Identifiable Steering via Sparse Autoencoding of Multi-Concept Shifts

• URL: http://arxiv.org/abs/2502.12179v1
• Date: Fri, 14 Feb 2025 08:49:41 GMT
• Title: Identifiable Steering via Sparse Autoencoding of Multi-Concept Shifts
• Authors: Shruti Joshi, Andrea Dittadi, Sébastien Lachapelle, Dhanya Sridhar,
• Abstract summary: Steering methods manipulate the representations of large language models (LLMs) to induce responses that have desired properties. Traditionally, steering has relied on supervision, such as from contrastive pairs of prompts that vary in a single target concept. We introduce Sparse Shift Autoencoders (SSAEs) that instead map the differences between embeddings to sparse representations.
• Score: 11.81523319216474
• License:
• Abstract: Steering methods manipulate the representations of large language models (LLMs) to induce responses that have desired properties, e.g., truthfulness, offering a promising approach for LLM alignment without the need for fine-tuning. Traditionally, steering has relied on supervision, such as from contrastive pairs of prompts that vary in a single target concept, which is costly to obtain and limits the speed of steering research. An appealing alternative is to use unsupervised approaches such as sparse autoencoders (SAEs) to map LLM embeddings to sparse representations that capture human-interpretable concepts. However, without further assumptions, SAEs may not be identifiable: they could learn latent dimensions that entangle multiple concepts, leading to unintentional steering of unrelated properties. We introduce Sparse Shift Autoencoders (SSAEs) that instead map the differences between embeddings to sparse representations. Crucially, we show that SSAEs are identifiable from paired observations that vary in \textit{multiple unknown concepts}, leading to accurate steering of single concepts without the need for supervision. We empirically demonstrate accurate steering across semi-synthetic and real-world language datasets using Llama-3.1 embeddings.

Related papers

• AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders [73.37603699731329]
  We introduce AxBench, a large-scale benchmark for steering and concept detection. For steering, we find that prompting outperforms all existing methods, followed by finetuning. For concept detection, representation-based methods such as difference-in-means, perform the best.
  arXiv  Detail & Related papers  (2025-01-28T18:51:24Z)
• LF-Steering: Latent Feature Activation Steering for Enhancing Semantic Consistency in Large Language Models [16.37602070339033]
  Large Language Models (LLMs) often generate inconsistent responses when prompted with semantically equivalent paraphrased inputs. We propose LF-Steering, a novel activation steering approach to precisely identify latent feature representations responsible for semantic inconsistency. Our method maps the hidden states of the relevant transformer layer into a sparsely activated, high-dimensional feature space based on a sparse autoencoder.
  arXiv  Detail & Related papers  (2025-01-19T13:06:51Z)
• LatentQA: Teaching LLMs to Decode Activations Into Natural Language [72.87064562349742]
  We introduce LatentQA, the task of answering open-ended questions about model activations in natural language. We propose Latent Interpretation Tuning (LIT), which finetunes a decoder LLM on a dataset of activations and associated question-answer pairs. Our decoder also specifies a differentiable loss that we use to control models, such as debiasing models on stereotyped sentences and controlling the sentiment of generations.
  arXiv  Detail & Related papers  (2024-12-11T18:59:33Z)
• Interpret the Internal States of Recommendation Model with Sparse Autoencoder [26.021277330699963]
  RecSAE is an automatic, generalizable probing method for interpreting the internal states of Recommendation models. We train an autoencoder with sparsity constraints to reconstruct internal activations of recommendation models. We automated the construction of concept dictionaries based on the relationship between latent activations and input item sequences.
  arXiv  Detail & Related papers  (2024-11-09T08:22:31Z)
• 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)
• MASA: Motion-aware Masked Autoencoder with Semantic Alignment for Sign Language Recognition [94.56755080185732]
  We propose a Motion-Aware masked autoencoder with Semantic Alignment (MASA) that integrates rich motion cues and global semantic information. Our framework can simultaneously learn local motion cues and global semantic features for comprehensive sign language representation.
  arXiv  Detail & Related papers  (2024-05-31T08:06:05Z)
• Language-Enhanced Latent Representations for Out-of-Distribution Detection in Autonomous Driving [1.3499500088995464]
  multimodal inputs offer the possibility of taking human language as a latent representation. In this paper, we use the cosine similarity of image and text representations encoded by the multimodal model CLIP as a new representation. Our experiments on realistic driving data show that the language-based latent representation performs better than the traditional representation of the vision encoder.
  arXiv  Detail & Related papers  (2024-05-02T19:27:28Z)
• Learning Common Rationale to Improve Self-Supervised Representation for Fine-Grained Visual Recognition Problems [61.11799513362704]
  We propose learning an additional screening mechanism to identify discriminative clues commonly seen across instances and classes. We show that a common rationale detector can be learned by simply exploiting the GradCAM induced from the SSL objective.
  arXiv  Detail & Related papers  (2023-03-03T02:07:40Z)
• Measuring the Interpretability of Unsupervised Representations via Quantized Reverse Probing [97.70862116338554]
  We investigate the problem of measuring interpretability of self-supervised representations. We formulate the latter as estimating the mutual information between the representation and a space of manually labelled concepts. We use our method to evaluate a large number of self-supervised representations, ranking them by interpretability.
  arXiv  Detail & Related papers  (2022-09-07T16:18:50Z)
• 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.