Steered Generation via Gradient-Based Optimization on Sparse Query Features

• URL: http://arxiv.org/abs/2605.23040v1
• Date: Thu, 21 May 2026 21:13:14 GMT
• Title: Steered Generation via Gradient-Based Optimization on Sparse Query Features
• Abstract summary: Prototype-Based Sparse Steering is a framework that applies Sparse Autoencoders specifically to query activations.<n>We demonstrate that sparse query representations provide the necessary disentanglement for unified, interpretable control over both logical planning and stylistic nuance.
• Score: 1.587869707099313
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Latent steering exploits internal representations of Large Language Models (LLMs) to guide generation, yet interventions on dense states can entangle distinct semantic features. In this paper, we investigate attention query activations as a high-fidelity site for precise control, hypothesizing that manipulating the attention mechanism itself offers sharper steerability than general state interventions. We introduce Prototype-Based Sparse Steering, a framework that applies Sparse Autoencoders (SAEs) specifically to query activations, to decompose them into interpretable features, then apply gradient-based optimization during inference to align the sparse representation with class prototypes of target behaviors. To validate this architectural insight, we first analyze the mechanism in Textualized Gridworld, a controlled environment for verifiable planning constraints. We demonstrate that optimizing sparse query features enables effective navigation of rigid planning requirements (i.e., safe vs. short paths), confirming the method's ability to satisfy objective rules. We then demonstrate the framework's versatility by training SAEs on a high-dimensional educational domain, where the framework steers the cognitive complexity of feedback (i.e., Bloom's Taxonomy). Our experiments establish that sparse query representations provide the necessary disentanglement for unified, interpretable control over both logical planning and stylistic nuance.

Related papers

• SAGE-Nav: Leveraging LLM Planning and Alignment Fusion for Hierarchical Scene Graph-Guided Navigation [26.212246536363363]
  Object-Goal Navigation (Nav) requires embodied agents to autonomously locate specified targets using only egocentric visual observations.<n>We propose SAGE-Nav, a novel hierarchical framework that integrates the reasoning capabilities of Large Language Models (LLMs) with dynamic scene graphs.<n>We show that SAGE-Nav achieves substantial gains in navigation efficiency and zero-shot generalization while maintaining the low control latency required for physical robotic deployment.
  arXiv  Detail & Related papers  (2026-06-24T07:24:39Z)
• The Art of Interrogation: Consistency Amplifies Factuality in Spatial Reasoning [65.67266333751569]
  Current Large Reasoning Models (LRMs) exhibit remarkable general capabilities but significantly underperform in spatial reasoning tasks.<n>We propose a self-supervised reinforcement learning framework that targets the internal reasoning process without requiring ground-truth annotations.<n>We show that this label-free consistency training approaches the accuracy of models trained with ground-truth supervision and achieves similar generalization across diverse tasks and data domains.
  arXiv  Detail & Related papers  (2026-06-10T10:50:06Z)
• SAGE: A Service Agent Graph-guided Evaluation Benchmark [27.342044311161654]
  We propose SAGE (Service Agent Graph-guided Evaluation), a universal benchmark for automated, dual-axis assessment.<n>SAGE formalizes unstructured SOPs into Dynamic Dialogue Graphs, enabling precise verification of logical compliance.<n>We also observe Empathy Resilience'', a phenomenon where models maintain polite conversational facades despite underlying logical failures.
  arXiv  Detail & Related papers  (2026-04-10T12:55:23Z)
• Structured prototype regularization for synthetic-to-real driving scene parsing [7.087302867255982]
  Driving scene parsing is critical for autonomous vehicles to operate reliably in real-world traffic environments.<n>To reduce the reliance on costly pixel-level annotations, synthetic datasets with automatically generated labels have become a popular alternative.<n>Models trained on synthetic data often perform poorly when applied to real-world scenes due to the synthetic-to-real domain gap.
  arXiv  Detail & Related papers  (2026-03-17T03:05:28Z)
• CLAIRE: Compressed Latent Autoencoder for Industrial Representation and Evaluation -- A Deep Learning Framework for Smart Manufacturing [51.56484100374058]
  CLAIRE is a hybrid end-to-end learning framework that integrates unsupervised deep representation learning with supervised classification for intelligent quality control in smart manufacturing systems.<n>It employs an optimized deep autoencoder to transform raw input into a compact latent space, effectively capturing the intrinsic data structure while suppressing irrelevant or noisy features.<n>The proposed framework highlights the potential of integrating explainable AI with feature-aware regularization for robust fault detection.
  arXiv  Detail & Related papers  (2026-03-06T15:11:58Z)
• StepVAR: Structure-Texture Guided Pruning for Visual Autoregressive Models [98.72926158261937]
  We propose a training-free token pruning framework for Visual AutoRegressive models.<n>We employ a lightweight high-pass filter to capture local texture details, while leveraging Principal Component Analysis (PCA) to preserve global structural information.<n>To maintain valid next-scale prediction under sparse tokens, we introduce a nearest neighbor feature propagation strategy.
  arXiv  Detail & Related papers  (2026-03-02T11:35:05Z)
• OOWM: Structuring Embodied Reasoning and Planning via Object-Oriented Programmatic World Modeling [62.3458279176813]
  We propose Object-Oriented World Modeling (OOWM), a novel framework that structures embodied reasoning through the lens of software engineering formalism.<n>OOWM employs Class Diagrams to ground visual perception into rigorous object hierarchies, and Activity Diagrams to operationalize planning into executable control flows.
  arXiv  Detail & Related papers  (2026-02-25T16:01:07Z)
• Agentic Spatio-Temporal Grounding via Collaborative Reasoning [80.83158605034465]
  Temporal Video Grounding aims to retrieve thetemporal tube of a target object or person in a video given a text query.<n>We propose the Agentic Spatio-Temporal Grounder (ASTG) framework for the task of STVG towards an open-world and training-free scenario.<n>Specifically, two specialized agents SRA (Spatial Reasoning Agent) and TRA (Temporal Reasoning Agent) constructed leveraging on modern Multimoal Large Language Models (MLLMs)<n>Experiments on popular benchmarks demonstrate the superiority of the proposed approach where it outperforms existing weakly-supervised and zero-shot approaches by a margin
  arXiv  Detail & Related papers  (2026-02-10T10:16:27Z)
• Mechanistic Knobs in LLMs: Retrieving and Steering High-Order Semantic Features via Sparse Autoencoders [8.188989044347595]
  We propose a Sparse Autoencoder-based framework for retrieving and steering semantically interpretable internal features.<n>Using the Big Five personality traits as a case study, we demonstrate that our method enables precise, bidirectional steering of model behavior.
  arXiv  Detail & Related papers  (2026-01-06T12:40:37Z)
• GLOW: Graph-Language Co-Reasoning for Agentic Workflow Performance Prediction [51.83437071408662]
  We propose GLOW, a unified framework for AW performance prediction.<n>GLOW combines the graph-structure modeling capabilities of GNNs with the reasoning power of LLMs.<n>Experiments on FLORA-Bench show that GLOW outperforms state-of-the-art baselines in prediction accuracy and ranking utility.
  arXiv  Detail & Related papers  (2025-12-11T13:30:46Z)
• InVDriver: Intra-Instance Aware Vectorized Query-Based Autonomous Driving Transformer [12.441180142943328]
  InVDriver is a novel vectorized query-based system for intra-instance spatial dependencies.<n>We show that InVDriver achieves state-of-the-art performance, surpassing prior methods in both accuracy and safety.
  arXiv  Detail & Related papers  (2025-02-25T08:20:16Z)
• Target-Embedding Autoencoders for Supervised Representation Learning [111.07204912245841]
  This paper analyzes a framework for improving generalization in a purely supervised setting, where the target space is high-dimensional. We motivate and formalize the general framework of target-embedding autoencoders (TEA) for supervised prediction, learning intermediate latent representations jointly optimized to be both predictable from features as well as predictive of targets.
  arXiv  Detail & Related papers  (2020-01-23T02:37:10Z)

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.