The Mechanics of Interference: Defusing Distractors in RAG via Sparse Autoencoder Interventions (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models exhibit a critical vulnerability to distractor interference when processing retrieval-augmented contexts. |
| Approach: | They propose a mechanistic framework that corrects this failure mode through targeted interventions in the model’s latent space. |
| Outcome: | The proposed framework achieves recovery rates of up to 94% on distractor-vulnerable samples on Gemma-2 and Llama-3 model families across three QA benchmarks. |
Similar Papers
Sparse Latents Steer Retrieval-Augmented Generation (2025.acl-long)
Copied to clipboard
Chunlei Xin, Shuheng Zhou, Huijia Zhu, Weiqiang Wang, Xuanang Chen, Xinyan Guan, Yaojie Lu, Hongyu Lin, Xianpei Han, Le Sun
| Challenge: | In this study, we uncover interpretable latents that govern RAG behavior in large language models . Sparse Autoencoders are used to control large language model (LLM) behavior . |
| Approach: | They leverage Sparse Autoencoders within the LLaMA Scope to uncover latents that govern RAG behaviors. |
| Outcome: | The proposed model can be used to control large language models without architectural modifications. |
The Distracting Effect: Understanding Irrelevant Passages in RAG (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods to detect and use hard distracting passages in RAG can cause problems . retrieved passages contain irrelevant but semantically related information that may mislead the LLM . |
| Approach: | They propose a method to identify and use hard distracting passages to improve RAG . they find that adding retrieved passages is found to ground the LLM response . |
| Outcome: | The proposed method achieves up to 7.5% increase in answering accuracy compared to fine-tuned datasets. |
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. |
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. |
Towards Understanding the Robustness of Sparse Autoencoders (2026.findings-acl)
Copied to clipboard
| 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 . |
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. |
Interpret and Control Dense Retrieval with Sparse Latent Features (2025.naacl-short)
Copied to clipboard
| Challenge: | Dense embeddings deliver strong retrieval performance but lack interpretability and controllability. |
| Approach: | They propose a novel approach using sparse autoencoders to interpret and control dense embeddings via latent sparsity. |
| Outcome: | The proposed approach retains the same retrieval accuracy as the original dense vectors, affirming their faithfulness. |
Do RAG Systems Really Suffer From Positional Bias? (2025.emnlp-main)
Copied to clipboard
| Challenge: | Retrieval Augmented Generation (RAG) improves the factual accuracy of LLMs on knowledgeintensive tasks by including in the prompt passages retrieved from an external corpus. |
| Approach: | They propose to use a retrieval algorithm to add passages from an external corpus to the LLM prompt to improve the factual accuracy of LLMs. |
| Outcome: | The proposed approach improves the factual accuracy of LLMs on knowledgeintensive tasks by including in the prompt passages retrieved from an external corpus. |
Toward Efficient Sparse Autoencoder-Guided Steering for Improved In-Context Learning in Large Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Sparse autoencoders (SAEs) have emerged as a powerful analytical tool in mechanistic interpretability for large language models (LLMs). |
| Approach: | They propose a novel approach that leverages SAEs to enhance the general in-context learning performance of large language models (LLMs). |
| Outcome: | The proposed method yields a 3.5% improvement across diverse text classification tasks and exhibits greater robustness to hyperparameter variations compared to standard steering approaches. |
CRISP: Persistent Concept Unlearning via Sparse Autoencoders (2026.acl-long)
Copied to clipboard
| Challenge: | Recent work has explored sparse autoencoders (SAEs) to perform precise interventions on monosemantic features, but most SAE-based methods operate at inference time, which does not create persistent changes in the model’s parameters. |
| Approach: | They propose a parameter-efficient method for persistent concept unlearning using SAEs that automatically identifies salient SAE features across multiple layers and suppresses their activations. |
| Outcome: | The proposed method outperforms previous methods on safety-critical unlearning tasks from the WMDP benchmark, successfully removing harmful knowledge while preserving general and in-domain capabilities. |