Publications

(2024). Randomized Asymmetric Chain of LoRA: The First Meaningful Theoretical Framework for Low-Rank Adaptation. Featured Theory.

PDF

(2024). Model Merging and Safety Alignment: One Bad Model Spoils the Bunch. EMNLP 2024.

PDF

(2024). Look Listen and Attack: Backdoor Attacks Against Video Action Recognition. CVPRW 2024.

PDF

(2024). Towards Interpretable Deep Local Learning with Successive Gradient Reconciliation. ICML 2024.

PDF

(2024). On Pretraining Data Diversity for Self-Supervised Learning. ECCV 2024.

PDF

(2024). SynthCLIP: Are We Ready for a Fully Synthetic CLIP Training?. SynthCLIP.

PDF

(2023). CAMEL: Communicative Agents for "Mind" Exploration of Large Scale Language Model Society. In NeurIPS23.

PDF Code

(2023). Rapid Adaptation in Online Continual Learning: Are We Evaluating It Right?. in ICCV23.

PDF Code

(2023). Mindstorms in Natural Language-Based Societies of Mind. Arxiv.

PDF

(2023). Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs. In CVPRW23.

PDF

(2023). Real-Time Evaluation in Online Continual Learning: A New Hope. Highlight in CVPR23.

PDF Code Poster Video

(2023). Generalizability of Adversarial Robustness Under Distribution Shifts. Featured Certification.

PDF

(2023). Computationally Budgeted Continual Learning: What Does Matter?. In CVPR23.

PDF Code Poster Video

(2023). Large eddy simulations of NH3-H2 jet flame at elevated pressure using PCA with inclusion of NH3/H2 ratio variation. In AIAA SCITECH 2023 Forum.

PDF

(2023). Large eddy simulations of ammonia-hydrogen jet flames at elevated pressure using principal component analysis and deep neural networks. In AIAA SCITECH 2023 Forum.

PDF

(2022). On the Decision Boundaries of Neural Networks: A Tropical Geometry Perspective. In PAMI22.

PDF

(2022). PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies. in NeurIPS22.

PDF Code

(2022). Check Your Other Door! Creating Backdoor Attacks in the Frequency Domain. In BMVC2022.

PDF

(2021). ASSANet: An Anisotropic Separable Set Abstraction for Efficient Point Cloud Representation Learning. in NeurIPS21.

PDF Code