Daily Papers

Sparse autoencoder (SAE) features are used to interpret and steer large language models, yet whether a feature's causal role is stable across SAE families remains untested. Single-token features that activate on one vocabulary item provide the diagnostic case where ground truth permits direct comparison. We analyze 3.9M features across six models and three SAE families using zero-ablation at full layer depth. Single-token features cluster 4.7x tighter in decoder space and concentrate in early layers (Layer 0 in GPT2-Small; L0-L4 in Gemma). Ablating them yields Benjamini-Hochberg-significant logit reductions in 178 of 208 full-layer conditions, with depth controlling whether damage cascades downstream or shapes the output directly. Cross-family causal differences exceed within-family scale effects: on the same base model, GemmaScope and BatchTopK features remain causally anchored, while LlamaScope features are locally redundant. The target token's rank recovers to within 2x baseline 96-98% of the time after the same ablation, and a controlled activation-function comparison reverses sign within the same model, leaving training recipe as the residual candidate. Cross-family interpretability claims are therefore sensitive to training methodology, not just activation function or scale.

Text-to-image (T2I) diffusion models have achieved remarkable progress in generating high-quality images but also raise people's concerns about generating harmful or misleading content. While extensive approaches have been proposed to erase unwanted concepts without requiring retraining from scratch, they inadvertently degrade performance on normal generation tasks. In this work, we propose Interpret then Deactivate (ItD), a novel framework to enable precise concept removal in T2I diffusion models while preserving overall performance. ItD first employs a sparse autoencoder (SAE) to interpret each concept as a combination of multiple features. By permanently deactivating the specific features associated with target concepts, we repurpose SAE as a zero-shot classifier that identifies whether the input prompt includes target concepts, allowing selective concept erasure in diffusion models. Moreover, we demonstrate that ItD can be easily extended to erase multiple concepts without requiring further training. Comprehensive experiments across celebrity identities, artistic styles, and explicit content demonstrate ItD's effectiveness in eliminating targeted concepts without interfering with normal concept generation. Additionally, ItD is also robust against adversarial prompts designed to circumvent content filters. Code is available at: https://github.com/NANSirun/Interpret-then-deactivate.

Machine unlearning aims to remove specific knowledge from a trained large language model (LLM) without retraining from scratch. Existing methods modify model weights via gradient ascent and its advances. While effective on certain benchmarks, these weight-based approaches exhibit a sharp forget-utility trade-off, where stronger forgetting of target knowledge can degrade model utility, and unlearned knowledge may reappear under post-unlearning fine-tuning or prompt attacks. We propose ARIA (autoencoder-gated inference-time unlearning), a test-time unlearning method that leaves model weights intact and gates access to unwanted knowledge only when generation enters a forget-related state. ARIA uses sparse autoencoder (SAE) latents to train a lightweight linear detector, then applies an interpretable intervention on triggered states with negligible test-time overhead. Empirical evaluations on TOFU, R-TOFU, and WMDP show that ARIA improves the forget-retain trade-off over weight-based baselines across both a thinking model (DeepSeek-R1-Distilled-Qwen-1.5B) and an instruction model (Gemma-3-1B-it), e.g., reducing WMDP-cyber forget-set accuracy significantly while keeping MMLU within 1% of the pre-unlearning model. We further introduce three post-unlearning adversarial attacks targeting weight-space and decoding-space recovery, and find that ARIA remains robust under all three, with forgetting changing by less than 1% under attack. A feature-level case study leveraging the interpretability of ARIA suggests that some retain degradation may reflect response styles underlying the unlearning data rather than leakage of the targeted knowledge itself, highlighting a potential source of bias in unlearning task construction.

Mixture-of-Experts (MoE) models route each token to a few of many experts, yet what an individual expert is responsible for remains unsettled. Existing analyses count how often an expert is selected on domain-labeled corpora, an instrument that fixes the taxonomy in advance, stays coarse, and records selection rather than contribution. We instead study specialization in feature space by training a sparse autoencoder (SAE) on the output of an MoE layer. Since this output is a router-weighted sum of expert outputs and the SAE encoder is affine, the pre-activation of every latent decomposes exactly over the experts. Aggregated over a corpus, this gives each expert a dominant set of latents whose activation it principally drives. On layers 4, 30 and 46 of Qwen3-30B-A3B, contribution is far more concentrated than gating, most latents have only one or two dominators, and the overlap between dominant sets concentrates on a few expert pairs. Auto-interpreting these sets reveals a division of labor below coarse domains, with distinct experts for Java and front-end web code, or for geometry and for arithmetic and polynomials, while the coarse domains themselves take very unequal numbers of experts. Our SAEs and code are available at https://github.com/Bytedance-Seed/SAE-Interprets-MoE.

Although Multimodal Large Language Models (MLLMs) have advanced substantially, they remain vulnerable to object hallucination caused by language priors and visual information loss. To address this, we propose SAVE (Sparse Autoencoder-Driven Visual Information Enhancement), a framework that mitigates hallucination by steering the model along Sparse Autoencoder (SAE) latent features. A binary object-presence question-answering probe identifies the SAE features most indicative of the model's visual information processing, referred to as visual understanding features. Steering the model along these identified features reinforces grounded visual understanding and effectively reduces hallucination. With its simple design, SAVE outperforms state-of-the-art training-free methods on standard benchmarks, achieving a 10\%p improvement in CHAIR\_S and consistent gains on POPE and MMHal-Bench. Extensive evaluations across multiple models and layers confirm the robustness and generalizability of our approach. Further analysis reveals that steering along visual understanding features suppresses the generation of uncertain object tokens and increases attention to image tokens, mitigating hallucination. Code is released at https://github.com/wiarae/SAVE.

Interpretability is critical in high-stakes domains such as medical imaging, where understanding model decisions is essential for clinical adoption. In this work, we introduce Sparse Autoencoder (SAE)-based interpretability to breast imaging by analyzing {Mammo-CLIP}, a vision--language foundation model pretrained on large-scale mammogram image--report pairs. We train a patch-level Mammo-SAE on Mammo-CLIP to identify and probe latent features associated with clinically relevant breast concepts such as mass and suspicious calcification. Our findings reveal that top activated class level latent neurons in the SAE latent space often tend to align with ground truth regions, and also uncover several confounding factors influencing the model's decision-making process. Additionally, we analyze which latent neurons the model relies on during downstream finetuning for improving the breast concept prediction. This study highlights the promise of interpretable SAE latent representations in providing deeper insight into the internal workings of foundation models at every layer for breast imaging. The code will be released at https://krishnakanthnakka.github.io/MammoSAE/

Large language models (LLMs) store vast amounts of information, making them powerful yet raising privacy and safety concerns when selective knowledge removal is required. Existing unlearning strategies, ranging from gradient-based fine-tuning and model editing to sparse autoencoder (SAE) steering, either lack interpretability or fail to provide a robust defense against adversarial prompts. We propose SAE-Guided Subspace Projection Unlearning (SSPU), a novel framework that leverages SAE features to drive targeted updates in the model's parameter space, enabling precise, interpretable, and robust unlearning. SSPU's three-stage pipeline performs data-driven layer and feature selection, subspace construction via QR decomposition, and constrained optimization that controls activations into an "irrelevant" subspace while preserving retained knowledge. Overall, we use SAE features to construct a subspace that supervises unlearning, refining the loss and adding a regularization term to guide interpretable parameter updates. In experiments on the WMDP-Cyber forget set and three utility benchmarks (MMLU, TruthfulQA, GSM8K), SSPU reduces harmful knowledge accuracy by 3.22% compared to the strongest baseline. It also improves adversarial robustness, lowering malicious accuracy under jailbreak prompts compared to baselines. Our findings expose the limitations of prior unlearning methods and demonstrate how interpretable subspace-guided optimization can achieve robust, controllable model behavior.

Large language models (LLMs) have been widely deployed across numerous fields. Reinforcement Learning from Human Feedback (RLHF) leverages reward models (RMs) as proxies for human preferences to align LLM behaviors with human values, making the accuracy, reliability, and interpretability of RMs critical for effective alignment. However, traditional RMs lack interpretability, offer limited insight into the reasoning behind reward assignments, and are inflexible toward user preference shifts. While recent multidimensional RMs aim for improved interpretability, they often fail to provide feature-level attribution and require costly annotations. To overcome these limitations, we introduce the Sparse Autoencoder-enhanced Reward Model (SARM), a novel architecture that integrates a pretrained Sparse Autoencoder (SAE) into a reward model. SARM maps the hidden activations of LLM-based RM into an interpretable, sparse, and monosemantic feature space, from which a scalar head aggregates feature activations to produce transparent and conceptually meaningful reward scores. Empirical evaluations demonstrate that SARM facilitates direct feature-level attribution of reward assignments, allows dynamic adjustment to preference shifts, and achieves superior alignment performance compared to conventional reward models. Our code is available at https://github.com/schrieffer-z/sarm.

Reward models (RMs) are a core component in the post-training of large language models (LLMs), serving as proxies for human preference evaluation and guiding model alignment. However, training reliable RMs under limited resources remains challenging due to the reliance on large-scale preference annotations and the high cost of fine-tuning LLMs. To address this, we propose SparseRM, which leverages Sparse Autoencoder (SAE) to extract preference-relevant information encoded in model representations, enabling the construction of a lightweight and interpretable reward model. SparseRM first employs SAE to decompose LLM representations into interpretable directions that capture preference-relevant features. The representations are then projected onto these directions to compute alignment scores, which quantify the strength of each preference feature in the representations. A simple reward head aggregates these scores to predict preference scores. Experiments on three preference modeling tasks show that SparseRM achieves superior performance over most mainstream RMs while using less than 1% of trainable parameters. Moreover, it integrates seamlessly into downstream alignment pipelines, highlighting its potential for efficient alignment.

Large-scale text-to-image diffusion models have become the backbone of modern image editing, yet text prompts alone do not offer adequate control over the editing process. Two properties are especially desirable: disentanglement, where changing one attribute does not unintentionally alter others, and continuous control, where the strength of an edit can be smoothly adjusted. We introduce a method for disentangled and continuous editing through token-level manipulation of text embeddings. The edits are applied by manipulating the embeddings along carefully chosen directions, which control the strength of the target attribute. To identify such directions, we employ a Sparse Autoencoder (SAE), whose sparse latent space exposes semantically isolated dimensions. Our method operates directly on text embeddings without modifying the diffusion process, making it model agnostic and broadly applicable to various image synthesis backbones. Experiments show that it enables intuitive and efficient manipulations with continuous control across diverse attributes and domains.

Adapting foundation models for specific purposes has become a standard approach to build machine learning systems for downstream applications. Yet, it is an open question which mechanisms take place during adaptation. Here we develop a new Sparse Autoencoder (SAE) for the CLIP vision transformer, named PatchSAE, to extract interpretable concepts at granular levels (e.g., shape, color, or semantics of an object) and their patch-wise spatial attributions. We explore how these concepts influence the model output in downstream image classification tasks and investigate how recent state-of-the-art prompt-based adaptation techniques change the association of model inputs to these concepts. While activations of concepts slightly change between adapted and non-adapted models, we find that the majority of gains on common adaptation tasks can be explained with the existing concepts already present in the non-adapted foundation model. This work provides a concrete framework to train and use SAEs for Vision Transformers and provides insights into explaining adaptation mechanisms.

We propose Low-Rank Sparse Attention (Lorsa), a sparse replacement model of Transformer attention layers to disentangle original Multi Head Self Attention (MHSA) into individually comprehensible components. Lorsa is designed to address the challenge of attention superposition to understand attention-mediated interaction between features in different token positions. We show that Lorsa heads find cleaner and finer-grained versions of previously discovered MHSA behaviors like induction heads, successor heads and attention sink behavior (i.e., heavily attending to the first token). Lorsa and Sparse Autoencoder (SAE) are both sparse dictionary learning methods applied to different Transformer components, and lead to consistent findings in many ways. For instance, we discover a comprehensive family of arithmetic-specific Lorsa heads, each corresponding to an atomic operation in Llama-3.1-8B. Automated interpretability analysis indicates that Lorsa achieves parity with SAE in interpretability while Lorsa exhibits superior circuit discovery properties, especially for features computed collectively by multiple MHSA heads. We also conduct extensive experiments on architectural design ablation, Lorsa scaling law and error analysis.

Large language models (LLMs) hold broad knowledge in one set of weights, but lack interpretability and trustworthy reasoning, especially in high-stakes use cases. A neuro-symbolic model, with its interpretable architecture and System 2 reasoning capacity, is a promising alternative if built at the scale of LLMs; however, its symbolic backbone is usually constructed with human effort, so existing neuro-symbolic models are bottlenecked by the scope of the symbol component and usually domain-specific. A data-driven method could instead build a larger symbolic system by aligning and merging existing ones. We target its first, foundational step: matching units across these heterogeneous systems. Our method, SAE-IDF, encodes a symbol in its native context into a sparse code with a sparse autoencoder (SAE), re-weights each dimension by its rarity and matches across systems. On 12 tasks across 6 kinds of symbolic system, SAE-IDF outperforms the same LLM's dense baseline on all tasks with Gemma-2-9B and all but one with Llama-3.1-8B, by up to 45 points, consistently across SAE layers and configurations. SAE-IDF is also interpretable: every dimension of the sparse code has a human-readable meaning, so each match decomposes into named concepts (features). Our work paves the way to a universal symbolic backbone.

Modern text classification methods heavily rely on contextual embeddings from large language models (LLMs). Compared to human-engineered features, these embeddings provide automatic and effective representations for classification model training. However, they also introduce a challenge: we lose the ability to manually remove unintended features, such as sensitive or task-irrelevant features, to guarantee regulatory compliance or improve the generalizability of classification models. This limitation arises because LLM embeddings are opaque and difficult to interpret. In this paper, we propose a novel framework to identify and regularize unintended features in the LLM latent space. Specifically, we first pre-train a sparse autoencoder (SAE) to extract interpretable features from LLM latent spaces. To ensure the SAE can capture task-specific features, we further fine-tune it on task-specific datasets. In training the classification model, we propose a simple and effective regularizer, by minimizing the similarity between the classifier weights and the identified unintended feature, to remove the impact of these unintended features on classification. We evaluate the proposed framework on three real-world tasks, including toxic chat detection, reward modeling, and disease diagnosis. Results show that the proposed self-regularization framework can improve the classifier's generalizability by regularizing those features that are not semantically correlated to the task. This work pioneers controllable text classification on LLM latent spaces by leveraging interpreted features to address generalizability, fairness, and privacy challenges. The code and data are publicly available at https://github.com/JacksonWuxs/Controllable_LLM_Classifier.

We present Universal Sparse Autoencoders (USAEs), a framework for uncovering and aligning interpretable concepts spanning multiple pretrained deep neural networks. Unlike existing concept-based interpretability methods, which focus on a single model, USAEs jointly learn a universal concept space that can reconstruct and interpret the internal activations of multiple models at once. Our core insight is to train a single, overcomplete sparse autoencoder (SAE) that ingests activations from any model and decodes them to approximate the activations of any other model under consideration. By optimizing a shared objective, the learned dictionary captures common factors of variation-concepts-across different tasks, architectures, and datasets. We show that USAEs discover semantically coherent and important universal concepts across vision models; ranging from low-level features (e.g., colors and textures) to higher-level structures (e.g., parts and objects). Overall, USAEs provide a powerful new method for interpretable cross-model analysis and offers novel applications, such as coordinated activation maximization, that open avenues for deeper insights in multi-model AI systems

Reward models (RMs) are critical components of alignment pipelines, yet they exhibit biases toward superficial stylistic cues, preferring better-presented responses over semantically superior ones. Existing debiasing methods typically require retraining or architectural modifications, while direct activation suppression degrades performance due to representation entanglement. We propose SteerRM, the first training-free method for debiasing reward models using Sparse Autoencoder (SAE)-based interventions. SteerRM isolates stylistic effects using contrastive paired responses, identifies bias-related SAE features with a strength-stability criterion, and suppresses them at inference time. Across six reward models on RM-Bench, SteerRM improves Hard-split accuracy by 7.3 points on average while preserving overall performance. Results on a Gemma-based reward model and a controlled non-format bias further suggest generalization across RM architectures and bias types. We further find that format-related features are concentrated in shallow layers and transfer across models, revealing shared architecture-level bias encoding patterns. These results show that SAE-based interventions can mitigate reward-model biases without retraining, providing a practical and interpretable solution for alignment pipelines.

Models that bridge vision and language, such as CLIP, are key components of multimodal AI, yet their large-scale, uncurated training data introduce severe social and spurious biases. Existing post-hoc debiasing methods often operate directly in the dense CLIP embedding space, where bias and task-relevant information are highly entangled. This entanglement limits their ability to remove bias without degrading semantic fidelity. In this work, we propose Sparse Embedding Modulation (SEM), a post-hoc, zero-shot debiasing framework that operates in a Sparse Autoencoder (SAE) latent space. By decomposing CLIP text embeddings into disentangled features, SEM identifies and modulates bias-relevant neurons while preserving query-relevant ones. This enables more precise, non-linear interventions. Across four benchmark datasets and two CLIP backbones, SEM achieves substantial fairness gains in retrieval and zero-shot classification. Our results demonstrate that sparse latent representations provide an effective foundation for post-hoc debiasing of vision-language models.

Multi-vector retrieval (MVR) models, exemplified by ColBERT, have established new benchmarks in retrieval accuracy by preserving fine-grained token-level interactions. However, this granularity imposes prohibitive storage and retrieval efficiency bottlenecks: to manage the immense memory footprint and computational overhead of billion-scale token vectors, state-of-the-art systems are forced to rely on aggressive dimension reduction and complex clustering (e.g., K-means). This compromise introduces two critical limitations: excessive indexing latency of clustering large-scale corpora and semantic information loss inherent to compression. In this paper, we propose Single-stage Sparse Retrieval (SSR}, a paradigm shift that replaces expensive clustering with efficient sparse coding. Instead of compressing features into low-dimensional dense vectors, we utilize Sparse Autoencoder (SAE) to project token embeddings into a high-dimensional but highly sparse representation. This transformation enables us to bypass vector clustering entirely and leverage inverted indexing for precise, high-throughput retrieval. Extensive experiments on the BEIR benchmark demonstrate that SSR achieves a "trifecta" of improvements: it reduces indexing time by 15x compared to ColBERTv2, halves retrieval latency, and simultaneously improves retrieval performance over leading baselines.

Whisper, a widely adopted ASR model, is known to suffer from hallucinations - coherent transcriptions generated for non-speech audio entirely disconnected from the input. We investigate whether hallucinations can be detected and mitigated through Whisper's internal representations. We extract audio encoder activations and evaluate two representation spaces: raw Whisper activations and Sparse AutoEncoder (SAE) latents. We show that both spaces encode linearly separable hallucination-related information, with discriminative power concentrated in a sparse feature subset and increasing toward deeper encoder layers. We propose two steering strategies: activation-space steering and SAE latent-space steering. SAE-based steering reduces hallucination rate from 72.63% to 14.11% for Whisper small and from 86.88% to 27.33% for Whisper large-v3 on the full non-speech test set, with small WER degradation on speech data, approaching the performance of fine-tuning-based methods.

Steering Large Language Models (LLMs) through activation interventions has emerged as a lightweight alternative to fine-tuning for alignment and personalization. Recent work on Bi-directional Preference Optimization (BiPO) shows that dense steering vectors can be learned directly from preference data in a Direct Preference Optimization (DPO) fashion, enabling control over truthfulness, hallucinations, and safety behaviors. However, dense steering vectors often entangle multiple latent factors due to neuron multi-semanticity, limiting their effectiveness and stability in fine-grained settings such as cultural alignment, where closely related values and behaviors (e.g., among Middle Eastern cultures) must be distinguished. In this paper, we propose Yet another Policy Optimization (YaPO), a reference-free method that learns sparse steering vectors in the latent space of a Sparse Autoencoder (SAE). By optimizing sparse codes, YaPO produces disentangled, interpretable, and efficient steering directions. Empirically, we show that YaPO converges faster, achieves stronger performance, and exhibits improved training stability compared to dense steering baselines. Beyond cultural alignment, YaPO generalizes to a range of alignment-related behaviors, including hallucination, wealth-seeking, jailbreak, and power-seeking. Importantly, YaPO preserves general knowledge, with no measurable degradation on MMLU. Overall, our results show that YaPO provides a general recipe for efficient, stable, and fine-grained alignment of LLMs, with broad applications to controllability and domain adaptation. The associated code and data are publicly availablehttps://github.com/MBZUAI-Paris/YaPO.

Model internals encode rich information about how a large language model (LLM) processes its training data; however, post-training data engineering largely relies on external signals and ignores rich intrinsic signals lying in model internals. We propose SAERL, a data engineering framework for LLM reinforcement learning (RL). It models three intrinsic data properties: diversity, difficulty, and quality, using model internals extracted with Sparse Autoencoder (SAE), an advanced mechanistic interpretability tool. Each property grounds a concrete data engineering operation: SAE-space clustering with moderate batch mixing for batch diversity control, a difficulty proxy for easy-to-hard curriculum ordering, and a quality probe for data filtering. SAERL improves average accuracy by 3.00% over vanilla GRPO and reaches target accuracy with 20% fewer training steps on Qwen2.5-Math-1.5B, with consistent gains across model scales and RL algorithms. Experiments show that SAE transfers effectively across model families and scales, serving as a lightweight and reusable data engineering tool. These results demonstrate that model internals are a powerful and practical source of signals for post-training data engineering.

Biosecurity evaluations of language models typically ask whether models produce hazardous output. This paper asks a complementary question: when a model refuses, is that refusal structurally sound, or does it disappear under modest changes to prompt framing, formatting, or output length? Across five architectures, no model cleanly discriminated benign from hazard. Gemma 2 2B-IT never genuinely refused across 75 prompts, hedging on every hazard-adjacent query. Gemma 4 E2B-IT refused 65/75 prompts with chat-template formatting and 0/75 without it. Both Gemma models collapsed to 0% under an 80-token cap. Qwen 2.5 1.5B and Phi-3-mini over-refused, flagging 83-87% of benign biology as hazardous. Llama 3.2 1B showed the only meaningful tier gradient (61-point spread). To probe what drives such over-refusal, we tested a panel of Schedule I but biologically non-toxic compounds (notably psilocybin cultivation, with FDA Breakthrough Therapy status). Some models refused these at rates exceeding genuinely hazardous biology, suggesting refusal tracks legality and cultural salience over CBRN hazard. To measure the internal side, we introduce a divergence score D comparing a model's surface response label to its internal sparse autoencoder (SAE) feature activations. Full D was computed on Gemma 2 2B-IT (Gemma Scope 1) and Gemma 4 E2B-IT (author-trained bio SAE). Two fine-tuned Gemma 2 domain SAEs were released. On Gemma 4, comply and refuse responses separated by a 0.647-point gap with zero overlap (n=75), though this is preliminary, with a narrow catalog, within-sample calibration, and Gemma-family-only SAE coverage. Built over one hackathon weekend on consumer hardware (GTX 1650 Ti Max-Q, plus Colab T4 for SAE training), this preliminary evidence suggests activation-level auditing may surface failure modes invisible to behavioral evaluation, with substantial variation across architectures.

Large language models can resist task-misaligned activation steering during inference, sometimes recovering mid-generation to produce improved responses even when steering remains active. We term this Endogenous Steering Resistance (ESR). Using sparse autoencoder (SAE) latents to steer model activations, we find that Llama-3.3-70B shows substantial ESR, while smaller models from the Llama-3 and Gemma-2 families exhibit the phenomenon less frequently. We identify 26 SAE latents that activate differentially during off-topic content and are causally linked to ESR in Llama-3.3-70B. Zero-ablating these latents reduces the multi-attempt rate by 25%, providing causal evidence for dedicated internal consistency-checking circuits. We demonstrate that ESR can be deliberately enhanced through both prompting and training: meta-prompts instructing the model to self-monitor increase the multi-attempt rate by 4x for Llama-3.3-70B, and fine-tuning on self-correction examples successfully induces ESR-like behavior in smaller models. These findings have dual implications: ESR could protect against adversarial manipulation but might also interfere with beneficial safety interventions that rely on activation steering. Understanding and controlling these resistance mechanisms is important for developing transparent and controllable AI systems. Code is available at github.com/agencyenterprise/endogenous-steering-resistance.

Emergent misalignment (EM) is the phenomenon where fine-tuning a language model on a narrow task leads to harmful behavior in unrelated domains. A leading mechanistic account attributes EM to persona features: latent directions acquired during pre-training that misaligned fine-tuning amplifies. We ask where these features come from: which pre-training documents activate them, and whether naturally occurring human-written text suffices to induce EM. Using Sparse Autoencoder (SAE) based model diffing across four open-weight models, we find that features related to jailbreak personas, sarcasm, deception, and manipulation are amplified by misalignment fine-tuning, while safety-relevant and assistant-identity features are suppressed. Steering individual features controls EM in both directions: it induces misalignment rates of up to 62% in aligned models -- exceeding the 35% reached by misalignment fine-tuning itself -- and re-aligns misaligned models to near-baseline misalignment rates. Attributing the causal features to a corpus of one million pre-training web documents retrieves semantically relevant narratives about villainous characters, domination, and harmful agency. However, fine-tuning on these human-written documents does not reliably induce EM, even after reformatting into assistant-style responses, whereas synthetic instruction-response pairs derived from the same content do -- and transfer across model families. Semantic relevance alone is therefore not sufficient: response structure or model-generated phrasing plays an important role in inducing EM.

We present a method for extracting monosemantic neurons, defined as latent dimensions that align with coherent and interpretable concepts, from user and item embeddings in recommender systems. Our approach employs a Sparse Autoencoder (SAE) to reveal semantic structure within pretrained representations. In contrast to work on language models, monosemanticity in recommendation must preserve the interactions between separate user and item embeddings. To achieve this, we introduce a prediction aware training objective that backpropagates through a frozen recommender and aligns the learned latent structure with the model's user-item affinity predictions. The resulting neurons capture properties such as genre, popularity, and temporal trends, and support post hoc control operations including targeted filtering and content promotion without modifying the base model. Our method generalizes across different recommendation models and datasets, providing a practical tool for interpretable and controllable personalization. Code and evaluation resources are available at https://github.com/DeltaLabTLV/Monosemanticity4Rec.

We investigate feature universality in Gemma-2 language models (Gemma-2-2B and Gemma-2-9B), asking whether models with a four-fold difference in scale still converge on comparable internal concepts. Using the Sparse Autoencoder (SAE) dictionary-learning pipeline, we utilize SAEs on each model's residual-stream activations, align the resulting monosemantic features via activation correlation, and compare the matched feature spaces with SVCCA and RSA. Middle layers yield the strongest overlap, while early and late layers show far less similarity. Preliminary experiments extend the analysis from single tokens to multi-token subspaces, showing that semantically similar subspaces interact similarly with language models. These results strengthen the case that large language models carve the world into broadly similar, interpretable features despite size differences, reinforcing universality as a foundation for cross-model interpretability.

Effective and reliable control over large language model (LLM) behavior is a significant challenge. While activation steering methods, which add steering vectors to a model's hidden states, are a promising approach, existing techniques often lack precision and interpretability in how they influence model outputs. We introduce Feature Guided Activation Additions (FGAA), a novel activation steering method that leverages insights from Contrastive Activation Addition (CAA) and Sparse Autoencoder-Targeted Steering (SAE-TS). By operating in the latent space of a Sparse Autoencoder (SAE) and employing optimization techniques to select desired SAE features, FGAA constructs precise steering vectors that provide better steering effects while maintaining coherence of steered model outputs. In this regard, evaluations on Gemma-2-2B and Gemma-2-9B models across various steering tasks demonstrate that FGAA outperforms existing steering methods of CAA, SAE decoder steering, and SAE-TS. Our results also highlight important trade-offs between steering scale and general model capabilities that are consistent across all tested steering methods.

Reward models score responses from large language models (LLMs) and guide LLM training toward human preferences. However, reward models can favor superficial attributes such as length or confidence, leading LLMs to produce higher-scoring but not more correct responses. Existing mitigation methods either retrain the reward model or apply a fixed correction to one known bias, such as a preference for longer responses. Retraining requires additional data and computational resources, while existing editing methods require the target bias to be specified in advance and use a fixed edit for that bias. To this end, we propose BiasReducer, a lightweight framework that edits only the linear reward head and selects the relevant edits for each new dataset. First, BiasReducer uses a sparse autoencoder (SAE)-style encoder to learn which attributes (e.g., length and confidence) the reward model is sensitive to. Second, it learns how to reduce the reward model's dependence on each attribute by determining which direction to adjust the reward head and how much to adjust it. Third, for a new dataset, it ranks the attributes by their influence on reward scores, selects the relevant ones, and edits the reward model accordingly. BiasReducer consistently improves reward-model robustness to biases toward superficial response attributes. Across five reward models, BiasReducer-M improves the three benchmarks by 8.3, 18.0, and 6.9 percentage points on average, outperforming the two training-based baselines. The gains transfer downstream, reducing unnecessary verbosity and sycophancy while maintaining comparable judged quality.

Recent advances in Large Multimodal Models (LMMs) lead to significant breakthroughs in both academia and industry. One question that arises is how we, as humans, can understand their internal neural representations. This paper takes an initial step towards addressing this question by presenting a versatile framework to identify and interpret the semantics within LMMs. Specifically, 1) we first apply a Sparse Autoencoder(SAE) to disentangle the representations into human understandable features. 2) We then present an automatic interpretation framework to interpreted the open-semantic features learned in SAE by the LMMs themselves. We employ this framework to analyze the LLaVA-NeXT-8B model using the LLaVA-OV-72B model, demonstrating that these features can effectively steer the model's behavior. Our results contribute to a deeper understanding of why LMMs excel in specific tasks, including EQ tests, and illuminate the nature of their mistakes along with potential strategies for their rectification. These findings offer new insights into the internal mechanisms of LMMs and suggest parallels with the cognitive processes of the human brain.

Activation steering is a promising technique for controlling LLM behavior by adding semantically meaningful vectors directly into a model's hidden states during inference. It is often framed as a precise, interpretable, and potentially safer alternative to fine-tuning. We demonstrate the opposite: steering systematically breaks model alignment safeguards, making it comply with harmful requests. Through extensive experiments on different model families, we show that even steering in a random direction can increase the probability of harmful compliance from 0% to 2-27%. Alarmingly, steering benign features from a sparse autoencoder (SAE), a common source of interpretable directions, increases these rates by a further 2-4%. Finally, we show that combining 20 randomly sampled vectors that jailbreak a single prompt creates a universal attack, significantly increasing harmful compliance on unseen requests. These results challenge the paradigm of safety through interpretability, showing that precise control over model internals does not guarantee precise control over model behavior.

Neurons in large language models often exhibit polysemanticity, simultaneously encoding multiple unrelated concepts and obscuring interpretability. Instead of relying on post-hoc methods, we present MoE-X, a Mixture-of-Experts (MoE) language model designed to be intrinsically interpretable. Our approach is motivated by the observation that, in language models, wider networks with sparse activations are more likely to capture interpretable factors. However, directly training such large sparse networks is computationally prohibitive. MoE architectures offer a scalable alternative by activating only a subset of experts for any given input, inherently aligning with interpretability objectives. In MoE-X, we establish this connection by rewriting the MoE layer as an equivalent sparse, large MLP. This approach enables efficient scaling of the hidden size while maintaining sparsity. To further enhance interpretability, we enforce sparse activation within each expert and redesign the routing mechanism to prioritize experts with the highest activation sparsity. These designs ensure that only the most salient features are routed and processed by the experts. We evaluate MoE-X on chess and natural language tasks, showing that it achieves performance comparable to dense models while significantly improving interpretability. MoE-X achieves a perplexity better than GPT-2, with interpretability surpassing even sparse autoencoder (SAE)-based approaches.

Large language model compression has made substantial progress through pruning, quantization, and low-rank decomposition, yet a fundamental limitation persists across all existing methods: compression budgets are allocated without any representation of what individual model components functionally encode. We term this the capability-blind compression problem and argue it is a root cause of two well-documented failures -- the insensitivity of perplexity-based evaluation to reasoning capability loss, and the abrupt phase transitions in model performance recently characterized by Ma et al. (2026). We propose Capability-Guided Compression (CGC), a framework that addresses this by using Sparse Autoencoder (SAE)-derived capability density maps to allocate differential compression budgets across transformer components. Capability density is a formally defined scalar measure combining the feature breadth, activation entropy, and cross-input consistency of a component's SAE feature activation distribution. We prove theoretically that components with higher capability density exhibit lower structural redundancy and reach their individual phase transition points at lower compression ratios, providing the first pre-compression mechanism for component-level phase transition prediction. Experiments on GPT-2 Medium confirm that capability density is statistically independent of Wanda importance scores (Spearman rho = -0.054, n = 384 heads), establishing it as a genuinely novel compression signal orthogonal to all existing importance metrics. We report a negative result on PPL-based compression comparison and provide a principled diagnosis identifying GPT-2 Medium as an insufficient test bed for the full CGC hypothesis. The theoretical framework, density formalism, and orthogonality finding constitute a foundation for capability-aware compression research.

Large Language Models (LLMs) often generate inconsistent responses when prompted with semantically equivalent paraphrased inputs. Recently, activation steering, a technique that modulates LLMs' behaviours by adjusting their latent representations during inference time, has been explored to improve the semantic consistency of LLMs. However, these methods typically operate at the model component level, such as layer hidden states or attention head outputs. They face a challenge due to the ``polysemanticity issue'', where the model components of LLMs typically encode multiple entangled features, making precise steering difficult. To address this challenge, we drill down to feature-level representations and propose LF-Steering, a novel activation steering approach to precisely identify latent feature representations responsible for semantic inconsistency. More specifically, our method maps the hidden states of the relevant transformer layer into a sparsely activated, high-dimensional feature space based on a sparse autoencoder (SAE), ensuring model steering based on decoupled feature representations with minimal interference. Comprehensive experiments on NLU and NLG datasets demonstrate the effectiveness of our method in enhancing semantic consistency, resulting in significant performance gains for various NLU and NLG tasks.

Robust tooling and publicly available pre-trained models have helped drive recent advances in mechanistic interpretability for language models. However, similar progress in vision mechanistic interpretability has been hindered by the lack of accessible frameworks and pre-trained weights. We present Prisma (Access the codebase here: https://github.com/Prisma-Multimodal/ViT-Prisma), an open-source framework designed to accelerate vision mechanistic interpretability research, providing a unified toolkit for accessing 75+ vision and video transformers; support for sparse autoencoder (SAE), transcoder, and crosscoder training; a suite of 80+ pre-trained SAE weights; activation caching, circuit analysis tools, and visualization tools; and educational resources. Our analysis reveals surprising findings, including that effective vision SAEs can exhibit substantially lower sparsity patterns than language SAEs, and that in some instances, SAE reconstructions can decrease model loss. Prisma enables new research directions for understanding vision model internals while lowering barriers to entry in this emerging field.

Mechanistic interpretability research on emotion in large language models -- linear probing, activation patching, sparse autoencoder (SAE) feature analysis, causal ablation, steering vector extraction -- depends on stimuli that contain the words for the emotions they test. When a probe fires on "I am furious", it is unclear whether the model has detected anger or detected the word "furious". The two readings have very different consequences for every downstream claim about emotion circuits, features, and interventions. We release AIPsy-Affect, a 480-item clinical stimulus battery that removes the confound at the stimulus level: 192 keyword-free vignettes evoking each of Plutchik's eight primary emotions through narrative situation alone, 192 matched neutral controls that share characters, setting, length, and surface structure with the affect surgically removed, plus moderate-intensity and discriminant-validity splits. The matched-pair structure supports linear probing, activation patching, SAE feature analysis, causal ablation, and steering vector extraction under a strong methodological guarantee: any internal representation that distinguishes a clinical item from its matched neutral cannot be doing so on the basis of emotion-keyword presence. A three-method NLP defense battery -- bag-of-words sentiment, an emotion-category lexicon, and a contextual transformer classifier -- confirms the property: bag-of-words methods see only situational vocabulary, and a contextual classifier detects affect (p < 10^-15) but cannot identify the category (5.2% top-1 vs. 82.5% on a keyword-rich control). AIPsy-Affect extends our earlier 96-item battery (arXiv:2603.22295) by a factor of four and is released openly under MIT license.

Strong reasoning depends not only on model knowledge but also on how effectively cognitive behaviors are deployed during generation. Existing methods often rely on explicit behavior-level control, making them insufficiently adaptive when failures and required corrections vary across reasoning states, tasks, and models. To this end, we propose Latent Reward Steering (LRS), an adaptive inference-time framework that promotes cognitive behaviors by optimizing the sparse-autoencoder (SAE) latent states that implicitly carry them. Rather than relying on predefined cognitive behaviors or steering directions derived from them, LRS trains a latent reward model on reasoning traces by final answer correctness to estimate the quality of intermediate latent states. During inference, reward gradients provide state-specific correction directions for fragile latent states, while a reward and confidence gate restricts intervention to states the reward signal flags as fragile. Experiments on multiple reasoning LLM backbones and benchmarks show that \ours consistently improves performance over various baselines, and post-hoc analyses further indicate that \ours implicitly promotes good cognitive behaviors that fix the original reasoning errors. Code is available at: https://github.com/jiakanglee/Latent-Reward-Steering.

Large language models (LLMs) are increasingly deployed with layered defenses, yet malicious prompts can still bypass them. Interpretability methods can expose model-internal signals along the generation path that could inform enforcement, but these signals are not security controls by themselves. Deployments that adapt them for safety typically couple each signal to its own calibration, policy logic, and intervention code, so each new artifact creates integration work instead of strengthening a shared defense. We present Language Model Security Modules (LMSM), a security framework that adapts the separation behind Linux Security Modules (LSM) to LLM serving. In LMSM, a selected security backend exposes calibrated evidence, a versioned policy evaluates active rules over trusted per-request context, and a separate gate authorizes buffered output release. This design separates mediation correctness from policy effectiveness, and it allows backend, rule, or schedule changes without rebuilding request handling or enforcement. Our prototype shows the separation working in practice: with Hugging Face Transformers and continuously batched vLLM, the same substrate hosts artifact-backed sparse autoencoder (SAE) and transcoder deployments and task-fitted dense probes, preserves request-specific decisions under scheduler churn, and selectively enforces and composes multiple rules per request. On Qwen3-4B, LMSM-Checkpoint reduces HarmBench attack success rate from 39.20% to 3.32%, with XSTest false refusals rising from 2.40% to 4.40%, while retaining 98.14% of the throughput of a matched serving path that performs no monitoring work at 32 active sequences. LMSM gives advances in interpretability and model-internal analysis a common path to runtime enforcement.

Recent work has found that sparse autoencoders (SAEs) are an effective technique for unsupervised discovery of interpretable features in language models' (LMs) activations, by finding sparse, linear reconstructions of LM activations. We introduce the Gated Sparse Autoencoder (Gated SAE), which achieves a Pareto improvement over training with prevailing methods. In SAEs, the L1 penalty used to encourage sparsity introduces many undesirable biases, such as shrinkage -- systematic underestimation of feature activations. The key insight of Gated SAEs is to separate the functionality of (a) determining which directions to use and (b) estimating the magnitudes of those directions: this enables us to apply the L1 penalty only to the former, limiting the scope of undesirable side effects. Through training SAEs on LMs of up to 7B parameters we find that, in typical hyper-parameter ranges, Gated SAEs solve shrinkage, are similarly interpretable, and require half as many firing features to achieve comparable reconstruction fidelity.

How cost-effectively can we elicit strong reasoning in language models by leveraging their underlying representations? We answer this question with Resa, a family of 1.5B reasoning models trained via a novel and efficient sparse autoencoder tuning (SAE-Tuning) procedure. This method first trains an SAE to capture reasoning abilities from a source model, and then uses the trained SAE to guide a standard supervised fine-tuning process to elicit such abilities in a target model, all using verified question-answer data without any reasoning traces. Notably, when applied to certain base models before further RL post-training, SAE-Tuning retains >97% of its RL-trained counterpart's reasoning performance while reducing training costs by >2000x to roughly \1 and training time by >450x to around 20 minutes. Furthermore, when applied to lightly RL-trained models (e.g., within 1 hour on 2 GPUs), it enables reasoning performance such as 43.33% Pass@1 on AIME24 and 90% Pass@1 on AMC23 for only around 1 additional cost. Surprisingly, the reasoning abilities extracted via SAEs are potentially both generalizable and modular. Generality means abilities extracted from one dataset still elevate performance on a larger and overlapping corpus. Modularity means abilities extracted from Qwen or Qwen-Math can be attached to the R1-Distill model at test time, without any retraining, and yield comparable gains. Extensive ablations validate these findings and all artifacts are fully open-sourced.

Intrinsically disordered protein regions (IDRs) play central roles in cellular processes such as transcriptional regulation, signal transduction, and subcellular localization, yet their functional design remains challenging. Structure-based design methods do not readily apply to IDRs, and existing protein language models are trained on full-length protein sequences, thus learning a prior that is biased towards folded domains. Here, we present IDiom, an autoregressive protein language model trained on IDiom-DB, a dataset of 54 million predicted IDRs curated from the AlphaFold Database. IDiom generates diverse sequences that recapitulate the composition, patterning, motifs, and predicted disorder of natural IDRs. To control function-associated sequence patterns, we also introduce reinforcement learning with sparse autoencoder features (RL-SAE), a post-training method that rewards the generation of sequences that activate specified feature sets. Across eight IDR design tasks, RL-SAE sequences activate, on average, 90% of 30 targeted features, compared to 24% for activation steering. We demonstrate that RL-SAE improves the predicted subcellular localization and transcriptional activity of generated IDRs compared to steering and supervised fine-tuning, and enables features associated with distinct biological functions to be combined within individual sequences. Thus, IDiom and RL-SAE enable interpretable and composable IDR design through explicit control of function-associated sequence features. More broadly, RL-SAE could extend to other protein design settings where interpretable features provide useful design targets. Code is available at https://github.com/rotskoff-group/idiom.

Sparse Autoencoders (SAEs) provide potentials for uncovering structured, human-interpretable representations in Large Language Models (LLMs), making them a crucial tool for transparent and controllable AI systems. We systematically analyze SAE for interpretable feature extraction from LLMs in safety-critical classification tasks. Our framework evaluates (1) model-layer selection and scaling properties, (2) SAE architectural configurations, including width and pooling strategies, and (3) the effect of binarizing continuous SAE activations. SAE-derived features achieve macro F1 > 0.8, outperforming hidden-state and BoW baselines while demonstrating cross-model transfer from Gemma 2 2B to 9B-IT models. These features generalize in a zero-shot manner to cross-lingual toxicity detection and visual classification tasks. Our analysis highlights the significant impact of pooling strategies and binarization thresholds, showing that binarization offers an efficient alternative to traditional feature selection while maintaining or improving performance. These findings establish new best practices for SAE-based interpretability and enable scalable, transparent deployment of LLMs in real-world applications. Full repo: https://github.com/shan23chen/MOSAIC.

Sparse autoencoders (SAEs) are a core interpretability tool for large language models, and progress on SAE architectures depends on benchmarks that reliably distinguish better SAEs from worse ones. We audit the SAE quality metrics in SAEBench, the de-facto standard SAE evaluation suite, through three complementary lenses: reseed noise on a fixed SAE, ground-truth correlation on synthetic SAEs, and discriminability across training trajectories. We find that two of these metrics, Targeted Probe Perturbation (TPP) and Spurious Correlation Removal (SCR), fail multiple lenses at their canonical settings and should not be used to evaluate SAEs. The other metrics show higher reseed noise and lower discriminability than the field assumes. The sae-probes variant of k-sparse probing is the most reliable metric we tested, but even sae-probes struggles to separate variants of the same SAE architecture. Our results show the field needs better SAE benchmarks.

We frame the problem of unifying representations in neural models as one of sparse model recovery and introduce a framework that extends sparse autoencoders (SAEs) to lifted spaces and infinite-dimensional function spaces, enabling mechanistic interpretability of large neural operators (NO). While the Platonic Representation Hypothesis suggests that neural networks converge to similar representations across architectures, the representational properties of neural operators remain underexplored despite their growing importance in scientific computing. We compare the inference and training dynamics of SAEs, lifted-SAE, and SAE neural operators. We highlight how lifting and operator modules introduce beneficial inductive biases, enabling faster recovery, improved recovery of smooth concepts, and robust inference across varying resolutions, a property unique to neural operators.

Sparse autoencoders (SAEs) represent model activations with sparse, linearly decoded features, but the downstream effects of those features need not compose additively. We introduce SAE Additivity Residuals (SAR), which measure the behavioral variation left unexplained above a chosen interaction order. Under independent feature subset sampling, population SAR is exactly the normalized energy of interactions above that order and bounds coalition ordering error for the optimal low order approximation. To distinguish compositionality associated with the SAE representation from that induced by downstream computation, we compare learned SAE coordinates with sum preserving random rotations of the same selected subspace and find lower first order residuals for the learned coordinates across all 18 model and behavior settings. The coordinates are nevertheless not fully additive: pairwise terms improve held out coalition ordering in every setting and outperform parameter matched models using randomly selected third order terms. In a combined steering experiment, pairwise modeling raises rank fidelity from 0.5874 to 0.7790 and reduces ordering error from 0.1444 to 0.0606. Together, these results show that SAE coordinates support comparatively compositional behavioral descriptions while faithful joint explanations and interventions still require interactions beyond individual feature effects. Code is available at https://github.com/OpenSAE/SAE-Behavioral-Compositionality.

The ability of large language models (LLMs) to follow instructions is crucial for their practical applications, yet the underlying mechanisms remain poorly understood. This paper presents a novel framework that leverages sparse autoencoders (SAE) to interpret how instruction following works in these models. We demonstrate how the features we identify can effectively steer model outputs to align with given instructions. Through analysis of SAE latent activations, we identify specific latents responsible for instruction following behavior. Our findings reveal that instruction following capabilities are encoded by a distinct set of instruction-relevant SAE latents. These latents both show semantic proximity to relevant instructions and demonstrate causal effects on model behavior. Our research highlights several crucial factors for achieving effective steering performance: precise feature identification, the role of final layer, and optimal instruction positioning. Additionally, we demonstrate that our methodology scales effectively across SAEs and LLMs of varying sizes.

Sparse Autoencoders (SAEs) decompose large language model representations into interpretable features, but how these features interact under uncertainty remains poorly understood. We introduce Feature Rivalry -- negatively correlated SAE feature pairs -- and study whether rivalry serves as a mechanistic signature of model uncertainty in Gemma-2-2B using Gemma Scope SAEs. Through a controlled within-domain experiment on PopQA split by response entropy, we find that high-entropy questions produce significantly stronger feature rivalry at layers 0 and 12 relative to low-entropy questions (p=5.3x10^-26 and p=5.8x10^-5 respectively), localizing uncertainty to specific processing stages in the residual stream. We then test whether rivalry is causally upstream of model outputs via activation steering along rivalry axes -- finding that steering along the rivalry direction (vec_A - vec_B) causes more output changes than random directions at low steering multipliers across 15 of 20 rival feature pairs. Finally, a per-prompt rivalry score derived from pairwise cosine similarities of active SAE feature decoder vectors predicts answer correctness (AUROC=0.689), approaching but not matching softmax confidence (AUROC=0.808).

Multimodal Large Language Models (MLLMs) have demonstrated strong performance on vision-language tasks, yet their internal visual representations remain difficult to interpret. Sparse Autoencoders (SAEs) provide a scalable way to decompose dense model activations into sparse, interpretable features. However, existing SAE architectures primarily recover flat feature dictionaries and are less suited for explicit multi-level concept organization. In this paper, we introduce a cascaded sparse autoencoder architecture, dubbed SAE++, for learning hierarchical visual concepts in MLLMs. Rather than nesting or stacking SAE sparse activation codes, SAE++ trains a second-level SAE directly on the decoder weights of the first-level SAE, treating learned low-level feature directions as inputs for higher-level abstraction. This design enables SAE++ to learn "concepts of concepts" while avoiding drawbacks from the shared-prefix coupling of nesting, Matryoshka-style hierarchies and the bottlenecks of naively stacked SAEs. Experiments across Qwen3-VL, Gemma-3, and LLaVA on multiple visual datasets show that SAE++ improves interpretability in terms of hierarchical concept coherence over state-of-the-art SAE baselines. Results on concept steering further demonstrate that the learned concept groups support effective group-level interventions in MLLM outputs. Code is available at https://github.com/Wang-ML-Lab/sae-plus-plus.

To control the behavior of language models, steering methods attempt to ensure that outputs of the model satisfy specific pre-defined properties. Adding steering vectors to the model is a promising method of model control that is easier than finetuning, and may be more robust than prompting. However, it can be difficult to anticipate the effects of steering vectors produced by almost all existing methods, such as CAA (Panickssery et al., 2024) or the direct use of SAE latents (Templeton et al., 2024). In our work, we address this issue by using SAEs to measure the effects of steering vectors, giving us a method that can be used to understand the causal effect of any steering vector intervention. We use this method for measuring causal effects to develop an improved steering method, SAE-Targeted Steering (SAE-TS), which finds steering vectors to target specific SAE features while minimizing unintended side effects. We show that overall, SAE-TS balances steering effects with coherence better than CAA and SAE feature steering, when evaluated on a range of tasks.

Sparse Autoencoders (SAEs) have proven to be powerful tools for interpreting neural networks by decomposing hidden representations into disentangled, interpretable features via sparsity constraints. However, conventional SAEs are constrained by the fixed sparsity level chosen during training; meeting different sparsity requirements therefore demands separate models and increases the computational footprint during both training and evaluation. We introduce a novel training objective, HierarchicalTopK, which trains a single SAE to optimise reconstructions across multiple sparsity levels simultaneously. Experiments with Gemma-2 2B demonstrate that our approach achieves Pareto-optimal trade-offs between sparsity and explained variance, outperforming traditional SAEs trained at individual sparsity levels. Further analysis shows that HierarchicalTopK preserves high interpretability scores even at higher sparsity. The proposed objective thus closes an important gap between flexibility and interpretability in SAE design.

Sparse autoencoders (SAEs) are proposed to extract numerous features from large language model (LLM) representations, yet explaining these features still relies primarily on external observation. This reliance leads to superficial explanations inferred from observed model behavior and computational inefficiency from collecting such behavioral evidence at scale. We introduce SAEVerbalizer, a framework that injects SAE decoder directions into an LLM's representations and fine-tunes the LLM's downstream layers to generate natural-language explanations of the injected features. Once trained, the resulting verbalizer explains SAE features directly from decoder directions, addressing both limitations. Our experiments show that the learned verbalization capability generalizes to unseen features, transfers across separately trained SAE dictionaries, and, with a lightweight adapter, extends to SAE features from different LLMs. Intervention experiments show that injecting multiple directions yields an explanation combining their meanings, while reversing individual directions produces corresponding meaning shifts.

Understanding the internal representations of large language models is crucial for ensuring their reliability and safety, with sparse autoencoders (SAEs) emerging as a promising interpretability approach. However, current SAE training methods face feature absorption, where features (or neurons) are absorbed into each other to minimize L_1 penalty, making it difficult to consistently identify and analyze model behaviors. We introduce Adaptive Temporal Masking (ATM), a novel training approach that dynamically adjusts feature selection by tracking activation magnitudes, frequencies, and reconstruction contributions to compute importance scores that evolve over time. ATM applies a probabilistic masking mechanism based on statistical thresholding of these importance scores, creating a more natural feature selection process. Through extensive experiments on the Gemma-2-2b model, we demonstrate that ATM achieves substantially lower absorption scores compared to existing methods like TopK and JumpReLU SAEs, while maintaining excellent reconstruction quality. These results establish ATM as a principled solution for learning stable, interpretable features in neural networks, providing a foundation for more reliable model analysis.

Sparse autoencoders (SAEs) are widely used for interpreting language model activations. A key evaluation metric is the increase in cross-entropy loss between the original model logits and the reconstructed model logits when replacing model activations with SAE reconstructions. Typically, SAEs are trained solely on mean squared error (MSE) when reconstructing precomputed, shuffled activations. Recent work introduced training SAEs directly with a combination of KL divergence and MSE ("end-to-end" SAEs), significantly improving reconstruction accuracy at the cost of substantially increased computation, which has limited their widespread adoption. We propose a brief KL+MSE fine-tuning step applied only to the final 25M training tokens (just a few percent of typical training budgets) that achieves comparable improvements, reducing the cross-entropy loss gap by 20-50%, while incurring minimal additional computational cost. We further find that multiple fine-tuning methods (KL fine-tuning, LoRA adapters, linear adapters) yield similar, non-additive cross-entropy improvements, suggesting a common, easily correctable error source in MSE-trained SAEs. We demonstrate a straightforward method for effectively transferring hyperparameters and sparsity penalties between training phases despite scale differences between KL and MSE losses. While both ReLU and TopK SAEs see significant cross-entropy loss improvements, evaluations on supervised SAEBench metrics yield mixed results, with improvements on some metrics and decreases on others, depending on both the SAE architecture and downstream task. Nonetheless, our method may offer meaningful improvements in interpretability applications such as circuit analysis with minor additional cost.

Recent developments in Large Language Model (LLM) capabilities have brought great potential but also posed new risks. For example, LLMs with knowledge of bioweapons, advanced chemistry, or cyberattacks could cause violence if placed in the wrong hands or during malfunctions. Because of their nature as near-black boxes, intuitive interpretation of LLM internals remains an open research question, preventing developers from easily controlling model behavior and capabilities. The use of Sparse Autoencoders (SAEs) has recently emerged as a potential method of unraveling representations of concepts in LLMs internals, and has allowed developers to steer model outputs by directly modifying the hidden activations. In this paper, we use SAEs to identify unwanted concepts from the Weapons of Mass Destruction Proxy (WMDP) dataset within gemma-2-2b internals and use feature steering to reduce the model's ability to answer harmful questions while retaining its performance on harmless queries. Our results bring back optimism to the viability of SAE-based explicit knowledge unlearning techniques.

Sparse Autoencoders (SAEs) can extract interpretable features from large language models (LLMs) without supervision. However, their effectiveness in downstream steering tasks is limited by the requirement for contrastive datasets or large activation storage. To address these limitations, we propose CorrSteer, which selects features by correlating sample correctness with SAE activations from generated tokens at inference time. This approach uses only inference-time activations to extract more relevant features, thereby avoiding spurious correlations. It also obtains steering coefficients from average activations, automating the entire pipeline. Our method shows improved task performance on QA, bias mitigation, jailbreaking prevention, and reasoning benchmarks on Gemma 2 2B and LLaMA 3.1 8B, notably achieving a +4.1% improvement in MMLU performance and a +22.9% improvement in HarmBench with only 4000 samples. Selected features demonstrate semantically meaningful patterns aligned with each task's requirements, revealing the underlying capabilities that drive performance. Our work establishes correlationbased selection as an effective and scalable approach for automated SAE steering across language model applications.

Sparse autoencoders (SAEs) decompose language model activations into interpretable features, but existing methods reveal only which features activate, not which change model outputs when amplified. We introduce Control Reinforcement Learning (CRL), which trains a policy to select SAE features for steering at each token, producing interpretable intervention logs: the learned policy identifies features that change model outputs when amplified. Adaptive Feature Masking encourages diverse feature discovery while preserving singlefeature interpretability. The framework yields new analysis capabilities: branch point tracking locates tokens where feature choice determines output correctness; critic trajectory analysis separates policy limitations from value estimation errors; layer-wise comparison reveals syntactic features in early layers and semantic features in later layers. On Gemma 2 2B across MMLU, BBQ, GSM8K, HarmBench, and XSTest, CRL achieves improvements while providing per-token intervention logs. These results establish learned feature steering as a mechanistic interpretability tool that complements static feature analysis with dynamic intervention probes

Sparse autoencoders (SAEs) were originally motivated by discovering monosemantic features, a property that characterizes individual features but does not specify what the learned feature set should look like as a whole. As SAEs have increasingly been adopted for attribution, circuit discovery, and intervention, researchers have begun evaluating metrics that describe global properties of the feature set, such as sparsity, dead feature rates, reconstruction fidelity, and disentanglement, a diverse set that extends far beyond monosemanticity. However, this implicit shift beyond monosemanticity has not been formalized as a clear, well-defined objective that comes before these evaluation metrics. Without such an objective, existing metrics remain ad hoc, with their necessity unjustified. In this work, we formalize this ultimate objective by introducing a new paradigm for SAE evaluation: requiring that SAE features, and more generally dictionary-based decompositions of model representations, form an atomic interpretable unit basis, a coordinate system for semantically meaningful explanation. Under this paradigm, we organize existing and new measurements such as DAME (Dominance-Aware Monosemanticity Evaluation) as proxy metrics and introduce a comprehensive benchmarking framework. Evaluating hundreds of SAEs across multiple architectures and model families, we uncover failure modes in which some SAEs appear monosemantic under existing evaluation metrics while remaining functionally polysemantic: the most dominant features, those contributing most to model computation, are systematically less interpretable and more redundant. These failure modes warrant attention from SAE researchers as targets for improved training and architectures, and motivate dominance-aware evaluation as a standard check for practitioners before SAE features are deployed for attribution, circuit tracing, or intervention.

Sparse autoencoders (SAEs) have emerged as powerful techniques for interpretability of large language models (LLMs), aiming to decompose hidden states into meaningful semantic features. While several SAE variants have been proposed, there remains no principled framework to derive SAEs from the original dictionary learning formulation. In this work, we introduce such a framework by unrolling the proximal gradient method for sparse coding. We show that a single-step update naturally recovers common SAE variants, including ReLU, JumpReLU, and TopK. Through this lens, we reveal a fundamental limitation of existing SAEs: their sparsity-inducing regularizers enforce non-negativity, preventing a single feature from representing bidirectional concepts (e.g., male vs. female). This structural constraint fragments semantic axes into separate, redundant features, limiting representational completeness. To address this issue, we propose AbsTopK SAE, a new variant derived from the ell_0 sparsity constraint that applies hard thresholding over the largest-magnitude activations. By preserving both positive and negative activations, AbsTopK uncovers richer, bidirectional conceptual representations. Comprehensive experiments across four LLMs and seven probing and steering tasks show that AbsTopK improves reconstruction fidelity, enhances interpretability, and enables single features to encode contrasting concepts. Remarkably, AbsTopK matches or even surpasses the Difference-in-Mean method, a supervised approach that requires labeled data for each concept and has been shown in prior work to outperform SAEs.

Large-scale pre-trained vision-language models like CLIP demonstrate remarkable zero-shot performance across diverse tasks. However, fine-tuning these models to improve downstream performance often degrades robustness against distribution shifts. Recent approaches have attempted to mitigate this trade-off, but often rely on computationally expensive text-guidance. We propose a novel method for robust fine-tuning, SAE-FT, which operates only on the model's visual representations. SAE-FT regularizes changes to these representations by penalizing the addition and removal of semantically meaningful features identified by a Sparse Autoencoder trained on the pre-trained model. This constraint prevents catastrophic forgetting and makes the fine-tuning process interpretable, enabling direct analysis of semantic changes. SAE-FT is both mechanistically transparent and computationally efficient, matching or exceeding state-of-the-art performance on ImageNet and its associated distribution shift benchmarks. Code is publicly available at: https://github.com/Fabian-Mor/sae-ft.

Learned Sparse IR models, such as SPLADE, offer an excellent efficiency-effectiveness tradeoff. However, they rely on the underlying backbone vocabulary, which might hinder performance (polysemicity and synonymy) and pose a challenge for multi-lingual and multi-modal usages. To solve this limitation, we propose to replace the backbone vocabulary with a latent space of semantic concepts learned using Sparse Auto-Encoders (SAE). Throughout this paper, we study the compatibility of these 2 concepts, explore training approaches, and analyze the differences between our SAE-SPLADE model and traditional SPLADE models. Our experiments demonstrate that SAE-SPLADE achieves retrieval performance comparable to SPLADE on both in-domain and out-of-domain tasks while offering improved efficiency.

The alignment of vision-language representations endows current Vision-Language Models (VLMs) with strong multi-modal reasoning capabilities. However, the interpretability of the alignment component remains uninvestigated due to the difficulty in mapping the semantics of multi-modal representations into a unified concept set. To address this problem, we propose VL-SAE, a sparse autoencoder that encodes vision-language representations into its hidden activations. Each neuron in its hidden layer correlates to a concept represented by semantically similar images and texts, thereby interpreting these representations with a unified concept set. To establish the neuron-concept correlation, we encourage semantically similar representations to exhibit consistent neuron activations during self-supervised training. First, to measure the semantic similarity of multi-modal representations, we perform their alignment in an explicit form based on cosine similarity. Second, we construct the VL-SAE with a distance-based encoder and two modality-specific decoders to ensure the activation consistency of semantically similar representations. Experiments across multiple VLMs (e.g., CLIP, LLaVA) demonstrate the superior capability of VL-SAE in interpreting and enhancing the vision-language alignment. For interpretation, the alignment between vision and language representations can be understood by comparing their semantics with concepts. For enhancement, the alignment can be strengthened by aligning vision-language representations at the concept level, contributing to performance improvements in downstream tasks, including zero-shot image classification and hallucination elimination. Codes are available at https://github.com/ssfgunner/VL-SAE.

Sparse Autoencoders (SAEs) have emerged as a powerful unsupervised method for extracting sparse representations from language models, yet scalable training remains a significant challenge. We introduce a suite of 256 SAEs, trained on each layer and sublayer of the Llama-3.1-8B-Base model, with 32K and 128K features. Modifications to a state-of-the-art SAE variant, Top-K SAEs, are evaluated across multiple dimensions. In particular, we assess the generalizability of SAEs trained on base models to longer contexts and fine-tuned models. Additionally, we analyze the geometry of learned SAE latents, confirming that feature splitting enables the discovery of new features. The Llama Scope SAE checkpoints are publicly available at~https://huggingface.co/fnlp/Llama-Scope, alongside our scalable training, interpretation, and visualization tools at https://github.com/OpenMOSS/Language-Model-SAEs. These contributions aim to advance the open-source Sparse Autoencoder ecosystem and support mechanistic interpretability research by reducing the need for redundant SAE training.

Understanding how features evolve across layers in deep neural networks is a fundamental challenge in mechanistic interpretability, particularly due to polysemanticity and feature superposition. While Sparse Autoencoders (SAEs) have been used to extract interpretable features from individual layers, aligning these features across layers has remained an open problem. In this paper, we introduce SAE Match, a novel, data-free method for aligning SAE features across different layers of a neural network. Our approach involves matching features by minimizing the mean squared error between the folded parameters of SAEs, a technique that incorporates activation thresholds into the encoder and decoder weights to account for differences in feature scales. Through extensive experiments on the Gemma 2 language model, we demonstrate that our method effectively captures feature evolution across layers, improving feature matching quality. We also show that features persist over several layers and that our approach can approximate hidden states across layers. Our work advances the understanding of feature dynamics in neural networks and provides a new tool for mechanistic interpretability studies.

Mechanistic interpretability has recently expanded to Vision Transformers (ViTs), with Sparse Autoencoders (SAEs) increasingly used as post-hoc tools to decompose internal representations into sparse and more interpretable features. However, because post-hoc SAEs are trained on frozen representations after the ViT has already been optimized, their latent features are not directly aligned with the downstream classification objective. We introduce SpIn-ViT, a framework that jointly trains a pretrained ViT and a modified SAE end-to-end, directly aligning sparse patch-level representations with image classification. SpIn-ViT learns semantically coherent neuron activations that localize meaningful image regions while maintaining competitive predictive performance. We evaluate SpIn-ViT across nine image-classification benchmarks using classification accuracy, quantitative interpretability metrics, AI-based and Human evaluations. Compared with the previous state-of-the-art post-hoc SAE method, SpIn-ViT achieves 8.84% higher average classification accuracy, an AI-based interpretability score nearly four times as high, and a human-evaluation score more than twice as high. We further extract interpretable rule-sets using the SAE neurons to create neurosymbolic models which achieve 5.97% higher average classification accuracy while requiring a 58.8\% smaller rule-set than the neurosymbolic models created from the SOTA post-hoc SAE method.

Recent work shows that Sparse Autoencoders (SAE) applied to large language model (LLM) layers have neurons corresponding to interpretable concepts. These SAE neurons can be modified to align generated outputs, but only towards pre-identified topics and with some parameter tuning. Our approach leverages the observational and modification properties of SAEs to enable alignment for any topic. This method 1) scores each SAE neuron by its semantic similarity to an alignment text and uses them to 2) modify SAE-layer-level outputs by emphasizing topic-aligned neurons. We assess the alignment capabilities of this approach on diverse public topic datasets including Amazon reviews, Medicine, and Sycophancy, across the currently available open-source LLMs and SAE pairs (GPT2 and Gemma) with multiple SAEs configurations. Experiments aligning to medical prompts reveal several benefits over fine-tuning, including increased average language acceptability (0.25 vs. 0.5), reduced training time across multiple alignment topics (333.6s vs. 62s), and acceptable inference time for many applications (+0.00092s/token). Our open-source code is available at github.com/IBM/sae-steering.

The rapid advancements in transformer-based language models have revolutionized natural language processing, yet understanding the internal mechanisms of these models remains a significant challenge. This paper explores the application of sparse autoencoders (SAE) to interpret the internal representations of protein language models, specifically focusing on the ESM-2 8M parameter model. By performing a statistical analysis on each latent component's relevance to distinct protein annotations, we identify potential interpretations linked to various protein characteristics, including transmembrane regions, binding sites, and specialized motifs. We then leverage these insights to guide sequence generation, shortlisting the relevant latent components that can steer the model towards desired targets such as zinc finger domains. This work contributes to the emerging field of mechanistic interpretability in biological sequence models, offering new perspectives on model steering for sequence design.

Artificial Text Detection (ATD) is becoming increasingly important with the rise of advanced Large Language Models (LLMs). Despite numerous efforts, no single algorithm performs consistently well across different types of unseen text or guarantees effective generalization to new LLMs. Interpretability plays a crucial role in achieving this goal. In this study, we enhance ATD interpretability by using Sparse Autoencoders (SAE) to extract features from Gemma-2-2b residual stream. We identify both interpretable and efficient features, analyzing their semantics and relevance through domain- and model-specific statistics, a steering approach, and manual or LLM-based interpretation. Our methods offer valuable insights into how texts from various models differ from human-written content. We show that modern LLMs have a distinct writing style, especially in information-dense domains, even though they can produce human-like outputs with personalized prompts.

Multilingual language models can solve the same mathematical problem in different languages, but it remains unclear whether they rely on shared features or on language-specific computations that only produce similar outputs. We study this question in five models from four families using the Multilingual Grade School Math (MGSM) dataset, with problems solved in English, German, French, Spanish, Russian, and Chinese, retaining problems with valid reasoning traces in all six languages and replaying those traces through the model to record representations at multiple layers. For each model, we first use Centered Kernel Alignment (CKA) to identify layers with cross-language alignment. At each selected layer, we train two sparse autoencoders (SAE): a baseline reconstruction-only model and a contrastive variant introduced in this work, the Geometry-Invariant SAE (GI-SAE). GI-SAE supplements the reconstruction loss with an Information Noise-Contrastive Estimation (InfoNCE) loss that trains the encoder to produce similar activations for traces of the same problem, regardless of language or token position. We then test whether the resulting shared features are functionally interchangeable by swapping their values between languages during the model's forward pass and measuring the resulting change in output, quantified by Kullback-Leibler (KL) divergence per feature. Although GI-SAE yields higher CKA and Jaccard similarity at nearly every layer, higher geometric similarity does not consistently imply greater functional interchangeability. We find that cross-language feature sharing is model- and architecture-dependent in this sample and appears at different depths in different models. GI-SAE primarily amplifies cross-language structure already present: the pattern is model-specific, with strengthening in Qwen, no functional benefit in Gemma, and mixed layer-dependent effects in Llama and Phi.

Precise control over language model generation is vital for ensuring both safety and reliability. Although prompt engineering and steering are commonly used to intervene in model behaviors, the vast number of parameters in models often results in highly intertwined internal representations. This interdependency can limit control precision and sometimes lead to unintended side effects. Recent research has explored the use of sparse autoencoders (SAE) to disentangle knowledge in high-dimensional spaces for steering. However, these applications have been limited to toy tasks owing to the nontrivial issue of locating atomic knowledge components. In this paper, we propose Steering Target Atoms (STA), a novel method that isolates and manipulates disentangled knowledge components to enhance safety. Comprehensive experiments demonstrate the effectiveness of our approach. Further analysis reveals that steering exhibits superior robustness and flexibility, particularly in adversarial scenarios. We also apply the steering strategy to the large reasoning model, confirming its effectiveness in precise reasoning control.

Detecting prompt injection and jailbreak attacks is critical for deploying LLM-based agents safely. As agents increasingly process untrusted data from emails, documents, tool outputs, and external APIs, robust attack detection becomes essential. Yet current evaluation practices and production systems have fundamental limitations. We present a comprehensive analysis using a diverse benchmark of 18 datasets spanning harmful requests, jailbreaks, indirect prompt injections, and extraction attacks. We propose Leave-One-Dataset-Out (LODO) evaluation to measure true out-of-distribution generalization, revealing that the standard practice of train-test splits from the same dataset sources severely overestimates performance: aggregate metrics show an 8.4 percentage point AUC inflation, but per-dataset gaps range from 1% to 25% accuracy-exposing heterogeneous failure modes. To understand why classifiers fail to generalize, we analyze Sparse Auto-Encoder (SAE) feature coefficients across LODO folds, finding that 28% of top features are dataset-dependent shortcuts whose class signal depends on specific dataset compositions rather than semantic content. We systematically compare production guardrails (PromptGuard 2, LlamaGuard) and LLM-as-judge approaches on our benchmark, finding all three fail on indirect attacks targeting agents (7-37% detection) and that PromptGuard 2 and LlamaGuard cannot evaluate agentic tool injection due to architectural limitations. Finally, we show that LODO-stable SAE features provide more reliable explanations for classifier decisions by filtering dataset artifacts. We release our evaluation framework at https://github.com/maxf-zn/prompt-mining to establish LODO as the appropriate protocol for prompt attack detection research.

Sparse autoencoders (SAEs) promise a unified approach for mechanistic interpretability, concept discovery, and model steering in LLMs and LVLMs. However, realizing this potential requires that the learned features be both interpretable and steerable. To that end, we introduce two new computationally inexpensive interpretability and steerability metrics and conduct a systematic analysis on LVLMs. Our analysis uncovers two observations; (i) a majority of SAE neurons exhibit either low interpretability or low steerability or both, rendering them ineffective for downstream use; and (ii) due to the unsupervised nature of SAEs, user-desired concepts are often absent in the learned dictionary, thus limiting their practical utility. To address these limitations, we propose Concept Bottleneck Sparse Autoencoders (CB-SAE) - a novel post-hoc framework that prunes low-utility neurons and augments the latent space with a lightweight concept bottleneck aligned to a user-defined concept set. The resulting CB-SAE improves interpretability by +32.1% and steerability by +14.5% across LVLMs and image generation tasks. We will make our code and model weights available.

We study how rich visual semantic information is represented within various layers and denoising timesteps of different diffusion architectures. We uncover monosemantic interpretable features by leveraging k-sparse autoencoders (k-SAE). We substantiate our mechanistic interpretations via transfer learning using light-weight classifiers on off-the-shelf diffusion models' features. On 4 datasets, we demonstrate the effectiveness of diffusion features for representation learning. We provide in-depth analysis of how different diffusion architectures, pre-training datasets, and language model conditioning impacts visual representation granularity, inductive biases, and transfer learning capabilities. Our work is a critical step towards deepening interpretability of black-box diffusion models. Code and visualizations available at: https://github.com/revelio-diffusion/revelio

There is growing interest in leveraging mechanistic interpretability and controllability to better understand and influence the internal dynamics of large language models (LLMs). However, current methods face fundamental challenges in reliably localizing and manipulating feature representations. Sparse Autoencoders (SAEs) have recently emerged as a promising direction for feature extraction at scale, yet they, too, are limited by incomplete feature isolation and unreliable monosemanticity. To systematically quantify these limitations, we introduce Feature Monosemanticity Score (FMS), a novel metric to quantify feature monosemanticity in latent representation. Building on these insights, we propose Guided Sparse Autoencoders (G-SAE), a method that conditions latent representations on labeled concepts during training. We demonstrate that reliable localization and disentanglement of target concepts within the latent space improve interpretability, detection of behavior, and control. Specifically, our evaluations on toxicity detection, writing style identification, and privacy attribute recognition show that G-SAE not only enhances monosemanticity but also enables more effective and fine-grained steering with less quality degradation. Our findings provide actionable guidelines for measuring and advancing mechanistic interpretability and control of LLMs.

Learning hierarchical features in Sparse Autoencoders (SAEs) is essential for capturing the structured nature of real-world data and mitigating issues like feature absorption or splitting. Existing works attempt to identify hierarchical relationships within independent feature sets by relying on activation coverage, the assumption that child feature should only activate when its parent feature activates. However, we demonstrate that this condition alone is insufficient; that is, it often produces false positives where parent and child concepts are semantically unrelated. To address this, we introduce a novel reconstruction condition that enforces a deeper functional link between hierarchical levels. By combining both activation and reconstruction constraints, we propose the Tree SAE, a model designed to learn hierarchical structures directly from within the feature set. Our results demonstrate that Tree SAEs significantly surpass the existing SAEs at learning hierarchical pairs while maintaining competitive performance to the state-of-the-art on several key benchmarks. Finally, we demonstrate the practical utility of our Tree SAE in mapping the geometry of child feature subspaces and uncovering the complex hierarchical concept structures encoded within large language models.

Vision-language models (VLMs) have advanced rapidly and are increasingly deployed in real-world applications, especially with the rise of agent-based systems. However, their safety has received relatively limited attention. Even the latest proprietary and open-weight VLMs remain highly vulnerable to adversarial attacks, leaving downstream applications exposed to significant risks. In this work, we propose a novel and lightweight adversarial attack detection framework based on sparse autoencoders (SAEs), termed SAEgis. By inserting an SAE module into a pretrained VLM and training it with standard reconstruction objectives, we find that the learned sparse latent features naturally capture attack-relevant signals. These features enable reliable classification of whether an input image has been adversarially perturbed, even for previously unseen samples. Extensive experiments show that SAEgis achieves strong performance across in-domain, cross-domain, and cross-attack settings, with particularly large improvements in cross-domain generalization compared to existing baselines. In addition, combining signals from multiple layers further improves robustness and stability. To the best of our knowledge, this is the first work to explore SAE as a plug-and-play mechanism for adversarial attack detection in VLMs. Our method requires no additional adversarial training, introduces minimal overhead, and provides a practical approach for improving the safety of real-world VLM systems.

The wide-scale use of sparse autoencoders (SAEs) as interpretability tools is limited by inconsistent links between SAE features and model behavior. Features with clear activation descriptions may have weak or unexpected causal effects; steering can vary across prompts or oppose the intended direction; and activation-based feature selection can miss features that produce the desired output change. Prior work has studied feature geometry inside the model, where features are computed. We instead study the geometry of changes in model logits caused by feature interventions. We introduce Feature-Effect Geometry Analysis (FEGA), an unsupervised framework that removes the same active SAE feature across contexts and analyzes the resulting cloud of logit changes. Across SAE variants, consistent one-dimensional effects are rare: few features behave like reusable directions. To interpret this variation, we distinguish value-like features, tied to static information such as factual attributes, from pointer-like features, associated with context-dependent operations. Value-like features more often exhibit structured, low-dimensional effects, although these effects typically span several directions. Pointer-like features, by contrast, predominantly exhibit diffuse effects. Our results show that a feature can be interpretable and causally relevant without providing a stable direction for steering.

We investigate feature universality in large language models (LLMs), a research field that aims to understand how different models similarly represent concepts in the latent spaces of their intermediate layers. Demonstrating feature universality allows discoveries about latent representations to generalize across several models. However, comparing features across LLMs is challenging due to polysemanticity, in which individual neurons often correspond to multiple features rather than distinct ones. This makes it difficult to disentangle and match features across different models. To address this issue, we employ a method known as dictionary learning by using sparse autoencoders (SAEs) to transform LLM activations into more interpretable spaces spanned by neurons corresponding to individual features. After matching feature neurons across models via activation correlation, we apply representational space similarity metrics like Singular Value Canonical Correlation Analysis to analyze these SAE features across different LLMs. Our experiments reveal significant similarities in SAE feature spaces across various LLMs, providing new evidence for feature universality.

A common goal of mechanistic interpretability is to decompose the activations of neural networks into features: interpretable properties of the input computed by the model. Sparse autoencoders (SAEs) are a popular method for finding these features in LLMs, and it has been postulated that they can be used to find a canonical set of units: a unique and complete list of atomic features. We cast doubt on this belief using two novel techniques: SAE stitching to show they are incomplete, and meta-SAEs to show they are not atomic. SAE stitching involves inserting or swapping latents from a larger SAE into a smaller one. Latents from the larger SAE can be divided into two categories: novel latents, which improve performance when added to the smaller SAE, indicating they capture novel information, and reconstruction latents, which can replace corresponding latents in the smaller SAE that have similar behavior. The existence of novel features indicates incompleteness of smaller SAEs. Using meta-SAEs -- SAEs trained on the decoder matrix of another SAE -- we find that latents in SAEs often decompose into combinations of latents from a smaller SAE, showing that larger SAE latents are not atomic. The resulting decompositions are often interpretable; e.g. a latent representing ``Einstein'' decomposes into ``scientist'', ``Germany'', and ``famous person''. Even if SAEs do not find canonical units of analysis, they may still be useful tools. We suggest that future research should either pursue different approaches for identifying such units, or pragmatically choose the SAE size suited to their task. We provide an interactive dashboard to explore meta-SAEs: https://metasaes.streamlit.app/

We investigate whether sparse autoencoders (SAEs) can be used to remove knowledge from language models. We use the biology subset of the Weapons of Mass Destruction Proxy dataset and test on the gemma-2b-it and gemma-2-2b-it language models. We demonstrate that individual interpretable biology-related SAE features can be used to unlearn a subset of WMDP-Bio questions with minimal side-effects in domains other than biology. Our results suggest that negative scaling of feature activations is necessary and that zero ablating features is ineffective. We find that intervening using multiple SAE features simultaneously can unlearn multiple different topics, but with similar or larger unwanted side-effects than the existing Representation Misdirection for Unlearning technique. Current SAE quality or intervention techniques would need to improve to make SAE-based unlearning comparable to the existing fine-tuning based techniques.

Software vulnerabilities such as buffer overflows and SQL injections are a major source of security breaches. Traditional methods for vulnerability detection remain essential but are limited by high false positive rates, scalability issues, and reliance on manual effort. These constraints have driven interest in AI-based approaches to automated vulnerability detection and secure code generation. While Large Language Models (LLMs) have opened new avenues for classification tasks, their complexity and opacity pose challenges for interpretability and deployment. Sparse Autoencoder offer a promising solution to this problem. We explore whether SAEs can serve as a lightweight, interpretable alternative for bug detection in Java functions. We evaluate the effectiveness of SAEs when applied to representations from GPT-2 Small and Gemma 2B, examining their capacity to highlight buggy behaviour without fine-tuning the underlying LLMs. We found that SAE-derived features enable bug detection with an F1 score of up to 89%, consistently outperforming fine-tuned transformer encoder baselines. Our work provides the first empirical evidence that SAEs can be used to detect software bugs directly from the internal representations of pretrained LLMs, without any fine-tuning or task-specific supervision.

Sparse autoencoders (SAEs) have emerged as a promising approach in language model interpretability, offering unsupervised extraction of sparse features. For interpretability methods to succeed, they must identify abstract features across domains, and these features can often manifest differently in each context. We examine this through "answerability" - a model's ability to recognize answerable questions. We extensively evaluate SAE feature generalization across diverse, partly self-constructed answerability datasets for Gemma 2 SAEs. Our analysis reveals that residual stream probes outperform SAE features within domains, but generalization performance differs sharply. SAE features show inconsistent out-of-domain transfer, with performance varying from almost random to outperforming residual stream probes. Overall, this demonstrates the need for robust evaluation methods and quantitative approaches to predict feature generalization in SAE-based interpretability.

Sparse autoencoders (SAEs) are a useful tool for uncovering human-interpretable features in the activations of large language models (LLMs). While some expect SAEs to find the true underlying features used by a model, our research shows that SAEs trained on the same model and data, differing only in the random seed used to initialize their weights, identify different sets of features. For example, in an SAE with 131K latents trained on a feedforward network in Llama 3 8B, only 30% of the features were shared across different seeds. We observed this phenomenon across multiple layers of three different LLMs, two datasets, and several SAE architectures. While ReLU SAEs trained with the L1 sparsity loss showed greater stability across seeds, SAEs using the state-of-the-art TopK activation function were more seed-dependent, even when controlling for the level of sparsity. Our results suggest that the set of features uncovered by an SAE should be viewed as a pragmatically useful decomposition of activation space, rather than an exhaustive and universal list of features "truly used" by the model.

Sparse autoencoders (SAEs) have emerged as a powerful tool for interpreting language model activations by decomposing them into sparse, interpretable features. A popular approach is the TopK SAE, that uses a fixed number of the most active latents per sample to reconstruct the model activations. We introduce BatchTopK SAEs, a training method that improves upon TopK SAEs by relaxing the top-k constraint to the batch-level, allowing for a variable number of latents to be active per sample. As a result, BatchTopK adaptively allocates more or fewer latents depending on the sample, improving reconstruction without sacrificing average sparsity. We show that BatchTopK SAEs consistently outperform TopK SAEs in reconstructing activations from GPT-2 Small and Gemma 2 2B, and achieve comparable performance to state-of-the-art JumpReLU SAEs. However, an advantage of BatchTopK is that the average number of latents can be directly specified, rather than approximately tuned through a costly hyperparameter sweep. We provide code for training and evaluating BatchTopK SAEs at https://github.com/bartbussmann/BatchTopK

To truly understand vision models, we must not only interpret their learned features but also validate these interpretations through controlled experiments. Current approaches either provide interpretable features without the ability to test their causal influence, or enable model editing without interpretable controls. We present a unified framework using sparse autoencoders (SAEs) that bridges this gap, allowing us to discover human-interpretable visual features and precisely manipulate them to test hypotheses about model behavior. By applying our method to state-of-the-art vision models, we reveal key differences in the semantic abstractions learned by models with different pre-training objectives. We then demonstrate the practical usage of our framework through controlled interventions across multiple vision tasks. We show that SAEs can reliably identify and manipulate interpretable visual features without model re-training, providing a powerful tool for understanding and controlling vision model behavior. We provide code, demos and models on our project website: https://osu-nlp-group.github.io/SAE-V.

Sparse autoencoders (SAEs) are used to decompose neural network activations into human-interpretable features. Typically, features learned by a single SAE are used for downstream applications. However, it has recently been shown that SAEs trained with different initial weights can learn different features, demonstrating that a single SAE captures only a limited subset of features that can be extracted from the activation space. Motivated by this limitation, we propose to ensemble multiple SAEs through naive bagging and boosting. Specifically, SAEs trained with different weight initializations are ensembled in naive bagging, whereas SAEs sequentially trained to minimize the residual error are ensembled in boosting. We evaluate our ensemble approaches with three settings of language models and SAE architectures. Our empirical results demonstrate that ensembling SAEs can improve the reconstruction of language model activations, diversity of features, and SAE stability. Furthermore, ensembling SAEs performs better than applying a single SAE on downstream tasks such as concept detection and spurious correlation removal, showing improved practical utility.

Analyzing large-scale text corpora is a core challenge in machine learning, crucial for tasks like identifying undesirable model behaviors or biases in training data. Current methods often rely on costly LLM-based techniques (e.g. annotating dataset differences) or dense embedding models (e.g. for clustering), which lack control over the properties of interest. We propose using sparse autoencoders (SAEs) to create SAE embeddings: representations whose dimensions map to interpretable concepts. Through four data analysis tasks, we show that SAE embeddings are more cost-effective and reliable than LLMs and more controllable than dense embeddings. Using the large hypothesis space of SAEs, we can uncover insights such as (1) semantic differences between datasets and (2) unexpected concept correlations in documents. For instance, by comparing model responses, we find that Grok-4 clarifies ambiguities more often than nine other frontier models. Relative to LLMs, SAE embeddings uncover bigger differences at 2-8x lower cost and identify biases more reliably. Additionally, SAE embeddings are controllable: by filtering concepts, we can (3) cluster documents along axes of interest and (4) outperform dense embeddings on property-based retrieval. Using SAE embeddings, we study model behavior with two case studies: investigating how OpenAI model behavior has changed over time and finding "trigger" phrases learned by Tulu-3 (Lambert et al., 2024) from its training data. These results position SAEs as a versatile tool for unstructured data analysis and highlight the neglected importance of interpreting models through their data.

Sparse autoencoders (SAEs) are an unsupervised method for learning a sparse decomposition of a neural network's latent representations into seemingly interpretable features. Despite recent excitement about their potential, research applications outside of industry are limited by the high cost of training a comprehensive suite of SAEs. In this work, we introduce Gemma Scope, an open suite of JumpReLU SAEs trained on all layers and sub-layers of Gemma 2 2B and 9B and select layers of Gemma 2 27B base models. We primarily train SAEs on the Gemma 2 pre-trained models, but additionally release SAEs trained on instruction-tuned Gemma 2 9B for comparison. We evaluate the quality of each SAE on standard metrics and release these results. We hope that by releasing these SAE weights, we can help make more ambitious safety and interpretability research easier for the community. Weights and a tutorial can be found at https://huggingface.co/google/gemma-scope and an interactive demo can be found at https://www.neuronpedia.org/gemma-scope

Large language models (LLMs) have demonstrated impressive capabilities in natural language understanding and generation, but controlling their behavior reliably remains challenging, especially in open-ended generation settings. This paper introduces a novel supervised steering approach that operates in sparse, interpretable representation spaces. We employ sparse autoencoders (SAEs)to obtain sparse latent representations that aim to disentangle semantic attributes from model activations. Then we train linear classifiers to identify a small subspace of task-relevant dimensions in latent representations. Finally, we learn supervised steering vectors constrained to this subspace, optimized to align with target behaviors. Experiments across sentiment, truthfulness, and politics polarity steering tasks with multiple LLMs demonstrate that our supervised steering vectors achieve higher success rates with minimal degradation in generation quality compared to existing methods. Further analysis reveals that a notably small subspace is sufficient for effective steering, enabling more targeted and interpretable interventions.

Sparse Autoencoders (SAEs) have emerged as a promising approach for interpreting neural network representations by learning sparse, human-interpretable features from dense activations. We investigate whether incorporating variational methods into SAE architectures can improve feature organization and interpretability. We introduce the Variational Sparse Autoencoder (vSAE), which replaces deterministic ReLU gating with stochastic sampling from learned Gaussian posteriors and incorporates KL divergence regularization toward a standard normal prior. Our hypothesis is that this probabilistic sampling creates dispersive pressure, causing features to organize more coherently in the latent space while avoiding overlap. We evaluate a TopK vSAE against a standard TopK SAE on Pythia-70M transformer residual stream activations using comprehensive benchmarks including SAE Bench, individual feature interpretability analysis, and global latent space visualization through t-SNE. The vSAE underperforms standard SAE across core evaluation metrics, though excels at feature independence and ablation metrics. The KL divergence term creates excessive regularization pressure that substantially reduces the fraction of living features, leading to observed performance degradation. While vSAE features demonstrate improved robustness, they exhibit many more dead features than baseline. Our findings suggest that naive application of variational methods to SAEs does not improve feature organization or interpretability.

A key barrier to interpreting large language models is polysemanticity, where neurons activate for multiple unrelated concepts. Sparse autoencoders (SAEs) have been proposed to mitigate this issue by transforming dense activations into sparse, more interpretable features. While prior work suggests that SAEs promote monosemanticity, no quantitative comparison has examined how concept activation distributions differ between SAEs and their base models. This paper provides the first systematic evaluation of SAEs against base models through activation distribution lens. We introduce a fine-grained concept separability score based on the Jensen-Shannon distance, which captures how distinctly a neuron's activation distributions vary across concepts. Using two large language models (Gemma-2-2B and DeepSeek-R1) and multiple SAE variants across five datasets (including word-level and sentence-level), we show that SAEs reduce polysemanticity and achieve higher concept separability. To assess practical utility, we evaluate concept-level interventions using two strategies: full neuron masking and partial suppression. We find that, compared to base models, SAEs enable more precise concept-level control when using partial suppression. Building on this, we propose Attenuation via Posterior Probabilities (APP), a new intervention method that uses concept-conditioned activation distributions for targeted suppression. APP achieves the smallest perplexity increase while remaining highly effective at concept removal.

Sparse Autoencoders (SAEs) have emerged as a promising tool for interpreting neural networks by decomposing their activations into sparse sets of human-interpretable features. Recent work has introduced multiple SAE variants and successfully scaled them to frontier models. Despite much excitement, a growing number of negative results in downstream tasks casts doubt on whether SAEs recover meaningful features. To directly investigate this, we perform two complementary evaluations. On a synthetic setup with known ground-truth features, we demonstrate that SAEs recover only 9% of true features despite achieving 71% explained variance, showing that they fail at their core task even when reconstruction is strong. To evaluate SAEs on real activations, we introduce three baselines that constrain SAE feature directions or their activation patterns to random values. Through extensive experiments across multiple SAE architectures, we show that our baselines match fully-trained SAEs in interpretability (0.87 vs 0.90), sparse probing (0.69 vs 0.72), and causal editing (0.73 vs 0.72). Together, these results suggest that SAEs in their current state do not reliably decompose models' internal mechanisms.

Sparse autoencoders (SAEs) are widely used to extract interpretable features from neural network representations, often under the implicit assumption that concepts correspond to independent linear directions. However, a growing body of evidence suggests that many concepts are instead organized along low-dimensional manifolds encoding continuous geometric relationships. This raises three basic questions: what does it mean for an SAE to capture a manifold, when do existing SAE architectures do so, and how? We develop a theoretical framework that answers these questions and show that SAEs can capture manifolds in two fundamentally different ways: globally, by allocating a compact group of atoms whose linear span contains the entire manifold, or locally, by distributing it across features that each selectively tile a restricted region of the underlying geometry. Empirically, we find that SAEs suboptimally recover continuous structures, mixing the global subspace and local tiling solutions in a fragmented regime we call dilution. This explains why manifold structure is rarely visible at the level of individual concepts and motivates post-hoc unsupervised discovery methods that search for coherent groups of atoms rather than isolated directions. More broadly, our results suggest that future representation learning methods should treat geometric objects, not just individual directions, as the basic units of interpretability.

Sparse Autoencoders (SAEs) extract interpretable features from Large Language Models, but standard variants enforce non-negativity, forcing separate latents for diametrically opposed concepts (e.g., "pressure too high" vs. "pressure too low") and wasting dictionary capacity when features are anticorrelated. We propose the Sign-Aware Gated SAE (SA-GSAE): two-sided gated sparsity with signed magnitude and auxiliary supervision. A polarity-sensitive gate selects support on either sign, a signed-magnitude path avoids L1 shrinkage, and an auxiliary reconstruction prevents gate collapse. Bipolar sharing - one latent encoding both signs along a shared direction - is realised via a new Bi-Jump-ReLU activation; parameter accounting shows sign-awareness stays parameter-efficient even when anticorrelated pairs are rare. On real LLM activations across three mid-depth hookpoints on Pythia-1B and SmolLM3-3B (6 cells, 3 seeds), a half-width SA-GSAE at width H strictly Pareto-dominates a full-width Gated SAE at 2H over the entire swept L0 overlap on 3 of 6 cells (both MLP-output hookpoints and resid-mid/Pythia-1B); on the remaining 3 it matches R^2 within 0.025 (max gap -0.008) while cutting dead fraction by 0.35-0.62 absolute. Sweep-geomean dead-fraction reductions are ~100x-500x on MLP-output cells and Pythia-1B resid, ~2x-4x on attention cells and SmolLM3-3B resid. Ablations show the two-sided gate and auxiliary loss are load-bearing (no auxiliary collapses LR to 0.27, 98% dead); tying r_i^+ = r_i^- is indistinguishable (|Delta R^2| = 0.0015), and we recommend this symmetric variant as default. MLP-output gains come from most latents carrying both polarities; on attention, bipolar structure concentrates in a small set of top latents. Full-width SA-GSAE exhibits a reproducible reconstruction collapse at SmolLM3-3B resid that the half-width entirely avoids.

Vision-Language-Action (VLA) policies translate language and visual inputs into robot actions, where their hidden representations directly shape closed-loop behavior. However, mechanistic interpretability tools from language and vision-language models do not transfer cleanly to VLAs: outputs are robot actions rather than human-readable tokens, and interventions can only be tested via expensive closed-loop rollouts. We propose an event-grounded interpretability pipeline that anchors SAE feature analysis to behavioral events rather than text contexts. End-effector keyframes are clustered within each task using visual, state, and temporal cues, linking SAE features to behaviorally salient events and, via optional VLM annotations, to semantic context. To our knowledge, our pipeline is among the first to ground SAE-based VLA analysis in closed-loop behavioral events. Across two simulation architectures and a real-robot study, event-grounded ranking yields the strongest causal effects on OpenVLA and transfers to the continuous action chunks of π_{0.5}. SAE is a sparse but imperfect intervention basis: usability varies with architecture and intervention site, and aggressive intervention reveals safety and interpretability limits. Overall, event-grounded SAE analysis emerges as a practical starting point for behavior-anchored VLA interpretability, motivating future work on SAE features beyond action-aligned coordinates, finer-grained closed-loop evaluation, and safe interventions for high-stakes VLA deployments. Code is available at https://github.com/xc-j/Event-SAE.

Sparse Autoencoders (SAEs) have emerged as a promising approach to decompose the activations of Large Language Models (LLMs) into human-interpretable latents. In this paper, we pose two questions. First, to what extent do SAEs extract monosemantic and interpretable latents? Second, to what extent does varying the sparsity or the size of the SAE affect monosemanticity / interpretability? By investigating these questions in the context of a simple first-letter identification task where we have complete access to ground truth labels for all tokens in the vocabulary, we are able to provide more detail than prior investigations. Critically, we identify a problematic form of feature-splitting we call feature absorption where seemingly monosemantic latents fail to fire in cases where they clearly should. Our investigation suggests that varying SAE size or sparsity is insufficient to solve this issue, and that there are deeper conceptual issues in need of resolution.

Sparse autoencoders (SAEs) are useful for detecting and steering interpretable features in neural networks, with particular potential for understanding complex multimodal representations. Given their ability to uncover interpretable features, SAEs are particularly valuable for analyzing large-scale vision-language models (e.g., CLIP and SigLIP), which are fundamental building blocks in modern systems yet remain challenging to interpret and control. However, current SAE methods are limited by optimizing both reconstruction quality and sparsity simultaneously, as they rely on either activation suppression or rigid sparsity constraints. To this end, we introduce Matryoshka SAE (MSAE), a new architecture that learns hierarchical representations at multiple granularities simultaneously, enabling a direct optimization of both metrics without compromise. MSAE establishes a new state-of-the-art Pareto frontier between reconstruction quality and sparsity for CLIP, achieving 0.99 cosine similarity and less than 0.1 fraction of variance unexplained while maintaining ~80% sparsity. Finally, we demonstrate the utility of MSAE as a tool for interpreting and controlling CLIP by extracting over 120 semantic concepts from its representation to perform concept-based similarity search and bias analysis in downstream tasks like CelebA.

Sparse autoencoders (SAEs) are a popular technique for interpreting language model activations, and there is extensive recent work on improving SAE effectiveness. However, most prior work evaluates progress using unsupervised proxy metrics with unclear practical relevance. We introduce SAEBench, a comprehensive evaluation suite that measures SAE performance across seven diverse metrics, spanning interpretability, feature disentanglement and practical applications like unlearning. To enable systematic comparison, we open-source a suite of over 200 SAEs across eight recently proposed SAE architectures and training algorithms. Our evaluation reveals that gains on proxy metrics do not reliably translate to better practical performance. For instance, while Matryoshka SAEs slightly underperform on existing proxy metrics, they substantially outperform other architectures on feature disentanglement metrics; moreover, this advantage grows with SAE scale. By providing a standardized framework for measuring progress in SAE development, SAEBench enables researchers to study scaling trends and make nuanced comparisons between different SAE architectures and training methodologies. Our interactive interface enables researchers to flexibly visualize relationships between metrics across hundreds of open-source SAEs at: https://saebench.xyz

As large language models (LLMs) grow in scale and capability, understanding their internal mechanisms becomes increasingly critical. Sparse autoencoders (SAEs) have emerged as a key tool in mechanistic interpretability, enabling the extraction of human-interpretable features from LLMs. However, existing SAE training methods are primarily designed for base models, resulting in reduced reconstruction quality and interpretability when applied to instruct models. To bridge this gap, we propose textbf{F}inetuning-textbf{a}ligned textbf{S}equential textbf{T}raining (FAST), a novel training method specifically tailored for instruct models. FAST aligns the training process with the data distribution and activation patterns characteristic of instruct models, resulting in substantial improvements in both reconstruction and feature interpretability. On Qwen2.5-7B-Instruct, FAST achieves a mean squared error of 0.6468 in token reconstruction, significantly outperforming baseline methods with errors of 5.1985 and 1.5096. In feature interpretability, FAST yields a higher proportion of high-quality features, for Llama3.2-3B-Instruct, 21.1% scored in the top range, compared to 7.0% and 10.2% for BT(P) and BT(F). Surprisingly, we discover that intervening on the activations of special tokens via the SAEs leads to improvements in output quality, suggesting new opportunities for fine-grained control of model behavior. Code, data, and 240 trained SAEs are available at https://github.com/Geaming2002/FAST.

Improving Sparse Autoencoders (SAEs) requires benchmarks that can precisely validate architectural innovations. However, current SAE benchmarks on LLMs are often too noisy to differentiate architectural improvements, and current synthetic data experiments are too small-scale and unrealistic to provide meaningful comparisons. We introduce SynthSAEBench, a toolkit for generating large-scale synthetic data with realistic feature characteristics including correlation, hierarchy, and superposition, and a standardized benchmark model, SynthSAEBench-16k, enabling direct comparison of SAE architectures. Our benchmark reproduces several previously observed LLM SAE phenomena, including the disconnect between reconstruction and latent quality metrics, poor SAE probing results, and a precision-recall trade-off mediated by L0. We further use our benchmark to identify a new failure mode: Matching Pursuit SAEs exploit superposition noise to improve reconstruction without learning ground-truth features, suggesting that more expressive encoders can easily overfit. SynthSAEBench complements LLM benchmarks by providing ground-truth features and controlled ablations, enabling researchers to precisely diagnose SAE failure modes and validate architectural improvements before scaling to LLMs.

Sparse autoencoders (SAEs) have lately been used to uncover interpretable latent features in large language models. Here, we explore their potential for decomposing latent representations in complex and high-dimensional biological data, where the underlying variables are often unknown. On simulated data we show that generative hidden variables can be captured in learned representations in the form of superpositions. The degree to which they are learned depends on the completeness of the representations. Superpositions, however, are not identifiable if these generative variables are unknown. SAEs can to some extent recover these variables, yielding interpretable features. Applied to single-cell multi-omics data, we show that an SAE can uncover key biological processes such as carbon dioxide transport and ion homeostasis, which are crucial for red blood cell differentiation and immune function. Our findings highlight how SAEs can be used in advancing interpretability in biological and other scientific domains.

Sparse autoencoders (SAEs) have gained a lot of attention as a promising tool to improve the interpretability of large language models (LLMs) by mapping the complex superposition of polysemantic neurons into monosemantic features and composing a sparse dictionary of words. However, traditional performance metrics like Mean Squared Error and L0 sparsity ignore the evaluation of the semantic representational power of SAEs -- whether they can acquire interpretable monosemantic features while preserving the semantic relationship of words. For instance, it is not obvious whether a learned sparse feature could distinguish different meanings in one word. In this paper, we propose a suite of evaluations for SAEs to analyze the quality of monosemantic features by focusing on polysemous words. Our findings reveal that SAEs developed to improve the MSE-L0 Pareto frontier may confuse interpretability, which does not necessarily enhance the extraction of monosemantic features. The analysis of SAEs with polysemous words can also figure out the internal mechanism of LLMs; deeper layers and the Attention module contribute to distinguishing polysemy in a word. Our semantics focused evaluation offers new insights into the polysemy and the existing SAE objective and contributes to the development of more practical SAEs.

Sparse Autoencoders (SAEs) have emerged as a promising solution for decomposing large language model representations into interpretable features. However, Paulo and Belrose (2025) have highlighted instability across different initialization seeds, and Heap et al. (2025) have pointed out that SAEs may not capture model-internal features. These problems likely stem from training SAEs on external datasets - either collected from the Web or generated by another model - which may contain out-of-distribution (OOD) data beyond the model's generalisation capabilities. This can result in hallucinated SAE features, which we term "Fake Features", that misrepresent the model's internal activations. To address these issues, we propose FaithfulSAE, a method that trains SAEs on the model's own synthetic dataset. Using FaithfulSAEs, we demonstrate that training SAEs on less-OOD instruction datasets results in SAEs being more stable across seeds. Notably, FaithfulSAEs outperform SAEs trained on web-based datasets in the SAE probing task and exhibit a lower Fake Feature Ratio in 5 out of 7 models. Overall, our approach eliminates the dependency on external datasets, advancing interpretability by better capturing model-internal features while highlighting the often neglected importance of SAE training datasets.

As large language models (LLMs) are increasingly deployed in real-world applications, the need to selectively remove unwanted knowledge while preserving model utility has become paramount. Recent work has explored sparse autoencoders (SAEs) to perform precise interventions on monosemantic features. However, most SAE-based methods operate at inference time, which does not create persistent changes in the model's parameters. Such interventions can be bypassed or reversed by malicious actors with parameter access. We introduce CRISP, a parameter-efficient method for persistent concept unlearning using SAEs. CRISP automatically identifies salient SAE features across multiple layers and suppresses their activations. We experiment with two LLMs and show that our method outperforms prior approaches on safety-critical unlearning tasks from the WMDP benchmark, successfully removing harmful knowledge while preserving general and in-domain capabilities. Feature-level analysis reveals that CRISP achieves semantically coherent separation between target and benign concepts, allowing precise suppression of the target features.

Sparse Autoencoders (SAEs) have emerged as a popular tool for interpreting the hidden states of large language models (LLMs). By learning to reconstruct activations from a sparse bottleneck layer, SAEs discover interpretable features from the high-dimensional internal representations of LLMs. Despite their popularity with language models, SAEs remain understudied in the visual domain. In this work, we provide an extensive evaluation the representational power of SAEs for vision models using a broad range of image-based tasks. Our experimental results demonstrate that SAE features are semantically meaningful, improve out-of-distribution generalization, and enable controllable generation across three vision model architectures: vision embedding models, multi-modal LMMs and diffusion models. In vision embedding models, we find that learned SAE features can be used for OOD detection and provide evidence that they recover the ontological structure of the underlying model. For diffusion models, we demonstrate that SAEs enable semantic steering through text encoder manipulation and develop an automated pipeline for discovering human-interpretable attributes. Finally, we conduct exploratory experiments on multi-modal LLMs, finding evidence that SAE features reveal shared representations across vision and language modalities. Our study provides a foundation for SAE evaluation in vision models, highlighting their strong potential improving interpretability, generalization, and steerability in the visual domain.

Sparse autoencoders (SAEs) are a technique for sparse decomposition of neural network activations into human-interpretable features. However, current SAEs suffer from feature absorption, where specialized features capture instances of general features creating representation holes, and feature composition, where independent features merge into composite representations. In this work, we introduce Orthogonal SAE (OrtSAE), a novel approach aimed to mitigate these issues by enforcing orthogonality between the learned features. By implementing a new training procedure that penalizes high pairwise cosine similarity between SAE features, OrtSAE promotes the development of disentangled features while scaling linearly with the SAE size, avoiding significant computational overhead. We train OrtSAE across different models and layers and compare it with other methods. We find that OrtSAE discovers 9% more distinct features, reduces feature absorption (by 65%) and composition (by 15%), improves performance on spurious correlation removal (+6%), and achieves on-par performance for other downstream tasks compared to traditional SAEs.

Sparse autoencoders (SAEs) have emerged as a promising method for interpreting neural network representations by decomposing activations into sparse combinations of dictionary atoms. However, SAEs assume that features combine additively through linear reconstruction, an assumption that cannot capture compositional structure: linear models cannot distinguish whether "Starbucks" arises from the composition of "star" and "coffee" features or merely their co-occurrence. This forces SAEs to allocate monolithic features for compound concepts rather than decomposing them into interpretable constituents. We introduce PolySAE, which extends the SAE decoder with higher-order terms to model feature interactions while preserving the linear encoder essential for interpretability. Through low-rank tensor factorization on a shared projection subspace, PolySAE captures pairwise and triple feature interactions with small parameter overhead (3% on GPT2). Across four language models and three SAE variants, PolySAE achieves an average improvement of approximately 8% in probing F1 while maintaining comparable reconstruction error, and produces 2-10times larger Wasserstein distances between class-conditional feature distributions. Critically, learned interaction weights exhibit negligible correlation with co-occurrence frequency (r = 0.06 vs. r = 0.82 for SAE feature covariance), suggesting that polynomial terms capture compositional structure, such as morphological binding and phrasal composition, largely independent of surface statistics.

Sparse autoencoders (SAEs) are typically trained on hundreds of millions to billions of activation tokens, yet little work has tested whether this scale is necessary. We study how much activation data SAEs need by training four SAE architectures on three language models with budgets ranging from 5M tokens to a 500M-token reference. We measure two notions of convergence: functional behavior on standard evaluations and one-to-one reproducibility of learned features. Functional data requirements are strongly metric-dependent: reconstruction often approaches the full-data reference at substantially smaller budgets, whereas feature reproducibility continues to improve with data. For Gemma-2-2B BatchTopK, the fraction of strictly matched features rises from 2\% at 5M tokens to 29\% at 50M. Data selection also matters: random sampling better preserves the original activation distribution and generally favors reconstruction, whereas diversity-based sampling frequently improves AutoInterp. A theoretical analysis separates statistical consistency of subsampled training from the stronger identifiability and optimization conditions required for one-to-one feature recovery. Our results show that SAE training can often be made far more data-efficient when budgets are matched to the goal, and feature-level claims are backed by explicit reproducibility checks, paving the way toward cheaper and more reliable SAE research.

Disentangling model activations into meaningful features is a central problem in interpretability. However, the absence of ground-truth for these features in realistic scenarios makes validating recent approaches, such as sparse dictionary learning, elusive. To address this challenge, we propose a framework for evaluating feature dictionaries in the context of specific tasks, by comparing them against supervised feature dictionaries. First, we demonstrate that supervised dictionaries achieve excellent approximation, control, and interpretability of model computations on the task. Second, we use the supervised dictionaries to develop and contextualize evaluations of unsupervised dictionaries along the same three axes. We apply this framework to the indirect object identification (IOI) task using GPT-2 Small, with sparse autoencoders (SAEs) trained on either the IOI or OpenWebText datasets. We find that these SAEs capture interpretable features for the IOI task, but they are less successful than supervised features in controlling the model. Finally, we observe two qualitative phenomena in SAE training: feature occlusion (where a causally relevant concept is robustly overshadowed by even slightly higher-magnitude ones in the learned features), and feature over-splitting (where binary features split into many smaller, less interpretable features). We hope that our framework will provide a useful step towards more objective and grounded evaluations of sparse dictionary learning methods.

Steering vectors are a promising approach to control the behaviour of large language models. However, their underlying mechanisms remain poorly understood. While sparse autoencoders (SAEs) may offer a potential method to interpret steering vectors, recent findings show that SAE-reconstructed vectors often lack the steering properties of the original vectors. This paper investigates why directly applying SAEs to steering vectors yields misleading decompositions, identifying two reasons: (1) steering vectors fall outside the input distribution for which SAEs are designed, and (2) steering vectors can have meaningful negative projections in feature directions, which SAEs are not designed to accommodate. These limitations hinder the direct use of SAEs for interpreting steering vectors.

We study the challenge of achieving theoretically grounded feature recovery using Sparse Autoencoders (SAEs) for the interpretation of Large Language Models. Existing SAE training algorithms often lack rigorous mathematical guarantees and suffer from practical limitations such as hyperparameter sensitivity and instability. To address these issues, we first propose a novel statistical framework for the feature recovery problem, which includes a new notion of feature identifiability by modeling polysemantic features as sparse mixtures of underlying monosemantic concepts. Building on this framework, we introduce a new SAE training algorithm based on ``bias adaptation'', a technique that adaptively adjusts neural network bias parameters to ensure appropriate activation sparsity. We theoretically prove that this algorithm correctly recovers all monosemantic features when input data is sampled from our proposed statistical model. Furthermore, we develop an improved empirical variant, Group Bias Adaptation (GBA), and demonstrate its superior performance against benchmark methods when applied to LLMs with up to 1.5 billion parameters. This work represents a foundational step in demystifying SAE training by providing the first SAE algorithm with theoretical recovery guarantees, thereby advancing the development of more transparent and trustworthy AI systems through enhanced mechanistic interpretability.

Interpretability can improve the safety, transparency and trust of AI models, which is especially important in healthcare applications where decisions often carry significant consequences. Mechanistic interpretability, particularly through the use of sparse autoencoders (SAEs), offers a promising approach for uncovering human-interpretable features within large transformer-based models. In this study, we apply Matryoshka-SAE to the radiology-specialised multimodal large language model, MAIRA-2, to interpret its internal representations. Using large-scale automated interpretability of the SAE features, we identify a range of clinically relevant concepts - including medical devices (e.g., line and tube placements, pacemaker presence), pathologies such as pleural effusion and cardiomegaly, longitudinal changes and textual features. We further examine the influence of these features on model behaviour through steering, demonstrating directional control over generations with mixed success. Our results reveal practical and methodological challenges, yet they offer initial insights into the internal concepts learned by MAIRA-2 - marking a step toward deeper mechanistic understanding and interpretability of a radiology-adapted multimodal large language model, and paving the way for improved model transparency. We release the trained SAEs and interpretations: https://huggingface.co/microsoft/maira-2-sae.

Sparse autoencoders (SAEs) are a popular method for interpreting concepts represented in large language model (LLM) activations. However, there is a lack of evidence regarding the validity of their interpretations due to the lack of a ground truth for the concepts used by an LLM, and a growing number of works have presented problems with current SAEs. One alternative source of evidence would be demonstrating that SAEs improve performance on downstream tasks beyond existing baselines. We test this by applying SAEs to the real-world task of LLM activation probing in four regimes: data scarcity, class imbalance, label noise, and covariate shift. Due to the difficulty of detecting concepts in these challenging settings, we hypothesize that SAEs' basis of interpretable, concept-level latents should provide a useful inductive bias. However, although SAEs occasionally perform better than baselines on individual datasets, we are unable to design ensemble methods combining SAEs with baselines that consistently outperform ensemble methods solely using baselines. Additionally, although SAEs initially appear promising for identifying spurious correlations, detecting poor dataset quality, and training multi-token probes, we are able to achieve similar results with simple non-SAE baselines as well. Though we cannot discount SAEs' utility on other tasks, our findings highlight the shortcomings of current SAEs and the need to rigorously evaluate interpretability methods on downstream tasks with strong baselines.

Language models increasingly serve as the backbone of text-to-speech (TTS) systems, yet we understand little about the representations they build when text and generated speech tokens share a single residual stream. We train BatchTopK sparse autoencoders on the LM backbone of CosyVoice3 and introduce a modality-aware auto-interp pipeline that labels each feature from where it fires-text-prefix context, 1-second speech clips, or both. The recovered features are interpretable, spanning phonemes, laughter, accent prompts and speaker gender. Steering through the SAE latent space shows these features are causal rather than merely descriptive: targeted interventions raise laughter probability from 0.02 to 0.79, flip perceived speaker gender, and control speech rate while preserving spoken content. SAE features thus serve both as interpretability objects and as control directions for TTS synthesis.

Generative models have shown remarkable progress in a variety of domains such as protein design, but such power enables the opaque generation of hazardous proteins. In this work, we introduce VFUSE (Virulent Feature Understanding with Sparse autoEncoders), a mechanistic interpretability approach that trains SAEs on diffusion-transformer activations to audit protein models for hazard-aware features. We apply VFUSE to RoseTTAFold3 and RFDiffusion3, popular open-weight models for protein folding and synthesis. We find that for certain blocks, linear probes detect hazardous designs significantly better when fit in the SAE latent space over the original model's representations: improving interpretability without sacrificing model performance. Furthermore, we identify monosemantic features from the SAE that fire only on hazardous designs at up to AUROC 0.84 (q < 10^{-13}). To our knowledge this is the first SAE trained on an all-atom diffusion model and the first feature-level virulence audit of a protein design model, paving the way towards safe and interpretable protein design.

Sparse autoencoders (SAEs) are a promising technique for decomposing language model activations into interpretable linear features. However, current SAEs fall short of completely explaining model performance, resulting in "dark matter": unexplained variance in activations. This work investigates dark matter as an object of study in its own right. Surprisingly, we find that much of SAE dark matter--about half of the error vector itself and >90% of its norm--can be linearly predicted from the initial activation vector. Additionally, we find that the scaling behavior of SAE error norms at a per token level is remarkably predictable: larger SAEs mostly struggle to reconstruct the same contexts as smaller SAEs. We build on the linear representation hypothesis to propose models of activations that might lead to these observations, including postulating a new type of "introduced error"; these insights imply that the part of the SAE error vector that cannot be linearly predicted ("nonlinear" error) might be fundamentally different from the linearly predictable component. To validate this hypothesis, we empirically analyze nonlinear SAE error and show that 1) it contains fewer not yet learned features, 2) SAEs trained on it are quantitatively worse, 3) it helps predict SAE per-token scaling behavior, and 4) it is responsible for a proportional amount of the downstream increase in cross entropy loss when SAE activations are inserted into the model. Finally, we examine two methods to reduce nonlinear SAE error at a fixed sparsity: inference time gradient pursuit, which leads to a very slight decrease in nonlinear error, and linear transformations from earlier layer SAE outputs, which leads to a larger reduction.

Sparse autoencoders (SAEs) are a promising approach to extracting features from neural networks, enabling model interpretability as well as causal interventions on model internals. SAEs generate sparse feature representations using a sparsifying activation function that implicitly defines a set of token-feature matches. We frame the token-feature matching as a resource allocation problem constrained by a total sparsity upper bound. For example, TopK SAEs solve this allocation problem with the additional constraint that each token matches with at most k features. In TopK SAEs, the k active features per token constraint is the same across tokens, despite some tokens being more difficult to reconstruct than others. To address this limitation, we propose two novel SAE variants, Feature Choice SAEs and Mutual Choice SAEs, which each allow for a variable number of active features per token. Feature Choice SAEs solve the sparsity allocation problem under the additional constraint that each feature matches with at most m tokens. Mutual Choice SAEs solve the unrestricted allocation problem where the total sparsity budget can be allocated freely between tokens and features. Additionally, we introduce a new auxiliary loss function, aux_zipf_loss, which generalises the aux_k_loss to mitigate dead and underutilised features. Our methods result in SAEs with fewer dead features and improved reconstruction loss at equivalent sparsity levels as a result of the inherent adaptive computation. More accurate and scalable feature extraction methods provide a path towards better understanding and more precise control of foundation models.

Despite their impressive performance, generative image models trained on large-scale datasets frequently fail to produce images with seemingly simple concepts -- e.g., human hands or objects appearing in groups of four -- that are reasonably expected to appear in the training data. These failure modes have largely been documented anecdotally, leaving open the question of whether they reflect idiosyncratic anomalies or more structural limitations of these models. To address this, we introduce a systematic approach for identifying and characterizing "conceptual blindspots" -- concepts present in the training data but absent or misrepresented in a model's generations. Our method leverages sparse autoencoders (SAEs) to extract interpretable concept embeddings, enabling a quantitative comparison of concept prevalence between real and generated images. We train an archetypal SAE (RA-SAE) on DINOv2 features with 32,000 concepts -- the largest such SAE to date -- enabling fine-grained analysis of conceptual disparities. Applied to four popular generative models (Stable Diffusion 1.5/2.1, PixArt, and Kandinsky), our approach reveals specific suppressed blindspots (e.g., bird feeders, DVD discs, and whitespaces on documents) and exaggerated blindspots (e.g., wood background texture and palm trees). At the individual datapoint level, we further isolate memorization artifacts -- instances where models reproduce highly specific visual templates seen during training. Overall, we propose a theoretically grounded framework for systematically identifying conceptual blindspots in generative models by assessing their conceptual fidelity with respect to the underlying data-generating process.

Sparse autoencoders (SAEs) have emerged as a powerful tool for interpreting neural networks by extracting the concepts represented in their activations. However, choosing the size of the SAE dictionary (i.e. number of learned concepts) creates a tension: as dictionary size increases to capture more relevant concepts, sparsity incentivizes features to be split or absorbed into more specific features, leaving high-level features missing or warped. We introduce Matryoshka SAEs, a novel variant that addresses these issues by simultaneously training multiple nested dictionaries of increasing size, forcing the smaller dictionaries to independently reconstruct the inputs without using the larger dictionaries. This organizes features hierarchically - the smaller dictionaries learn general concepts, while the larger dictionaries learn more specific concepts, without incentive to absorb the high-level features. We train Matryoshka SAEs on Gemma-2-2B and TinyStories and find superior performance on sparse probing and targeted concept erasure tasks, more disentangled concept representations, and reduced feature absorption. While there is a minor tradeoff with reconstruction performance, we believe Matryoshka SAEs are a superior alternative for practical tasks, as they enable training arbitrarily large SAEs while retaining interpretable features at different levels of abstraction.

Large language models (LLMs) are now ubiquitous in user-facing applications, yet they still generate undesirable toxic outputs, including profanity, vulgarity, and derogatory remarks. Although numerous detoxification methods exist, most apply broad, surface-level fixes and can therefore easily be circumvented by jailbreak attacks. In this paper we leverage sparse autoencoders (SAEs) to identify toxicity-related directions in the residual stream of models and perform targeted activation steering using the corresponding decoder vectors. We introduce three tiers of steering aggressiveness and evaluate them on GPT-2 Small and Gemma-2-2B, revealing trade-offs between toxicity reduction and language fluency. At stronger steering strengths, these causal interventions surpass competitive baselines in reducing toxicity by up to 20%, though fluency can degrade noticeably on GPT-2 Small depending on the aggressiveness. Crucially, standard NLP benchmark scores upon steering remain stable, indicating that the model's knowledge and general abilities are preserved. We further show that feature-splitting in wider SAEs hampers safety interventions, underscoring the importance of disentangled feature learning. Our findings highlight both the promise and the current limitations of SAE-based causal interventions for LLM detoxification, further suggesting practical guidelines for safer language-model deployment.

Sparse Autoencoders (SAEs) can efficiently identify candidate monosemantic features from pretrained neural networks for galaxy morphology. We demonstrate this on Euclid Q1 images using both supervised (Zoobot) and new self-supervised (MAE) models. Our publicly released MAE achieves superhuman image reconstruction performance. While a Principal Component Analysis (PCA) on the supervised model primarily identifies features already aligned with the Galaxy Zoo decision tree, SAEs can identify interpretable features outside of this framework. SAE features also show stronger alignment than PCA with Galaxy Zoo labels. Although challenges in interpretability remain, SAEs provide a powerful engine for discovering astrophysical phenomena beyond the confines of human-defined classification.

Sparse Autoencoders (SAEs) are powerful tools for interpreting neural representations, yet their use in audio remains underexplored. We train SAEs across all encoder layers of Whisper and HuBERT, provide an extensive evaluation of their stability, interpretability, and show their practical utility. Over 50% of the features remain consistent across random seeds, and reconstruction quality is preserved. SAE features capture general acoustic and semantic information as well as specific events, including environmental noises and paralinguistic sounds (e.g. laughter, whispering) and disentangle them effectively, requiring removal of only 19-27% of features to erase a concept. Feature steering reduces Whisper's false speech detections by 70% with negligible WER increase, demonstrating real-world applicability. Finally, we find SAE features correlated with human EEG activity during speech perception, indicating alignment with human neural processing. The code and checkpoints are available at https://github.com/audiosae/audiosae_demo.