Navigating the Concept Space of Language Models

• URL: http://arxiv.org/abs/2603.23524v1
• Date: Fri, 06 Mar 2026 20:52:50 GMT
• Title: Navigating the Concept Space of Language Models
• Abstract summary: We present Concept Explorer, a scalable interactive system for post-hoc exploration of SAE features.<n>Our approach constructs a multi-resolution manifold over SAE feature embeddings, enabling progressive navigation from coarse concept clusters to fine-grained neighborhoods.<n>We demonstrate the utility of Concept Explorer on SAE features extracted from SmolLM2, where it reveals coherent high-level structure, meaningful subclusters, and distinctive rare concepts.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Sparse autoencoders (SAEs) trained on large language model activations output thousands of features that enable mapping to human-interpretable concepts. The current practice for analyzing these features primarily relies on inspecting top-activating examples, manually browsing individual features, or performing semantic search on interested concepts, which makes exploratory discovery of concepts difficult at scale. In this paper, we present Concept Explorer, a scalable interactive system for post-hoc exploration of SAE features that organizes concept explanations using hierarchical neighborhood embeddings. Our approach constructs a multi-resolution manifold over SAE feature embeddings and enables progressive navigation from coarse concept clusters to fine-grained neighborhoods, supporting discovery, comparison, and relationship analysis among concepts. We demonstrate the utility of Concept Explorer on SAE features extracted from SmolLM2, where it reveals coherent high-level structure, meaningful subclusters, and distinctive rare concepts that are hard to identify with existing workflows.

Related papers

• Cascaded Sparse Autoencoders Learn Multi-Level Visual Concepts in Multimodal LLMs [12.597421665142646]
  We introduce cascaded autoencoders (CSAEs) for learning hierarchical visual concepts in MLLMs.<n> experiments across Qwen3-VL, Gemma-3, and LLaVA on multiple visual datasets show that CSAEs improve interpretability in terms of hierarchical concept coherence.<n>Results on concept steering further demonstrate that the learned concept groups support effective group-level interventions in MLLM outputs.
  arXiv  Detail & Related papers  (2026-06-15T04:10:40Z)
• Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models [80.45129499188461]
  We introduce Qwen-Scope, an open-source suite of sparse autoencoders (SAEs) built on the Qwen model family.<n>We show that SAEs can go beyond post-hoc analysis to serve as practical interfaces for model development along four directions.
  arXiv  Detail & Related papers  (2026-05-12T10:01:06Z)
• PCA-Seg: Revisiting Cost Aggregation for Open-Vocabulary Semantic and Part Segmentation [58.1914505657064]
  This paper proposes a simple yet effective parallel cost aggregation (PCA-Seg) paradigm to alleviate the challenge of knowledge interference between class-level semantics and spatial context.<n>Experiments on eight benchmarks show that each parallel block in PCA-Seg adds merely 0.35M parameters while achieving state-of-the-art OSPS performance.
  arXiv  Detail & Related papers  (2026-03-18T09:26:43Z)
• LUCID-SAE: Learning Unified Vision-Language Sparse Codes for Interpretable Concept Discovery [14.222802170483739]
  LUCID is a vision-language sparse autoencoder that learns a shared latent dictionary for image patch and text token representations.<n> LUCID yields interpretable shared features that support patch-level grounding, establish cross-modal neuron correspondence, and enhance robustness against the concept clustering problem.<n>Our analysis reveals that LUCID's shared features capture diverse semantic categories beyond objects, including actions, attributes, and abstract concepts.
  arXiv  Detail & Related papers  (2026-02-07T02:01:25Z)
• Visual Exploration of Feature Relationships in Sparse Autoencoders with Curated Concepts [8.768503486874623]
  We propose a focused exploration framework that prioritizes curated concepts and their corresponding SAE features over attempts to visualize all available features simultaneously.<n>We present an interactive visualization system that combines topology-based visual encoding with dimensionality reduction to faithfully represent both local and global relationships among selected features.
  arXiv  Detail & Related papers  (2025-11-08T15:36:57Z)
• ConceptScope: Characterizing Dataset Bias via Disentangled Visual Concepts [54.60525564599342]
  ConceptScope is a scalable and automated framework for analyzing visual datasets.<n>It categorizes concepts into target, context, and bias types based on their semantic relevance and statistical correlation to class labels.<n>It reliably detects known biases and uncovers previously unannotated ones.
  arXiv  Detail & Related papers  (2025-10-30T06:46:17Z)
• ConceptViz: A Visual Analytics Approach for Exploring Concepts in Large Language Models [18.456737929856125]
  ConceptViz is a visual analytics system designed to explore concepts in large language models (LLMs)<n>Our results show that ConceptViz enhances interpretability research by streamlining the discovery and validation of meaningful concept representations in LLMs.
  arXiv  Detail & Related papers  (2025-09-20T04:57:20Z)
• Instruction-Guided Fusion of Multi-Layer Visual Features in Large Vision-Language Models [50.98559225639266]
  We investigate the contributions of visual features from different encoder layers using 18 benchmarks spanning 6 task categories.<n>Our findings reveal that multilayer features provide complementary strengths with varying task dependencies, and uniform fusion leads to suboptimal performance.<n>We propose the instruction-guided vision aggregator, a module that dynamically integrates multi-layer visual features based on textual instructions.
  arXiv  Detail & Related papers  (2024-12-26T05:41:31Z)
• Disentangling Dense Embeddings with Sparse Autoencoders [0.0]
  Sparse autoencoders (SAEs) have shown promise in extracting interpretable features from complex neural networks. We present one of the first applications of SAEs to dense text embeddings from large language models. We show that the resulting sparse representations maintain semantic fidelity while offering interpretability.
  arXiv  Detail & Related papers  (2024-08-01T15:46:22Z)
• Discover-then-Name: Task-Agnostic Concept Bottlenecks via Automated Concept Discovery [52.498055901649025]
  Concept Bottleneck Models (CBMs) have been proposed to address the 'black-box' problem of deep neural networks. We propose a novel CBM approach -- called Discover-then-Name-CBM (DN-CBM) -- that inverts the typical paradigm. Our concept extraction strategy is efficient, since it is agnostic to the downstream task, and uses concepts already known to the model.
  arXiv  Detail & Related papers  (2024-07-19T17:50:11Z)
• Unsupervised Learning of Compositional Energy Concepts [70.11673173291426]
  We propose COMET, which discovers and represents concepts as separate energy functions. Comet represents both global concepts as well as objects under a unified framework.
  arXiv  Detail & Related papers  (2021-11-04T17:46:12Z)

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.