The paper discusses how large language models retrieve facts from their weights, proposing that this phenomenon reflects a broader operation termed feature recall. It argues that a linear projection can be interpreted as retrieving stored information scaled by input activations, and demonstrates that feature recall applies across various architectures, contrasting it with the traditional feature combination paradigm. The authors also explore potential mechanistic identification of feature recall cases and suggest new empirical directions for mechanistic interpretability research.
arXiv:2606. 06624v1 Announce Type: new Abstract: In the current era of deep learning and especially generative models, there is significant investment in training very large generative models.
arXiv:2606. 29951v1 Announce Type: new Abstract: Interpretable Mesomorphic Neural Networks (IMNs) offer a promising framework that combines the predictive power of deep neural networks with the interpretability of linear models.
arXiv:2411. 18343v3 Announce Type: replace Abstract: Personalized Federal learning(PFL) allows clients to cooperatively train a personalized model without disclosing their private dataset.
arXiv:2605. 28969v2 Announce Type: replace-cross Abstract: If an AI agent makes decisions on a person's behalf, those decisions must align with its user.
Post-training quantization (PTQ) has become a practical solution for deploying deep learning models on resource-constrained edge devices by compressing high-precision floating-point weights into low-precision representations without requiring retraining. Past research has demonstrated that quantization largely preserves classification accuracy; however, whether it also preserves the model's internal reasoning remains an open question.
arXiv:2607. 22872v1 Announce Type: new Abstract: Post-training quantization (PTQ) has become a practical solution for deploying deep learning models on resource-constrained edge devices by compressing high-precision floating-point weights into low-precision representations without requiring retraining.
arXiv:2607. 00089v1 Announce Type: new Abstract: Mechanistic interpretability has produced a rich inventory of component-level analyses that characterise what neural-network components encode and how they interact.
arXiv:2609.15064v1 Announce Type: new Abstract: Reinforcement learning (RL) is widely utilized in large language model training to improve targeted capabilities, yet how RL reshapes a model remains p...
arXiv:2608.29034v1 Announce Type: cross Abstract: A wide range of methods have been proposed for interpreting language models, delivering important insights into their inner workings. However, differ...
arXiv:2603. 04198v2 Announce Type: replace-cross Abstract: Sparse autoencoders (SAEs) are widely used to extract human-interpretable features from neural network activations, but their learned features can vary substantially across random seeds and training choices.
arXiv:2606. 07998v1 Announce Type: cross Abstract: Recent advances in generative AI, especially powerful Large Language Models (LLMs) and Large Reasoning Models (LRMs), raise concerns over the interpretability, safety and sustainability of these large and opaque AI models.
arXiv:2609.15533v1 Announce Type: cross Abstract: Mechanistic interpretability has become the dominant lens for understanding frontier language models, as their inner workings are complex and inheren...