Tom Goldstein
Professor
Studies optimization and the reliability of deep learning, including adversarial robustness, watermarking of LLM outputs and diffusion models.
Tom Goldstein's lab at UMD works on trustworthy & safe ML, learning theory & optimization and image & video generation. 72 papers since 2023, 62 at top venues, mostly NeurIPS and ICLR. 23 alumni, most often to Apple and NVIDIA.
At a glance
- papers, 2023–2026
- 19/ year
- at T1 venues
- 86%
- PhD students
- 0
- MS students
- 0
- alumni listed
- 23
Papers by yearfrom DBLP, 2023–2026
- 8 papers2026 · T1 7 · collecting
- 18 papers2025 · T1 15
- 20 papers2024 · T1 17
- 26 papers2023 · T1 23
T1 venuesOtherVenues: NeurIPS, ICLR, CVPR, ICML, ECCV
Research focuspaper titles that use each topic's words
- Learning theory & optimization11 papers
- Image & video generation9 papers
- Trustworthy & safe ML2 papers
- watermarking1 paper
- outputs0 papers
Paper list
From DBLP.
20268 papers
- ICLR (T1)DynaGuard: A Dynamic Guardian Model With User-Defined Policies (opens in a new tab)
- ICLR (T1)FictionalQA: A Dataset for Studying Memorization and Knowledge Acquisition (opens in a new tab)
- ECCV (T1)Flash-BoN: Instant Drafts for Inference-Time Scaling in Diffusion Models (opens in a new tab)
- ICLR (T1)Identifying and Evaluating Inactive Heads in Pretrained LLMs (opens in a new tab)
- CVPR (T1)SpeeDe3DGS: Speedy Deformable 3D Gaussian Splatting with Temporal Pruning and Motion Grouping (opens in a new tab)
- CVPR (T1)SplatSuRe: Selective Super-Resolution for Multi-view Consistent 3D Gaussian Splatting (opens in a new tab)
- FGTransFIRA: Transfer Learning for Face Image Recognizability Assessment (opens in a new tab)
- ICLR (T1)Zebra-CoT: A Dataset for Interleaved Vision-Language Reasoning (opens in a new tab)
202518 papers
- NeurIPS (T1)A Technical Report on "Erasing the Invisible": The 2024 NeurIPS Competition on Stress Testing Image Watermarks (opens in a new tab)
- ICCV (T1)ARGUS: Hallucination and Omission Evaluation in Video-LLMs (opens in a new tab)
- AAAI (T1)Can Watermarking Large Language Models Prevent Copyrighted Text Generation and Hide Training Data? (opens in a new tab)
- NeurIPS (T1)Dense Backpropagation Improves Training for Sparse Mixture-of-Experts (opens in a new tab)
- CVPR (T1)Efficient Fine-Tuning and Concept Suppression for Pruned Diffusion Models (opens in a new tab)
- Findings of NAACLEnhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement (opens in a new tab)
- NeurIPS (T1)FineGRAIN: Evaluating Failure Modes of Text-to-Image Models with Vision Language Model Judges (opens in a new tab)
- NeurIPS (T1)Gemstones: A Model Suite for Multi-Faceted Scaling Laws (opens in a new tab)
- ISC High PerformanceLeveraging AI for Productive and Trustworthy HPC Software: Challenges and Research Directions (opens in a new tab)
- ICLR (T1)LiveBench: A Challenging, Contamination-Limited LLM Benchmark (opens in a new tab)
- Findings of NAACLLLM-Generated Passphrases That Are Secure and Easy to Remember (opens in a new tab)
- AAMAS (T1)Multimodal Agentic Model Predictive Control (opens in a new tab)
202420 papers
- NeurIPS (T1)Be like a Goldfish, Don't Memorize! Mitigating Memorization in Generative LLMs (opens in a new tab)
- NeurIPS (T1)CALVIN: Improved Contextual Video Captioning via Instruction Tuning (opens in a new tab)
- SCDemocratizing AI: Open-source Scalable LLM Training on GPU-based Supercomputers (opens in a new tab)
- ENLSPDense Backpropagation Improves Routing for Sparsely-Gated Mixture-of-Experts (opens in a new tab)
- NeurIPS (T1)Easy2Hard-Bench: Standardized Difficulty Labels for Profiling LLM Performance and Generalization (opens in a new tab)
- ICASSP (T1)FedAQT: Accurate Quantized Training with Federated Learning (opens in a new tab)
- Trans. Mach. Learn. Res.Graph Neural Networks Formed via Layer-wise Ensembles of Heterogeneous Base Models (opens in a new tab)
- ICRA (T1)Hierarchical Point Attention for Indoor 3D Object Detection (opens in a new tab)
- ICML (T1)InstructZero: Efficient Instruction Optimization for Black-Box Large Language Models (opens in a new tab)
- ECCV (T1)Investigating Style Similarity in Diffusion Models (opens in a new tab)
- ICLR (T1)NEFTune: Noisy Embeddings Improve Instruction Finetuning (opens in a new tab)
- CVPR (T1)Object Recognition as Next Token Prediction (opens in a new tab)
202326 papers
- AAAI/ACM AIESA Deep Dive into Dataset Imbalance and Bias in Face Identification (opens in a new tab)
- NeurIPS (T1)A Performance-Driven Benchmark for Feature Selection in Tabular Deep Learning (opens in a new tab)
- ICML (T1)A Watermark for Large Language Models (opens in a new tab)
- NeurIPS (T1)Battle of the Backbones: A Large-Scale Comparison of Pretrained Models across Computer Vision Tasks (opens in a new tab)
- ICLR (T1)Canary in a Coalmine: Better Membership Inference with Ensembled Adversarial Queries (opens in a new tab)
- NeurIPS (T1)Cold Diffusion: Inverting Arbitrary Image Transforms Without Noise (opens in a new tab)
- ICML (T1)Cramming: Training a Language Model on a single GPU in one day (opens in a new tab)
- TPAMI (T1)Dataset Security for Machine Learning: Data Poisoning, Backdoor Attacks, and Defenses (opens in a new tab)
- ICLR (T1)Decepticons: Corrupted Transformers Breach Privacy in Federated Learning for Language Models (opens in a new tab)
- CVPR (T1)Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Models (opens in a new tab)
- ICLR (T1)Exploring and Exploiting Decision Boundary Dynamics for Adversarial Robustness (opens in a new tab)
- ICML (T1)GOAT: A Global Transformer on Large-scale Graphs (opens in a new tab)
Where alumni went23 listed, 81% to industry
Industry 13Academia 3
Most often to
- Apple
- NVIDIA
- Adobe
Also to
- MIT
- NYU
- University of Maryland
Apply to UMDFall 2027
Not announced yet
UMD Computer Science PhD
- Round
- Single round
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- GBImperial
Jiankang Deng
- Vision-language & multimodal
- USUMD
Soheil Feizi
- Trustworthy & safe ML
- USUSC
C.-C. Jay Kuo
- Video understanding
- CNPeking
Wen Gao
- Video understanding
- GBUCL
Fabio Pierazzi
- Trustworthy & safe ML
- GBOxford
Philip Torr
- Trustworthy & safe ML
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- DBLP
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.