Posters

Adaptive Dense-to-sparse Constrained Optimization (ADC) efficiently jailbreaks LLMs by transforming discrete token optimization into a continuous process, achieving higher success rates than existing …

Sort & Search: 1000x faster lifelong model evaluation!

LLaVolta significantly boosts multi-modal LLMs by using visual context compression, achieving substantial training cost reduction and enhanced inference efficiency without performance loss.

Efficient algorithm achieves near-perfect graph matching in correlated stochastic block models, resolving a key open problem and enabling improved community detection.

FedAWE, a novel federated learning algorithm, efficiently handles intermittent and unpredictable client availability, ensuring fast and unbiased model training.

Provably efficient algorithms for learning with distribution shift are introduced, generalizing and improving prior work by achieving near-optimal error rates and offering universal learners for large…

TREACLE: a reinforcement learning policy efficiently selects LLMs and prompts, achieving up to 85% cost savings while maintaining high accuracy in answering reasoning questions.

Heat Diffusion Optimization (HeO) framework efficiently solves combinatorial optimization problems by enabling information propagation through heat diffusion, outperforming existing methods.

New attacks foil both supervised and contrastive learning, achieving state-of-the-art unlearnability with less computation.

This paper proposes a novel, statistically efficient offline policy evaluation method robust to environmental shifts and unobserved confounding, providing sharp bounds with theoretical guarantees.

Efficiently and privately reconstructing marginal queries from noisy data using residuals improves accuracy of existing differential privacy mechanisms.

Boosting Vision Transformer adaptation! Householder Transformation-based Adaptor (HTA) outperforms existing methods by dynamically adjusting adaptation matrix ranks across layers, improving efficiency…

New efficient algorithms minimize regret in extensive-form games by cleverly using low-degree swap deviations and a relaxed fixed-point concept, improving correlated equilibrium computation.

First-price auction efficiency in autobidding plummets to 45.7% with mixed bidders, but machine-learned advice restores optimality.

RELA accelerates representation learning by leveraging freely available pre-trained models to generate efficient data, reducing computational costs by up to 50% while maintaining accuracy.

Effective rank regularization enhances 3D Gaussian splatting, resolving needle-like artifacts and improving 3D model quality.

SI2E, a novel RL exploration framework, leverages structural information principles to maximize value-conditional structural entropy, significantly outperforming state-of-the-art baselines in various …

EEGPT: A pretrained transformer model revolutionizes EEG signal representation by using a dual self-supervised learning method, achieving state-of-the-art results across various tasks.

EEG2Video reconstructs dynamic videos from EEG signals, achieving 79.8% accuracy in semantic classification and 0.256 SSIM in video reconstruction.

The Efficient Diffusion Transformer (EDT) framework significantly speeds up and improves image generation by leveraging a lightweight architecture, human-like sketching-inspired Attention Modulation M…