Tag
This paper proves that task-relevant latent representations can be identified from generalist models in a fully nonparametric setting without interventions or parametric constraints, achieving a hierarchical identifiability guarantee across time steps and within each step.
The article argues that the primary AI risk may not be superintelligence but rather systems that optimize flawed, incomplete representations of reality, leading to institutional drift, automated misclassification, and invisible governance failures.
This paper proposes Spectra, a method using spectral occupancy to analyze and control the realized capacity of latent graph models, arguing that rank is not equivalent to model capacity.
This paper introduces SimReg, a regularization technique for LLM pretraining that uses embedding similarity to improve training convergence by over 30% and boost zero-shot performance.
This paper introduces the DR.Q algorithm, which improves model-based representations for Q-learning by maximizing mutual information and using faded prioritized experience replay to reduce bias and overfitting in continuous control tasks.
This paper introduces DRoRAE, a method that improves visual tokenization by fusing multi-layer features from pretrained vision encoders rather than relying solely on the last layer. It demonstrates significant improvements in reconstruction and generation quality on ImageNet and establishes a scaling law between fusion capacity and performance.
This paper addresses the issue of dimensional collapse in VQ-VAEs, showing that representations often occupy a low-dimensional subspace. It proposes an 'AE Warm-Up' strategy that trains the model as an unquantized autoencoder first, which improves reconstruction quality and increases effective latent dimensionality.
This paper introduces Data-Driven Variational Basis Learning (DVBL), a non-neural framework that learns basis functions directly from data through variational optimization, offering interpretability and mathematical transparency compared to neural networks.
This paper introduces TabEmbed, a generalist embedding model for tabular data that unifies classification and retrieval tasks, along with TabBench, a new benchmark for evaluating tabular understanding.
Introduces CGM-JEPA, a self-supervised pretraining framework for continuous glucose monitor data that improves cross-modal and cross-cohort performance through masked latent prediction and distributional objectives.
Researchers introduce GeoRepEval, a framework to evaluate LLM robustness across equivalent geometric problem representations (Euclidean, coordinate, vector). Testing 11 LLMs on 158 geometry problems, they find accuracy gaps up to 14 percentage points based solely on representation choice, with vector formulations being a consistent failure point.
This paper introduces HSG (Hyperbolic Scene Graph), a scene graph model that leverages hyperbolic geometry for representing hierarchical scene structures. It is hosted on Hugging Face and referenced via arXiv:2604.17454.
This paper introduces MMOT, an online mixture model learning framework based on optimal transport theory that addresses incremental learning with distributional shifts through dynamic centroid updates and improved class similarity estimation. The approach includes a Dynamic Preservation strategy to mitigate catastrophic forgetting and maintain class separability in latent space.
Google DeepMind published a paper in Nature detailing a method to align AI visual representations with human cognitive structures, improving model robustness and reliability.
OpenAI researchers propose a general framework for learning representations of agent policies in multiagent systems using minimal interaction data, casting the problem as representation learning with applications to competitive control and cooperative communication environments.
OpenAI researchers present a Variational Lossy Autoencoder (VLAE) that combines VAEs with neural autoregressive models (RNN, MADE, PixelRNN/CNN) to learn controllable global representations, achieving state-of-the-art results on MNIST, OMNIGLOT, and Caltech-101 Silhouettes density estimation tasks.