Research
Recent publications and preprints, auto-sorted. For the latest updates, check out Google Scholar
2025
-
-
Answer Matching Outperforms Multiple Choice for Language Model Evaluationarxiv:2507.02856[cs], Jul 2025
-
Can You Finetune Your Binoculars? Embedding Text Watermarks into the Weights of Large Language Modelsarxiv:2504.06446[cs], Apr 2025
-
Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approacharxiv:2502.05171[cs], Feb 2025
-
-
LLM-Generated Passphrases That Are Secure and Easy to RememberIn The 2025 Annual Conference of the Nations of the Americas Chapter of the ACL, Jan 2025
-
-
-
Can Language Models Falsify? Evaluating Algorithmic Reasoning with Counterexample Creationarxiv:2502.19414[cs], Feb 2025
-
Is Your Batch Size the Problem? Revisiting the Adam-SGD Gap in Language Modelingarxiv:2506.12543[cs], Jun 2025
-
Fine, I’ll Merge It Myself: A Multi-Fidelity Framework for Automated Model Mergingarxiv:2502.04030[cs], Feb 2025
-
GPTailor: Large Language Model Pruning Through Layer Cutting and Stitchingarxiv:2506.20480[cs], Jun 2025
2024
-
-
-
Be like a Goldfish, Don’t Memorize! Mitigating Memorization in Generative LLMsIn The Thirty-eighth Annual Conference on Neural Information Processing Systems, Sep 2024
-
Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated TextIn Proceedings of the Forty-first International Conference on Machine Learning, Jan 2024
-
Bring Your Own Data! Self-Sensitivity Evaluation for Large Language ModelsIn First Conference on Language Modeling, Aug 2024
-
-
-
Transformers Can Do Arithmetic with the Right EmbeddingsIn The Thirty-eighth Annual Conference on Neural Information Processing Systems, Sep 2024
-
-
-
-
Democratizing AI: Open-source Scalable LLM Training on GPU-based SupercomputersIn 2024 SC24: International Conference for High Performance Computing, Networking, Storage and Analysis SC, Nov 2024
-
CALVIN: Improved Contextual Video Captioning via Instruction TuningIn The Thirty-eighth Annual Conference on Neural Information Processing Systems, Sep 2024
-
Investigating Style Similarity in Diffusion ModelsIn Proceedings of the European Conference on Computer Vision, Apr 2024
-
Generating Potent Poisons and Backdoors from Scratch with Guided Diffusionarxiv:2403.16365[cs], Mar 2024
-
Privacy Backdoors: Enhancing Membership Inference through Poisoning Pre-trained ModelsIn The Thirty-eighth Annual Conference on Neural Information Processing Systems, Sep 2024
-
Object Recognition as Next Token PredictionIn Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Sep 2024
2023
-
-
Cold Diffusion: Inverting Arbitrary Image Transforms Without NoiseIn Thirty-Seventh Conference on Neural Information Processing Systems, Nov 2023
-
Universal Guidance for Diffusion ModelsIn The Twelfth International Conference on Learning Representations, Oct 2023
-
A Performance-Driven Benchmark for Feature Selection in Tabular Deep LearningIn Thirty-Seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track, Nov 2023
-
Loss Landscapes Are All You Need: Neural Network Generalization Can Be Explained Without the Implicit Bias of Gradient DescentIn The Eleventh International Conference on Learning Representations, Feb 2023
-
Panning for Gold in Federated Learning: Targeted Text Extraction under Arbitrarily Large-Scale AggregationIn International Conference on Learning Representations, Feb 2023
-
Decepticons: Corrupted Transformers Breach Privacy in Federated Learning for Language ModelsIn International Conference on Learning Representations, Feb 2023
-
Cramming: Training a Language Model on a Single GPU in One Day.In Proceedings of the 40th International Conference on Machine Learning, Jul 2023
-
How Much Data Are Augmentations Worth? An Investigation into Scaling Laws, Invariance, and Implicit RegularizationIn International Conference on Learning Representations, Feb 2023
-
A Survey on the Possibilities & Impossibilities of AI-generated Text DetectionTransactions on Machine Learning Research, Oct 2023
-
Baseline Defenses for Adversarial Attacks Against Aligned Language Modelsarxiv:2309.00614[cs], Sep 2023
-
On the Reliability of Watermarks for Large Language ModelsIn The Twelfth International Conference on Learning Representations, Oct 2023
-
A Watermark for Large Language ModelsIn Proceedings of the 40th International Conference on Machine Learning, Jul 2023
-
Augmenters at SemEval-2023 Task 1: Enhancing CLIP in Handling Compositionality and Ambiguity for Zero-Shot Visual WSD through Prompt Augmentation and Text-To-Image DiffusionIn Proceedings of the The 17th International Workshop on Semantic Evaluation (SemEval-2023), Jul 2023
-
-
-
-
What Can We Learn from Unlearnable Datasets?In Thirty-Seventh Conference on Neural Information Processing Systems, Nov 2023
-
On the Exploitability of Instruction TuningIn Thirty-Seventh Conference on Neural Information Processing Systems, Nov 2023
-
-
Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion ModelsIn Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Nov 2023
-
Understanding and Mitigating Copying in Diffusion ModelsIn Thirty-Seventh Conference on Neural Information Processing Systems, Nov 2023
-
Canary in a Coalmine: Better Membership Inference with Ensembled Adversarial QueriesIn International Conference on Learning Representations, Feb 2023
-
Hard Prompts Made Easy: Gradient-Based Discrete Optimization for Prompt Tuning and DiscoveryIn Thirty-Seventh Conference on Neural Information Processing Systems, Nov 2023
-
STYX: Adaptive Poisoning Attacks Against Byzantine-Robust Defenses in Federated LearningIn ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Jun 2023
-
Tree-Rings Watermarks: Invisible Fingerprints for Diffusion ImagesIn Thirty-Seventh Conference on Neural Information Processing Systems, Nov 2023
2022
-
A Simple Strategy to Provable Invariance via Orbit MappingIn Asian Conference on Computer Vision (ACCV), Dec 2022
-
-
-
Autoregressive Perturbations for Data PoisoningIn Advances in Neural Information Processing Systems, Dec 2022
-
Poisons That Are Learned Faster Are More EffectiveIn 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Jun 2022
-
Fishing for User Data in Large-Batch Federated Learning via Gradient MagnificationIn Proceedings of the 39th International Conference on Machine Learning, Jun 2022
-
Thinking Two Moves Ahead: Anticipating Other Users Improves Backdoor Attacks in Federated LearningIn AdvML Frontiers Workshop at 39th International Conference on Machine Learning, Jun 2022
2021
-
DP-InstaHide: Provably Defusing Poisoning and Backdoor Attacks with Differentially Private Data AugmentationsIn ICLR 2021 Workshop on Security and Safety in Machine Learning Systems, Mar 2021
-
Strong Data Augmentation Sanitizes Poisoning and Backdoor Attacks Without an Accuracy TradeoffIn ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Jun 2021
-
Adversarial Examples Make Strong PoisonsIn Advances in Neural Information Processing Systems, Jun 2021
-
Preventing Unauthorized Use of Proprietary Data: Poisoning for Secure Dataset ReleaseIn ICLR 2021 Workshop on Security and Safety in Machine Learning Systems, Feb 2021
-
Robbing the Fed: Directly Obtaining Private Data in Federated Learning with Modified ModelsIn International Conference on Learning Representations, Sep 2021
-
DARTS for Inverse Problems: A Study on Hyperparameter SensitivityarXiv:2108.05647 [cs], Aug 2021
-
-
Stochastic Training Is Not Necessary for GeneralizationIn International Conference on Learning Representations, Sep 2021
-
What Doesn’t Kill You Makes You Robust(Er): Adversarial Training against Poisons and BackdoorsIn ICLR 2021 Workshop on Security and Safety in Machine Learning Systems, Feb 2021
-
Witches’ Brew: Industrial Scale Data Poisoning via Gradient MatchingIn International Conference on Learning Representations, Apr 2021
2020
-
Witchcraft: Efficient PGD Attacks with Random Step SizeIn ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), May 2020
-
Fast Convex Relaxations Using Graph DiscretizationsIn 31st British Machine Vision Conference (BMVC 2020, Oral Presentation), Sep 2020
-
Inverting Gradients - How Easy Is It to Break Privacy in Federated Learning?In Advances in Neural Information Processing Systems, Dec 2020
-
Truth or Backpropaganda? An Empirical Investigation of Deep Learning TheoryIn Eighth International Conference on Learning Representations (ICLR 2020, Oral Presentation), Apr 2020
-
MetaPoison: Practical General-purpose Clean-label Data PoisoningIn Advances in Neural Information Processing Systems, Dec 2020
2019
-
Parametric Majorization for Data-Driven Energy Minimization MethodsIn Proceedings of the IEEE International Conference on Computer Vision, Dec 2019
-
Piecewise Rigid Scene Flow with Implicit Motion SegmentationIn 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Nov 2019
2018
-
Composite Optimization by Nonconvex Majorization-MinimizationSIAM Journal on Imaging Sciences, Jan 2018
-
Multiframe Motion Coupling for Video Super ResolutionIn Energy Minimization Methods in Computer Vision and Pattern Recognition, Jan 2018
2016
-
Image Analysis of Neural Tissue Development: Variational Methods for Segmentation and 3D-Reconstruction from Large Pinhole Confocal Fluorescence MicroscopyWestfälischen Wilhelms-Universität Münster, Sep 2016
2014
-
Comparison of Topology-preserving Segmentation Methods and Application to Mitotic Cell TrackingSep 2014