Posters

Unrolled neural networks, trained via gradient descent, provably achieve optimal Bayesian inference for compressed sensing, surpassing prior-aware counterparts.

Strategic data contributors with varying data quality can cause collaborative learning systems to ‘unravel’, but a novel probabilistic verification method effectively mitigates this, ensuring a stable…

This paper reveals how large embedding dimensions and appropriate initialization guarantee convergence in Transformer training, highlighting Gaussian attention’s superior landscape over Softmax.

This study disentangles best practices for learning from preference feedback in LLMs, revealing that data quality, algorithm choice, and reward model significantly impact performance.

AI can unlock the potential of global human expertise for solving complex problems by combining diverse expert solutions using an evolutionary framework, resulting in better and more effective strateg…

Self-guidance boosts masked generative models’ image synthesis, achieving superior quality and diversity with fewer steps!

MARS, a novel plugin framework, unlocks model-free RL’s intermittent control ability by encoding action sequences into a compact latent space, improving learning efficiency and real-world robotic task…

DiffewS: a novel framework leverages diffusion models for few-shot semantic segmentation, significantly outperforming existing methods in multiple settings.

ProjDiff: A novel algorithm unleashes diffusion models’ denoising power for superior inverse problem solutions.

Unlocking intermediate layers in MLLMs improves referring expression generation by enhancing accuracy and detail while reducing hallucinations.

New MVUAV dataset and SemiMV semi-supervised learning model significantly improve multispectral video semantic segmentation!

New unlearnable framework secures 3D point cloud data by using class-wise transformations, enabling authorized training while preventing unauthorized access.

Adaptive gradient methods using AdaGrad stepsizes achieve optimal convergence rates for convex composite optimization problems, handling inexact oracles, acceleration, and variance reduction without n…

Linear model transfer learning achieves universal generalization error improvements, depending only on first and second-order target statistics, and defying Gaussian assumptions.

Universal Sample Coding revolutionizes data transmission by reducing bits needed to communicate multiple samples from an unknown distribution, achieving significant improvements in federated learning …

Active learning’s optimal rates are completely characterized, resolving an open problem and providing new algorithms achieving exponential and sublinear rates depending on combinatorial complexity mea…

Universal Physics Transformers (UPTs) offer a unified, scalable framework for efficiently training neural operators across diverse spatio-temporal physics problems, overcoming limitations of existing …

This paper introduces a novel universal online convex optimization algorithm needing only one projection per round, achieving optimal regret bounds for various function types, including general convex…

Universal Neural Functionals (UNFs) automatically construct permutation-equivariant models for any weight space, improving learned optimizer performance and generalization.

Fully recurrent neural networks can be universal in-context approximators, achieving the same capabilities as transformer models by cleverly using prompts.