Study finds SAE unstable features reflect reproducible subspaces, not pure noise
A new arXiv paper investigates feature stability in sparse autoencoders (SAEs), measuring the probability that individual learned features reappear across independent training runs. The authors find a functional asymmetry: stable features carry most reconstruction-relevant signal, while unstable features are individually non-reproducible but concentrate in reproducible lower-rank subspaces, suggesting seed dependence reflects basis ambiguity rather than noise. A synthetic model confirms that low-rank ground-truth features can be recovered at the subspace level even when individual SAE latents are non-identifiable across seeds. The work has direct implications for interpretability research that relies on SAE features as meaningful, stable units of analysis.
Related guides (2)
Related events (8)
SAEs as Stethoscopes: Interpretability-Guided Layer Selection for Task Vector Model Editing
This paper evaluates a Sparse Autoencoder (SAE)-guided model editing pipeline for mathematical reasoning on Gemma-3-4B-IT, finding that projecting task vectors onto SAE feature subspaces discards ~97% of modification energy due to geometric misalignment between activation-space SAE directions and weight-space task vectors. The authors reframe SAEs as diagnostic tools ('stethoscopes') rather than intervention filters ('scalpels'), using SAE-derived specificity scores to identify which layers to inject unfiltered task vectors into. This approach improves Number Theory accuracy from 29.6% to 39.4% on Minerva Math (p=0.0007), with 5 of 7 math subjects significantly improved and none degraded. The method is fully deterministic and adds no inference cost.
SAERL: Using Sparse Autoencoders to Guide LLM Reinforcement Learning Data Engineering
SAERL is a post-training data engineering framework that uses Sparse Autoencoders (SAEs) — a mechanistic interpretability tool — to extract intrinsic model signals for controlling data diversity, difficulty, and quality during RL fine-tuning. The framework applies SAE-space clustering for batch diversity, a difficulty proxy for curriculum ordering, and a quality probe for data filtering. On Qwen2.5-Math-1.5B with GRPO, SAERL achieves 3% average accuracy improvement and reaches target accuracy with 20% fewer training steps. SAE representations transfer across model families and scales, suggesting broad applicability as a lightweight data engineering tool.
Sparse AutoEncoder steering reduces Whisper hallucination rate by ~5x without fine-tuning
Researchers investigate hallucination detection and mitigation in OpenAI's Whisper ASR model by probing internal encoder representations. They find that both raw activations and Sparse AutoEncoder (SAE) latents encode linearly separable hallucination signals concentrated in deeper layers. SAE-based activation steering reduces hallucination rates from 72.6% to 14.1% (Whisper small) and 86.9% to 27.3% (Whisper large-v3) on non-speech audio, with minimal WER degradation, approaching fine-tuning-level performance without weight updates.
Stable Recovery Manifold hypothesis: catastrophic forgetting as accessibility problem, not information destruction
A new arXiv preprint investigates the geometric structure of recoverability in continual learning using Split CIFAR-100 and a sequentially trained ResNet-18. The authors introduce Recovery Subspace Dimensionality (k_t) and find that recovery dimensionality remains stable across tasks (mean k_t = 8.0) despite substantial representational drift, with principal-angle drift strongly predicting recoverability (r = -0.862). The findings support the Stable Recovery Manifold hypothesis: forgotten knowledge remains compactly decodable, reframing catastrophic forgetting as a manifold-alignment and accessibility problem rather than true information loss.
Extracting Concepts from GPT-4: 16 Million Patterns via Sparse Autoencoders
OpenAI applied scaled sparse autoencoders (SAEs) to GPT-4 to automatically identify approximately 16 million interpretable features or patterns in the model's internal computations. This represents a significant scaling of mechanistic interpretability techniques previously demonstrated on smaller models. The work advances the ability to understand what concepts and representations large frontier models encode internally.
Probing bioacoustic embeddings for speech-like acoustic features reveals no-free-lunch pattern
A new arXiv preprint investigates which acoustic features are encoded in pretrained bioacoustic audio embeddings using 88 eGeMAPS speech features across six taxonomic groups. Linear and nonlinear regression probes reveal that no single model captures the full acoustic feature space, with loudness best recovered (R²=0.76) and fundamental frequency hardest (R²=0.33). A concatenated embedding approach achieves highest overall performance, suggesting complementary coverage across models. The work provides data-driven model selection guidance for bioacoustics tasks involving rare species or low-resource domains.
Explainability pipeline reveals divergent cues used by deepfake speech detectors
Researchers propose an audio-native explainability pipeline using Integrated Gradients on time-aligned self-supervised representations to localize decision evidence in deepfake speech detectors. Applied to three WavLM-based detectors (AASIST, CA-MHFA, SLS) on the ASVspoof 5 benchmark, the method reveals that despite similar performance, each detector relies on fundamentally different cues: environmental noise, phoneme artifacts, and word boundaries respectively. Findings are validated via causal masking experiments that confirm performance degrades when primary cues are removed. The work advances interpretability of audio deepfake detection, relevant to AI safety and media authenticity.
SPEX and ProxySPEX: Scalable Interaction Discovery for LLM Interpretability
Researchers from BAIR introduce SPEX (Spectral Explainer) and ProxySPEX, algorithms for identifying influential feature, data, and model-component interactions in LLMs at scale. The approach exploits sparsity, low-degreeness, and hierarchy properties to reframe interaction discovery as a sparse recovery problem using tools from signal processing and coding theory. ProxySPEX achieves comparable performance to SPEX with roughly 10x fewer ablations by leveraging hierarchical structure. The methods are evaluated on feature attribution (sentiment analysis), data attribution, and mechanistic interpretability tasks, outperforming marginal methods like LIME at long context lengths.