Google DeepMind

Efficiently tracking a small, switching subset of experts in continuous action spaces is provably harder than theoretically possible, with polynomial-time algorithms facing inherent regret limitations.

Optimistic FTRL achieves linear convergence to Nash equilibria in harmonic games, a significant improvement over prior sublinear bounds.

High sample efficiency in learning-to-rank can be achieved without the headaches of custom gradients, making RL more accessible for practitioners.

MXSens redefines LLM quantization by achieving state-of-the-art accuracy with mixed precision, leveraging sensitivity-aware bitwidth allocation.

RLNS turns a classic heuristic into a powerful MCMC sampler, enabling efficient combinatorial optimization without the need for exact solutions.

Forget hand-engineered safety constraints: MARL agents trained for high-speed quadrotor racing not only beat human champions, but also cut collision rates in half.

Dynamic quantization, a widely adopted optimization for efficient ML serving, can leak your data to adversaries sharing the same batch.

Generative AI evaluation can be sped up by 8-65x without sacrificing accuracy by proactively focusing on the most informative test cases using a pre-trained Gaussian Process surrogate model.

DPP-based Monte Carlo integration can offer variance reduction, but choosing the right DPP—fixed vs. tailored to the integrand—determines whether you get a biased but faster converging estimator or an unbiased but standard-rate estimator.

Entropy regularization makes planning provably easy: SmoothCruiser achieves polynomial sample complexity in MDPs where standard methods fail.

Forget sub-Gaussian assumptions: this semi-bandit algorithm adapts to the true covariance structure of outcomes, leading to tighter regret bounds and better performance.

Refining generative models with discriminator guidance provably improves generalization, offering a theoretical justification for techniques like score-based diffusion.

Unlock asymptotically normal and semiparametrically efficient estimators in adaptive data collection by using a novel target-specific condition called "directional stability," which is weaker than previous target-agnostic conditions.

Robots can now learn long-horizon tasks far more effectively by distilling complex histories into a few key visual moments, outperforming standard imitation learning by 70% on real-world tasks.

Ditch reward models: Nash Mirror Prox achieves fast, stable convergence to a Nash equilibrium directly from human preferences, sidestepping the limitations of traditional RLHF.