Furong Huang
Associate Professor
Researches trustworthy foundation models, reinforcement learning and generative AI alignment.
Furong Huang's lab at UMD works on generative & foundation models, trustworthy & safe ML and interpretability, evaluation & safety of LMs. 143 papers since 2023, 1 at top venues, mostly Findings of EMNLP and Nature Communications. 12 PhD students. 18 alumni, most often to Amazon and OpenAI.
At a glance
- papers, 2023–2026
- 37/ year
- at T1 venues
- 1%
- PhD students
- 12
- MS students
- 0
- alumni listed
- 18
Papers by yearfrom the lab homepage, 2023–2026
- 37 papers2026 · T1 0 · collecting
- 39 papers2025 · T1 1
- 42 papers2024 · T1 0
- 25 papers2023 · T1 0
T1 venuesOtherVenues: Findings of EMNLP, Nature Communications, TKDD
Research focuspaper titles that use each topic's words
- Generative & foundation models30 papers
- alignment15 papers
- Interpretability, evaluation & safety of LMs9 papers
- Trustworthy & safe ML7 papers
- researches trustworthy0 papers
Paper list
From the lab homepage.
202637 papers
- The Ninth Annual ACM Conference on Fairness, Accountability, and Transparency (FAccT)Advancing Regulation in Artificial Intelligence: An Auction-Based Approach
- AAAI AI Alignment Track (AAAI), OralAdvBDGen: A Robust Framework for Generating Adaptive and Stealthy Backdoors in LLM Alignment Attacks
- Communications MedicineAssessing safety and trustworthiness of large language models in medicine
- IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)Beyond Episodic Evaluation: Memory Architectural Bottlenecks in Sequential Embodied Question Answering
- BiAlign Workshop, poster, NeurIPSCompliance2LoRA: Personalizable On-Demand Safety Alignment on Arbitrary Policy Subsets via Hypernetwork-Generated LoRA Adapters
- Workshop on Compositional Learning: Safety, Interpretability, and Agents, ICMLCompositional Adversarial Training for Robust Visual Watermarking
- Actionable Interpretability Workshop (AIW), COLMDeliberative Alignment is Deep, but Uncertainty Remains: Inference time safety improvement in reasoning via attribution of unsafe behavior to base model
- Findings of EMNLPDoes Reasoning Preserve Alignment? On the Trustworthiness of Large Reasoning Models
- Conference on Neural Information Processing Systems (NeurIPS), Main Track, PosterDynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation
- Conference on Robot Learning (CoRL)EgoScale: Scaling Dexterous Manipulation with Diverse Egocentric Human Data
- Findings, The Annual Meeting of the Association for Computational Linguistics (ACL)EnsemW2S: Enhancing Weak-to-Strong Generalization with Large Language Model Ensembles
- Conference on Neural Information Processing Systems (NeurIPS), Main Track, PosterFlowBank: Query-Adaptive Agentic Workflows Optimization through Precompute-and-Reuse
202539 papers
- The Thirty- Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track (NeurIPS)A Technical Report on "Erasing the Invisible": The 2024 NeurIPS Competition on Stress Testing Image Watermarks
- Nature Communications (T1)Adversarial Prompt and Fine-Tuning Attacks Threaten Medical Large Language Models
- Workshop BuildingTrust Workshop, ICLRAegisLLM: Scaling Agentic Systems for Self-Reflective Defense in LLM Security
- TKDDBalancing Fairness and Accuracy in Data-Restricted Binary Classification
- The Thirteenth International Conference on Learning Representations (ICLR)Benchmarking Vision Language Model Unlearning via Fictitious Facial Identity Dataset
- Workshop on New Perspective in Graph Machine Learning (NPGML), NeurIPSBridging the Divide: End-to-End Sequence-Graph Learning
- Best Paper Award at the Third Workshop on New Frontiers in Adversarial Machine Learning, NeurIPS . The Annual AAAI Conference on Artificial Intelligence (AAAI)Can Watermarking Large Language Models Prevent Copyrighted Text Generation and Hide Training Data?
- The Thirteenth International Conference on Learning Representations (ICLR)Collab: Controlled Decoding using Mixture of Agents for LLM Alignment
- The International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR)CSRec: Rethinking Sequential Recommendation from A Causal Perspective
- The Conference on Empirical Methods in Natural Language Processing (EMNLP)DISCO Balances the Scales: Adaptive Domain- and Difficulty-Aware Reinforcement Learning on Imbalanced Data
- The Thirty- Annual Conference on Neural Information Processing Systems (NeurIPS)Does Thinking More Always Help? Mirage of Test-Time Scaling in Reasoning Models
- Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society (AIES), 8(2):1875-1887Effort-aware Fairness: Incorporating a Philosophy-informed, Human-centered Notion of Effort into Algorithmic Fairness Metrics
202442 papers
- Journal of Biomedical Informatics : 104646A survey of recent methods for addressing AI fairness and bias in biomedicine
- Transactions on Machine Learning Research (TMLR)A Survey on the Possibilities & Impossibilities of AI-generated Text Detection
- Proceedings of the International Conference on Machine Learning (ICML)ACE: Off-Policy Actor-Critic with Causality-Aware Entropy Regularization
- Proceedings of the International Conference on Machine Learning (ICML)Adapting Static Fairness to Sequential Decision-Making: Bias Mitigation Strategies towards Equal Long-term Benefit Rate
- First Conference on Language Modeling (COLM)AutoDAN: Interpretable Gradient-Based Adversarial Attacks on Large Language Models
- The Conference on Empirical Methods in Natural Language Processing (EMNLP)AUTOHALLUSION: Automatic Generation of Hallucination Benchmarks for Vision-Language Models
- First Conference on Language Modeling (COLM)Automatic Pseudo-Harmful Prompt Generation for Evaluating False Refusals in Large Language Models
- International Conference on Pattern Recognition and Artificial Intelligence (ICPRAI)Balancing Label Imbalance in Federated Environments Using Only Mixup and Artificially-Labeled Noise
- Spotlight. The Twelfth International Conference on Learning Representations (ICLR)Beyond Worst-case Attacks: Robust RL with Adaptive Defense via Non-dominated Policies
- The Thirty- Annual Conference on Neural Information Processing Systems (NeurIPS)Boosting Sample Efficiency and Generalization in Multi-agent Reinforcement Learning via Equivariance
- The Twelfth International Conference on Learning Representations (ICLR)COPlanner: Plan to Roll Out Conservatively but to Explore Optimistically for Model-Based RL
- The Twelfth International Conference on Learning Representations (ICLR)Decodable and Sample Invariant Continuous Object Encoder
202325 papers
- AdvML-Frontiers workshop, ICMLAdapting Robust Reinforcement Learning to Handle Temporally-Coupled Perturbations
- The Thirty- Annual Conference on Neural Information Processing Systems (NeurIPS)C-Disentanglement: Discovering Causally-Independent Generative Factors under an Inductive Bias of Confounder
- The Eleventh International Conference on Learning Representations (ICLR)Certifiably Robust Multi-Agent Reinforcement Learning against Adversarial Communication
- Rebellion and Disobedience in AI (RaD-AI) Workshop, AAMASCertifiably Robust Policy Learning against Adversarial Multi-Agent Communication
- The Thirty- Annual Conference on Neural Information Processing Systems (NeurIPS)Cold Diffusion: Inverting Arbitrary Image Transforms Without Noise
- AdvML-Frontiers workshop, ICMLEqual Long-term Benefit Rate: Adapting Static Fairness Notions to Sequential Decision Making
- The Eleventh International Conference on Learning Representations (ICLR)Exploring and Exploiting Decision Boundary Dynamics for Adversarial Robustness
- The Eleventh International Conference on Learning Representations (ICLR)Is Model Ensemble Necessary? Model-based RL via a Single Model with Lipschitz Regularized Value Function
- The Thirty- Annual Conference on Neural Information Processing Systems (NeurIPS)Large-Scale Distributed Learning via Private On-Device LSH
- Proceedings of the International Conference on Machine Learning (ICML)Learning Unforeseen Robustness from Out-of-distribution Data Using Equivariant Domain Translator
- Proceedings of the International Conference on Machine Learning (ICML)Live in the Moment: Learning Dynamics Model Adapted to Evolving Policy
- Workshop on Efficient Systems for Foundation Models (ES-FoMO), ICMLMental Calibration: Discovering and Adjusting for Latent Factors Improves Zero-Shot Inference of CLIP
Where alumni went18 listed, 85% to industry
Industry 11Academia 2
Most often to
- Amazon
- OpenAI
- Apple
Also to
- MIT
- University of Chicago · Postdoctoral Researcher with Ce Zhang
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
- USCMU
Steven Wu
- Trustworthy & safe ML
- USUSC
Phebe Vayanos
- Trustworthy & safe ML
- USStanford
Ellen Vitercik
- Learning theory & optimization
- GBEdinburgh
Henry Gouk
- Trustworthy & safe ML
- USUCSD
Sanjoy Dasgupta
- Learning theory & optimization
- CNPeking
Tongyang Li
- Learning theory & optimization
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.