Posters

Unified Graph Augmentations (UGA) module boosts graph contrastive learning by unifying diverse augmentation strategies, improving model generalizability and efficiency.

Unified generative-discriminative training boosts multimodal large language models (MLLMs)! Sugar, a novel approach, leverages dynamic sequence alignment and a triple kernel to enhance global and fin…

Unified Domain Generalization and Adaptation (UDGA) tackles 3D object detection’s domain adaptation challenges by leveraging multi-view overlap and label-efficient learning, achieving state-of-the-art…

Unified Covariate Adjustment (UCA) offers a scalable, doubly robust estimator for a wide array of causal estimands beyond standard methods.

UniDSeg uses Visual Foundation Models to create a unified framework for adaptable and generalizable cross-domain 3D semantic segmentation, achieving state-of-the-art results.

UniBias unveils and mitigates LLM bias by identifying and eliminating biased internal components (FFN vectors and attention heads), significantly improving in-context learning performance and robustne…

UniAudio 1.5 uses a novel LLM-driven audio codec to enable frozen LLMs to perform various audio tasks with just a few examples, opening new avenues for efficient few-shot cross-modal learning.

UniAR: A unified model predicts human attention and preferences across diverse visual content (images, webpages, designs), achieving state-of-the-art performance and enabling human-centric improvement…

Uni-Med, a novel unified medical foundation model, tackles multi-task learning challenges by using Connector-MoE to efficiently bridge modalities, achieving competitive performance across six medical …

Researchers unveil unelicitable backdoors in language models, using cryptographic transformer circuits, defying conventional detection methods and raising crucial AI safety concerns.

Deep learning models favor simple features, hindering generalization; this paper introduces a new feature complexity metric revealing a spectrum of simple-to-complex features, their learning dynamics,…

LLMs’ inner workings remain elusive. This study uses N-gram statistics to approximate transformer predictions, revealing how LLMs learn from simple to complex statistical rules, and how model variance…

Transformers excel at graph reasoning, with logarithmic depth proving necessary and sufficient for parallelizable tasks; single-layer transformers solve retrieval tasks efficiently.

E-SSL’s generalization ability is rigorously analyzed via an information-theoretic lens, revealing key design principles for improved performance.

Vision-language models struggle with multi-object reasoning due to the binding problem; this paper reveals human-like capacity limits in VLMs and proposes solutions.

Repeated self-distillation significantly reduces excess risk in linear regression, achieving up to a ’d’ factor improvement over single-step methods.

A mean-field theory explains Fourier Neural Operator (FNO) behavior, linking expressivity to trainability by identifying ordered and chaotic phases that correspond to vanishing or exploding gradients,…

This work systematically investigates the approximation properties of Transformer networks for sequence modeling, revealing the distinct roles of key components (self-attention, positional encoding, f…

Unifying framework reveals hidden connections between attention, recurrent, and state-space models, boosting foundation model efficiency.

Deep learning scaling laws are explained by novel approximation and estimation theories for transformers on low-dimensional data, resolving discrepancies between theory and practice.