🏢 Tsinghua University

SCD-Tree & GBD: Unlocking interpretable rules for unsupervised anomaly detection!

Diff-Tuning: a simple yet effective approach transfers pre-trained diffusion models to various downstream tasks by leveraging the ‘chain of forgetting’ phenomenon, improving transferability and conver…

Diffusion models are certifiably robust classifiers due to their inherent O(1) Lipschitzness, a property further enhanced by generalizing to noisy data, achieving over 80% certified robustness on CIFA…

DACER, a novel online RL algorithm, uses diffusion models to learn complex policies and adaptively balances exploration-exploitation via entropy estimation, achieving state-of-the-art performance on M…

DI-MaskDINO: Novel model significantly boosts object detection & instance segmentation accuracy by addressing performance imbalance using a De-Imbalance module and Balance-Aware Tokens Optimization.

Boosting multimodal LLMs, the Dense Connector efficiently integrates multi-layer visual features for significantly enhanced performance.

Vision’s Mamba model demystified: Researchers unveil its surprising link to linear attention, improving efficiency and accuracy through design enhancements.

DeformableTST: a new Transformer model for time series forecasting that surpasses existing methods by reducing over-reliance on patching, enhancing performance and adaptability.

DeepLag improves fluid prediction by uniquely combining Lagrangian and Eulerian perspectives, tracking key particles to reveal hidden dynamics and improve prediction accuracy.

Multi-objective decoding (MOD) efficiently aligns language models to diverse user needs by decoding the next token from a weighted combination of predictions from multiple base models trained on indiv…

DDN: Dual-domain Dynamic Normalization dynamically improves time series forecasting accuracy by addressing data distribution changes in both time and frequency domains via a plug-in module.

DART-Math tackles LLM limitations in mathematical problem-solving by introducing Difficulty-Aware Rejection Tuning, a novel method that generates high-quality, bias-reduced datasets, resulting in supe…

DARNet: a dual attention network for auditory attention detection surpasses current state-of-the-art models, especially in short decision windows, achieving this with a 91% reduction in parameters.

COVE: Consistent high-quality video editing achieved by leveraging diffusion feature correspondence for temporal consistency.

COSMIC efficiently compresses satellite images via a lightweight encoder and diffusion compensation, enabling practical onboard processing and high compression ratios.

CooHOI: A two-phase learning framework enables physically simulated characters to perform cooperative object transportation tasks naturally and efficiently, overcoming the limitations of existing meth…

This paper introduces induced empirical risk minimization for contextual linear optimization with bandit feedback, providing theoretical guarantees and computationally tractable solutions for improved…

Consistency Diffusion Bridge Models (CDBMs) dramatically speed up diffusion bridge model sampling by learning a consistency function, achieving up to a 50x speedup with improved sample quality.

Language models struggle to learn facts; this study reveals they prioritize word co-occurrence over true factual associations, proposing new training strategies for improved factual knowledge generali…

CLOVER: A closed-loop visuomotor framework using generative visual plans & feedback mechanisms achieves state-of-the-art results in long-horizon robotic manipulation tasks.