🏢 National University of Singapore

X-Ray: A novel 3D representation generating complete object surfaces from a single image!

VCR-GauS: Novel view-consistent depth-normal regularizer for superior, real-time 3D surface reconstruction using Gaussian splatting.

Self-attention, a key component of transformers, is revealed to be a projection of query vectors onto the principal components of the key matrix, derived from kernel PCA. This novel perspective leads…

Uncertainty of Thoughts (UoT) algorithm significantly boosts LLMs’ information-seeking abilities, leading to substantial performance gains across diverse tasks.

Stochastic Taylor Derivative Estimator (STDE) drastically accelerates the optimization of neural networks involving high-dimensional, high-order differential operators by efficiently amortizing comput…

SpeedLoader: A groundbreaking I/O efficient scheme dramatically boosts LLM training & inference speed on diverse hardware, even with limited resources!

SparseLinUCB: First sparse regret bounds for adversarial action sets with unknown sparsity, achieving superior performance over existing methods!

SHORE: a novel two-stage framework efficiently solves sparse & high-dimensional output regression, boosting interpretability and scalability.

SlimSAM achieves near original SAM performance using 0.1% of its training data by employing a novel alternate slimming framework and disturbed Taylor pruning, significantly advancing data-efficient mo…

This paper introduces a novel, tractable semidefinite program (SDP) relaxation for the Gromov-Wasserstein distance, enabling the computation of globally optimal transportation plans.

RFLPA: Secure Federated Learning resists poisoning attacks via efficient secure aggregation.

SCHOOL: A novel SHGL framework enhancing spectral clustering with rank and dual consistency constraints, effectively mitigating noise and leveraging cluster-level information for improved downstream t…

Remix-DiT: Boosting diffusion model image generation quality by cleverly mixing smaller basis models into numerous specialized denoisers, improving efficiency and lowering costs!

A novel inventory reserving algorithm achieves near-optimal performance for bandit problems with knapsacks in piecewise-stationary settings, offering a competitive ratio of O(log(nmax/min)).

Tri-BBAI & Opt-BBAI achieve optimal asymptotic and near-optimal non-asymptotic sample & batch complexities in batched best arm identification.

OnlineTAS, a novel framework, achieves state-of-the-art performance in online temporal action segmentation by using an adaptive memory and a post-processing method to mitigate over-segmentation.

FedSD2C, a novel one-shot federated learning framework, tackles data heterogeneity and information loss by sharing synthetic distillates directly from local data, outperforming existing methods on com…

One-Sample-Fits-All (OFA) framework efficiently approximates all probabilistic values simultaneously, achieving faster convergence rates than existing methods.

Softmax-DPO boosts LM-based recommender performance by directly optimizing for personalized ranking using a novel loss function that incorporates multiple negative samples, significantly outperforming…

MVSDet uses efficient plane sweeps for accurate indoor 3D object detection from multiple images, significantly outperforming previous NeRF-based methods.