Posters

ReBorn revitalizes multi-agent reinforcement learning by tackling dormant neurons, boosting network expressivity and learning efficiency.

Softmax regression reveals in-context learning’s surprising similarity to gradient descent in self-attention Transformers, showing the models’ remarkable learning capabilities.

Physics-Informed Neural Networks (PINNs) training dynamics for nonlinear PDEs are fundamentally different than linear ones; this paper reveals why using second-order methods is crucial for solving non…

Researchers found that superior OOD detection performance comes at the cost of reduced generalization. Their novel Decoupled Uncertainty Learning (DUL) algorithm harmonizes OOD detection and generali…

Data balancing in foundation models surprisingly reduces variance, improving model training and performance.

Diverse inhibitory neurons in brain circuits enable faster Bayesian computation via Hamiltonian sampling.

TFS-NeRF: A template-free neural radiance field efficiently reconstructs semantically separable 3D geometries of dynamic scenes featuring multiple interacting entities from sparse RGB videos.

TFGDA: Leveraging graph topology and feature alignment for superior semi-supervised domain adaptation.

Hassle-Free Textual Training (HFTT) uses only textual data to effectively remove unwanted visual data from AI training datasets, significantly reducing human annotation needs.

Text2NKG: a novel framework for building N-ary relational knowledge graphs by performing fine-grained n-ary relation extraction, supporting multiple schemas, and achieving state-of-the-art accuracy.

Ti-FAD: a novel zero-shot temporal action detection model outperforms state-of-the-art methods by enhancing text-related visual focus and foreground awareness.

Text-Guided Attention for Zero-Shot Robustness (TGA-ZSR) significantly improves vision-language model robustness against adversarial attacks by aligning and constraining text-guided attention, achievi…

Text-Aware Diffusion for Policy Learning (TADPoLe) uses pretrained diffusion models for zero-shot reward generation, enabling natural language-driven policy learning without manual reward design.

This work presents statistically grounded methods to rank semantic concept importance in black-box models, using conditional independence testing for both global and local interpretations.

This paper presents nearly-linear time algorithms for testing model calibration, improving upon existing methods and providing theoretical lower bounds for various calibration measures.

Testably learning polynomial threshold functions efficiently, matching agnostic learning’s best guarantees, is achieved, solving a key problem in robust machine learning.

Test-Time Dynamic Image Fusion (TTD) paradigm provably improves image fusion by dynamically weighting source data based on their relative dominance, reducing generalization error without extra trainin…

Test-time adaptation strengthens the linear correlation between in- and out-of-distribution accuracy, enabling precise OOD performance prediction and hyperparameter optimization without labeled OOD da…

Ada-ReAlign: a novel algorithm for continual test-time adaptation that leverages non-stationary representation learning to effectively align unlabeled data streams with source data, enhancing model ad…

Prioritize crucial decisions in deep RL policy testing with a novel model-based method for rigorous state importance ranking, enabling efficient safety and performance verification.