6 accepted papers
Popular offline reinforcement learning (RL) methods rely on conservatism, penalizing out-of-dataset actions or restricting rollout horizons. We question the universality of this principle and revisit a complementary Bayesian perspective. By modeling a posterior over plausible world models and traini…
Adapting a pretrained diffusion model to new objectives at inference time remains an open problem in generative modeling. Existing steering methods suffer from inaccurate value estimation, especially at high noise levels, which biases guidance. Moreover, information from past runs is not reused to i…
Large language models (LLMs) are increasingly used across research and industry applications, yet their inference efficiency remains a significant challenge. As the computational power of modern GPU architectures continuously improves, their memory bandwidth and capacity have not scaled proportional…
We present a novel perspective on goal-conditioned reinforcement learning by framing it within the context of denoising diffusion models. Analogous to the diffusion process, where Gaussian noise is used to create random trajectories that walk away from the data manifold, we construct trajectories th…
Known for their impressive performance in generative modeling, diffusion models are attractive candidates for density-based anomaly detection. This paper investigates different variations of diffusion modeling for unsupervised and semi-supervised anomaly detection. In particular, we find that Denois…
We study a variety of notions of equivariance as an inductive bias in Reinforcement Learning (RL). In particular, we propose new mechanisms for learning representations that are equivariant to both the agent’s action, as well as symmetry transformations of the state-action pairs. Whereas prior work…