Identifying Linear Relational Concepts in Large Language Models
- URL: http://arxiv.org/abs/2311.08968v2
- Date: Fri, 29 Mar 2024 22:14:30 GMT
- Title: Identifying Linear Relational Concepts in Large Language Models
- Authors: David Chanin, Anthony Hunter, Oana-Maria Camburu,
- Abstract summary: Transformer language models (LMs) have been shown to represent concepts as directions in the latent space of hidden activations.
We present a technique called linear relational concepts (LRC) for finding concept directions corresponding to human-interpretable concepts.
- Score: 16.917379272022064
- License: http://creativecommons.org/licenses/by/4.0/
- Abstract: Transformer language models (LMs) have been shown to represent concepts as directions in the latent space of hidden activations. However, for any human-interpretable concept, how can we find its direction in the latent space? We present a technique called linear relational concepts (LRC) for finding concept directions corresponding to human-interpretable concepts by first modeling the relation between subject and object as a linear relational embedding (LRE). We find that inverting the LRE and using earlier object layers results in a powerful technique for finding concept directions that outperforms standard black-box probing classifiers. We evaluate LRCs on their performance as concept classifiers as well as their ability to causally change model output.
Related papers
- Concept Layers: Enhancing Interpretability and Intervenability via LLM Conceptualization [2.163881720692685]
We introduce a new methodology for incorporating interpretability and intervenability into an existing model by integrating Concept Layers into its architecture.
Our approach projects the model's internal vector representations into a conceptual, explainable vector space before reconstructing and feeding them back into the model.
We evaluate CLs across multiple tasks, demonstrating that they maintain the original model's performance and agreement while enabling meaningful interventions.
arXiv Detail & Related papers (2025-02-19T11:10:19Z) - Beyond Single Concept Vector: Modeling Concept Subspace in LLMs with Gaussian Distribution [23.594013836364628]
We propose an approach to approximate the subspace representing a specific concept.
We demonstrate GCS's effectiveness through measuring its faithfulness and plausibility across multiple large language models.
We also use representation intervention tasks to showcase its efficacy in real-world applications such as emotion steering.
arXiv Detail & Related papers (2024-09-30T18:52:53Z) - On the Origins of Linear Representations in Large Language Models [51.88404605700344]
We introduce a simple latent variable model to formalize the concept dynamics of the next token prediction.
Experiments show that linear representations emerge when learning from data matching the latent variable model.
We additionally confirm some predictions of the theory using the LLaMA-2 large language model.
arXiv Detail & Related papers (2024-03-06T17:17:36Z) - Interpretable Neural-Symbolic Concept Reasoning [7.1904050674791185]
Concept-based models aim to address this issue by learning tasks based on a set of human-understandable concepts.
We propose the Deep Concept Reasoner (DCR), the first interpretable concept-based model that builds upon concept embeddings.
arXiv Detail & Related papers (2023-04-27T09:58:15Z) - Unsupervised Interpretable Basis Extraction for Concept-Based Visual
Explanations [53.973055975918655]
We show that, intermediate layer representations become more interpretable when transformed to the bases extracted with our method.
We compare the bases extracted with our method with the bases derived with a supervised approach and find that, in one aspect, the proposed unsupervised approach has a strength that constitutes a limitation of the supervised one and give potential directions for future research.
arXiv Detail & Related papers (2023-03-19T00:37:19Z) - Linear Spaces of Meanings: Compositional Structures in Vision-Language
Models [110.00434385712786]
We investigate compositional structures in data embeddings from pre-trained vision-language models (VLMs)
We first present a framework for understanding compositional structures from a geometric perspective.
We then explain what these structures entail probabilistically in the case of VLM embeddings, providing intuitions for why they arise in practice.
arXiv Detail & Related papers (2023-02-28T08:11:56Z) - Concept Gradient: Concept-based Interpretation Without Linear Assumption [77.96338722483226]
Concept Activation Vector (CAV) relies on learning a linear relation between some latent representation of a given model and concepts.
We proposed Concept Gradient (CG), extending concept-based interpretation beyond linear concept functions.
We demonstrated CG outperforms CAV in both toy examples and real world datasets.
arXiv Detail & Related papers (2022-08-31T17:06:46Z) - Navigating Neural Space: Revisiting Concept Activation Vectors to
Overcome Directional Divergence [14.071950294953005]
Concept Activation Vectors (CAVs) have emerged as a popular tool for modeling human-understandable concepts in the latent space.
In this paper we show that such a separability-oriented leads to solutions, which may diverge from the actual goal of precisely modeling the concept direction.
We introduce pattern-based CAVs, solely focussing on concept signals, thereby providing more accurate concept directions.
arXiv Detail & Related papers (2022-02-07T19:40:20Z) - Translational Concept Embedding for Generalized Compositional Zero-shot
Learning [73.60639796305415]
Generalized compositional zero-shot learning means to learn composed concepts of attribute-object pairs in a zero-shot fashion.
This paper introduces a new approach, termed translational concept embedding, to solve these two difficulties in a unified framework.
arXiv Detail & Related papers (2021-12-20T21:27:51Z) - Closed-Form Factorization of Latent Semantics in GANs [65.42778970898534]
A rich set of interpretable dimensions has been shown to emerge in the latent space of the Generative Adversarial Networks (GANs) trained for synthesizing images.
In this work, we examine the internal representation learned by GANs to reveal the underlying variation factors in an unsupervised manner.
We propose a closed-form factorization algorithm for latent semantic discovery by directly decomposing the pre-trained weights.
arXiv Detail & Related papers (2020-07-13T18:05:36Z)
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.