-Arnold Fuzzy Cognitive Map (KA-FCM), a novel architecture that redefines the causal transmission mechanism.
  arXiv  Detail & Related papers  (2026-04-06T19:56:12Z)
• Certified Quantum Schrödinger Control via Hierarchical Tucker Models [2.033424698590539]
  This paper develops a framework for sampled-data feedback control implemented with fixed-rank HT projections.<n>By viewing each truncation as a bounded, rank-dependent of the nominal closed loop, we show that the HT-projected dynamics are practically exponentially stable.
  arXiv  Detail & Related papers  (2026-03-20T17:46:50Z)
• OrthoFormer: Instrumental Variable Estimation in Transformer Hidden States via Neural Control Functions [0.0]
  Transformer architectures excel at sequential modeling yet remain fundamentally limited by correlational learning.<n>We propose OrthoFormer, a causally grounded architecture that embeds instrumental variable estimation directly into Transformer blocks via neural control functions.
  arXiv  Detail & Related papers  (2026-03-08T03:05:16Z)
• MirrorLA: Reflecting Feature Map for Vision Linear Attention [49.41670925034762]
  Linear attention significantly reduces the computational complexity of Transformers from quadratic to linear, yet it consistently lags behind softmax-based attention in performance.<n>We propose MirrorLA, a geometric framework that substitutes passive truncation with active reorientation.<n>MirrorLA achieves state-of-the-art performance across standard benchmarks, demonstrating that strictly linear efficiency can be achieved without compromising representational fidelity.
  arXiv  Detail & Related papers  (2026-02-04T09:14:09Z)
• FlexCausal: Flexible Causal Disentanglement via Structural Flow Priors and Manifold-Aware Interventions [1.7114074082429929]
  Causal Disentangled Representation Learning aims to learn and disentangle low dimensional representations from observations.<n>We propose FlexCausal, a novel CDRL framework based on a block-diagonal covariance VAE.<n>Our framework ensures a precise structural correspondence between the learned latent subspaces and the ground-truth causal relations.
  arXiv  Detail & Related papers  (2026-01-29T11:30:53Z)
• SIGMA: Scalable Spectral Insights for LLM Collapse [51.863164847253366]
  We introduce SIGMA (Spectral Inequalities for Gram Matrix Analysis), a unified framework for model collapse.<n>By utilizing benchmarks that deriving and deterministic bounds on the matrix's spectrum, SIGMA provides a mathematically grounded metric to track the contraction of the representation space.<n>We demonstrate that SIGMA effectively captures the transition towards states, offering both theoretical insights into the mechanics of collapse.
  arXiv  Detail & Related papers  (2026-01-06T19:47:11Z)
• Random-Matrix-Induced Simplicity Bias in Over-parameterized Variational Quantum Circuits [72.0643009153473]
  We show that expressive variational ansatze enter a Haar-like universality class in which both observable expectation values and parameter gradients concentrate exponentially with system size.<n>As a consequence, the hypothesis class induced by such circuits collapses with high probability to a narrow family of near-constant functions.<n>We further show that this collapse is not unavoidable: tensor-structured VQCs, including tensor-network-based and tensor-hypernetwork parameterizations, lie outside the Haar-like universality class.
  arXiv  Detail & Related papers  (2026-01-05T08:04:33Z)
• On the Joint Minimization of Regularization Loss Functions in Deep Variational Bayesian Methods for Attribute-Controlled Symbolic Music Generation [47.38557855930304]
  Explicit latent variable models provide a flexible yet powerful framework for data synthesis.<n>We show that existing approaches struggle to jointly minimize both regularization objectives.<n>We show that suitable attribute transformations can help achieve both controllability and regularization of the target latent dimensions.
  arXiv  Detail & Related papers  (2025-11-10T14:09:25Z)
• Drift No More? Context Equilibria in Multi-Turn LLM Interactions [58.69551510148673]
  contexts drift is the gradual divergence of a model's outputs from goal-consistent behavior across turns.<n>Unlike single-turn errors, drift unfolds temporally and is poorly captured by static evaluation metrics.<n>We show that multi-turn drift can be understood as a controllable equilibrium phenomenon rather than as inevitable decay.
  arXiv  Detail & Related papers  (2025-10-09T04:48:49Z)

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.