🏢 Google DeepMind

LLMs’ success is explained by the self-similar, long-range dependent fractal structure of language; small-scale patterns reflect larger ones.

First oracle-free, computationally tractable algorithms for provably convergent multivariate distributional RL are introduced, achieving convergence rates matching scalar settings and offering insight…

FlexCap generates controllable, region-specific image descriptions of varying lengths, achieving state-of-the-art zero-shot visual question answering.

FineStyle enables fine-grained controllable style personalization for text-to-image models using a novel concept-oriented data scaling and parameter-efficient adapter tuning, mitigating content leakag…

Fast Tree-Field Integrators (FTFIs) revolutionize graph processing by enabling polylog-linear time computation for integrating tensor fields on trees, providing significant speedups for various machin…

EM Distillation (EMD) efficiently trains one-step diffusion models by using an Expectation-Maximization approach, achieving state-of-the-art image generation quality and outperforming existing methods…

Novel sketching algorithms enable scalable gradient and Hessian analysis for large language models, revealing insights into their intrinsic dimensionality and challenging existing assumptions.

Researchers developed semantic RSMs, a novel approach to measure semantic similarity in neural networks, improving image retrieval and aligning network representations with predicted class probabiliti…

LLMs can reason effectively without prompting by simply adjusting the decoding process to reveal inherent chain-of-thought paths.

CAT3D: Generate high-quality 3D scenes from as little as one image using a novel multi-view diffusion model, outperforming existing methods in speed and quality.

Large-scale transformers achieve grandmaster-level chess play via supervised learning on a new 10M game benchmark dataset, demonstrating impressive generalization beyond memorization.