| Challenge: | Large Language Models (LLMs) are vulnerable to optimization-based jailbreak attacks that exploit internal gradient structure. |
| Approach: | They propose to integrate pretrained Sparse Autoencoders into transformer residual streams at inference time without modifying model weights or blocking gradients. |
| Outcome: | The proposed model reduces jailbreak success rate by 5x compared to baseline models . compared with models with weak white-box attacks, the proposed model is more robust . |
Similar Papers
Sparse Autoencoder Features for Classifications and Transferability (2025.emnlp-main)
Copied to clipboard
| Challenge: | Sparse Autoencoders (SAEs) provide potential for uncovering structured, human-interpretable representations in Large Language Models (LLMs). |
| Approach: | They analyze SAEs for interpretable feature extraction from Large Language Models in safety-critical classification tasks. |
| Outcome: | The proposed framework outperforms hidden-state and BoW models while demonstrating cross-lingual toxicity detection and visual classification tasks. |
On the Versatility of Sparse Autoencoders for In-Context Learning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Sparse autoencoders (SAEs) are emerging as a key analytical tool in interpretability for large language models. |
| Approach: | They propose to use SAEs to extract knowledge from billions of tokens for sparse reconstruction. |
| Outcome: | The proposed model can extract knowledge from billions of tokens for sparse reconstruction. |
Evaluating Adversarial Robustness of Concept Representations in Sparse Autoencoders (2026.eacl-long)
Copied to clipboard
| Challenge: | Existing evaluations of SAEs focus on metrics such as reconstruction-sparsity tradeoff, human (auto-)interpretability, and feature disentanglement, but they neglect robustness of concept representations to input perturbations. |
| Approach: | They propose an unsupervised approach to map LLM embeddings to sparse interpretable concept embeddables via dictionary learning. |
| Outcome: | The proposed framework shows that sparse autoencoders can manipulate concept-based interpretations without denoising or postprocessing. |
Group-SAE: Efficient Training of Sparse Autoencoders for Large Language Models via Layer Groups (2025.emnlp-main)
Copied to clipboard
| Challenge: | Sparse Autoencoders (SAEs) are a promising unsupervised approach for understanding the representations of layers of Large Language Models (LLMs). |
| Approach: | They propose a method that groups similar models and trains a single SAE per group based on representational similarity across layers. |
| Outcome: | Experiments on Pythia family models show that the proposed method significantly accelerates training with minimal impact on reconstruction quality and comparable downstream task performance and interpretability over baseline SAEs trained layer by layer. |
Breaking Bad Tokens: Detoxification of LLMs Using Sparse Autoencoders (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) are ubiquitous in user-facing applications, yet they still generate undesirable toxic outputs, including profanity, vulgarity, and derogatory remarks. |
| Approach: | They leverage sparse autoencoders to identify toxicity-related directions in residual stream of large language models and perform targeted activation steering using the corresponding decoder vectors. |
| Outcome: | The proposed models surpass baselines in reducing toxicity by up to 20%, though fluency can degrade noticeably on GPT-2 Small and Gemma-2-2B. |
Model Unlearning via Sparse Autoencoder Subspace Guided Projections (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing unlearning strategies lack interpretability or fail to provide robust defense against adversarial prompts. |
| Approach: | They propose a framework that leverages SAE features to drive targeted updates in the model’s parameter space. |
| Outcome: | The proposed framework reduces harmful knowledge accuracy by 3.22% compared to baselines and improves adversarial robustness under jailbreak prompts. |
Features that Make a Difference: Leveraging Gradients for Improved Dictionary Learning (2025.findings-naacl)
Copied to clipboard
| Challenge: | Sparse Autoencoders (SAEs) are a promising approach for extracting neural network representations by learning a sparse and overcomplete decomposition of the network’s internal activations. |
| Approach: | They propose a method that learns a sparse and overcomplete decomposition of the network's internal activations and a gradient approach to learn latents. |
| Outcome: | The proposed algorithms improve the performance of the k-sparse autoencoder and the ability to learn latent features. |
Beyond Input Activations: Identifying Influential Latents by Gradient Sparse Autoencoders (2025.emnlp-main)
Copied to clipboard
| Challenge: | Sparse Autoencoders (SAEs) have recently emerged as powerful tools for interpreting and steering the internal representations of large language models (LLMs). |
| Approach: | They propose a method that identifies the most influential latents by incorporating output-side gradient information. |
| Outcome: | The proposed method identifies the most influential latents by incorporating output-side gradient information. |
Unveiling Language-Specific Features in Large Language Models via Sparse Autoencoders (2025.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) exhibit impressive abilities in various domains such as text generation, instruction following, and reasoning. |
| Approach: | They propose a method to decompose the activations of Large Language Models into a sparse linear combination of SAE features. |
| Outcome: | The proposed method shows that some features are strongly related to specific languages, while others are unaffected by ablating them. |
A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Sparse Autoencoders (SAEs) can disentangle complex features into more interpretable components. |
| Approach: | They propose to use Sparse Autoencoders to disentangle LLM features into more interpretable components. |
| Outcome: | The proposed method disentangles complex features into more interpretable components. |