Identifying the features learned by neural networks is a core challenge in mechanistic interpretability. Sparse autoencoders (SAEs), which learn a sparse, overcomplete dictionary that reconstructs a network's internal activations, have been used to identify these features. However, SAEs may learn more about the structure of the dataset than the computational structure of the network. There is therefore only indirect reason to believe that the directions found in these dictionaries are functionally important to the network. We propose end-to-end (e2e) sparse dictionary learning, a method for training SAEs that ensures the features learned are functionally important by minimizing the KL divergence between the output distributions of the original model and the model with SAE activations inserted.

Since P@k treats all the labels equally, it doesn't reveal the performance of the model on tail labels. Amazon-3M using only 4 GPUs as compared to XR-Transformer which leverages 8 GPUs. Table 10: Time taken to train the ensembles of the respective models. X-Transformer and XR-Transformer have been reported using 8 NVidia V100 GPUs. Thus, following XR-Transformer's lead we combine the features trained by CasadeXML with DiSMEC across datasets (Table: 1), we find DiSMEC to be more resource efficient than XR-Linear.

Identifying the features learned by neural networks is a core challenge in mechanistic interpretability. Sparse autoencoders (SAEs), which learn a sparse, overcomplete dictionary that reconstructs a network's internal activations, have been used to identify these features. However, SAEs may learn more about the structure of the dataset than the computational structure of the network. There is therefore only indirect reason to believe that the directions found in these dictionaries are functionally important to the network. We propose end-to-end (e2e) sparse dictionary learning, a method for training SAEs that ensures the features learned are functionally important by minimizing the KL divergence between the output distributions of the original model and the model with SAE activations inserted.

The rapid advancement and widespread adoption of generative artificial intelligence (AI) pose significant threats to the integrity of personal identity, including digital cloning, sophisticated impersonation, and the unauthorized monetization of identity-related data. Mitigating these risks necessitates the development of robust AI-generated content detection systems, enhanced legal frameworks, and ethical guidelines. This paper introduces the Digital Identity Rights Framework (DIRF), a structured security and governance model designed to protect behavioral, biometric, and personality-based digital likeness attributes to address this critical need. Structured across nine domains and 63 controls, DIRF integrates legal, technical, and hybrid enforcement mechanisms to secure digital identity consent, traceability, and monetization. We present the architectural foundations, enforcement strategies, and key use cases supporting the need for a unified framework. This work aims to inform platform builders, legal entities, and regulators about the essential controls needed to enforce identity rights in AI-driven systems.

Quantum computers offer a promising route to tackling problems that are classically intractable such as in prime-factorization, solving large-scale linear algebra and simulating complex quantum systems, but require fault-tolerant quantum hardware. On the other hand, variational quantum algorithms (VQAs) have the potential to provide a near-term route to quantum utility or advantage, and is usually constructed by using parametrized quantum circuits (PQCs) in combination with a classical optimizer for training. Although VQAs have been proposed for a multitude of tasks such as ground-state estimation, combinatorial optimization and unitary compilation, there remain major challenges in its trainability and resource costs on quantum hardware. Here we address these challenges by adopting Hardware Efficient and dynamical LIe algebra Supported Ansatz (HELIA), and propose two training schemes that combine an existing g-sim method (that uses the underlying group structure of the operators) and the Parameter-Shift Rule (PSR). Our improvement comes from distributing the resources required for gradient estimation and training to both classical and quantum hardware. We numerically test our proposal for ground-state estimation using Variational Quantum Eigensolver (VQE) and classification of quantum phases using quantum neural networks. Our methods show better accuracy and success of trials, and also need fewer calls to the quantum hardware on an average than using only PSR (upto 60% reduction), that runs exclusively on quantum hardware. We also numerically demonstrate the capability of HELIA in mitigating barren plateaus, paving the way for training large-scale quantum models.

Sensitive directions experiments attempt to understand the computational features of Language Models (LMs) by measuring how much the next token prediction probabilities change by perturbing activations along specific directions. We extend the sensitive directions work by introducing an improved baseline for perturbation directions. We demonstrate that KL divergence for Sparse Autoencoder (SAE) reconstruction errors are no longer pathologically high compared to the improved baseline. We also show that feature directions uncovered by SAEs have varying impacts on model outputs depending on the SAE's sparsity, with lower L0 SAE feature directions exerting a greater influence. Additionally, we find that end-to-end SAE features do not exhibit stronger effects on model outputs compared to traditional SAEs.

Neural Information Processing Systems http://nips.cc/

Identifying the features learned by neural networks is a core challenge in mechanistic interpretability. Sparse autoencoders (SAEs), which learn a sparse, overcomplete dictionary that reconstructs a network's internal activations, have been used to identify these features. However, SAEs may learn more about the structure of the datatset than the computational structure of the network. There is therefore only indirect reason to believe that the directions found in these dictionaries are functionally important to the network. We propose end-to-end (e2e) sparse dictionary learning, a method for training SAEs that ensures the features learned are functionally important by minimizing the KL divergence between the output distributions of the original model and the model with SAE activations inserted. Compared to standard SAEs, e2e SAEs offer a Pareto improvement: They explain more network performance, require fewer total features, and require fewer simultaneously active features per datapoint, all with no cost to interpretability. We explore geometric and qualitative differences between e2e SAE features and standard SAE features. E2e dictionary learning brings us closer to methods that can explain network behavior concisely and accurately.

In this work, we introduce Semantic Pyramid AutoEncoder (SPAE) for enabling frozen LLMs to perform both understanding and generation tasks involving non-linguistic modalities such as images or videos. SPAE converts between raw pixels and interpretable lexical tokens (or words) extracted from the LLM's vocabulary. The resulting tokens capture both the semantic meaning and the fine-grained details needed for visual reconstruction, effectively translating the visual content into a language comprehensible to the LLM, and empowering it to perform a wide array of multimodal tasks. Our approach is validated through in-context learning experiments with frozen PaLM 2 and GPT 3.5 on a diverse set of image understanding and generation tasks. Our method marks the first successful attempt to enable a frozen LLM to generate image content while surpassing state-of-the-art performance in image understanding tasks, under the same setting, by over 25%.