3 citations · 9 across the 20 of their papers we have counts for
14 papers · 1 filter
From Curiosity to Caution: Mitigating Reward Hacking for Best-of-N with Pessimism
Zhuohao Yu, Zhiwei Steven Wu, Adam Block
Inference-time compute scaling has emerged as a powerful paradigm for improving language model performance on a wide range of tasks, but the question of how best to use the additio…
Back to Blackwell: Closing the Loop on Intransitivity in Multi-Objective Preference Fine-Tuning
Jiahao Zhang, Lujing Zhang, Keltin Grimes +3
A recurring challenge in preference fine-tuning (PFT) is handling (i.e., cyclic) preferences. Intransitive preferences often stem from either …
From Firewalls to Frontiers: AI Red-Teaming is a Domain-Specific Evolution of Cyber Red-Teaming
Anusha Sinha, Keltin Grimes, James Lucassen +4
A red team simulates adversary attacks to help defenders find effective strategies to defend their systems in a real-world operational setting. As more enterprise systems adopt AI,…
Discretization-free Multicalibration through Loss Minimization over Tree Ensembles
Hongyi Henry Jin, Zijun Ding, Dung Daniel Ngo +1
In recent years, multicalibration has emerged as a desirable learning objective for ensuring that a predictor is calibrated across a rich collection of overlapping subpopulations.…
Dimension-Free Decision Calibration for Nonlinear Loss Functions
Jingwu Tang, Jiayun Wu, Zhiwei Steven Wu +1
When model predictions inform downstream decision making, a natural question is under what conditions can the decision-makers simply respond to the predictions as if they were the…
All Roads Lead to Likelihood: The Value of Reinforcement Learning in Fine-Tuning
Gokul Swamy, Sanjiban Choudhury, Wen Sun +2
From a first-principles perspective, it may seem odd that the strongest results in foundation model fine-tuning (FT) are achieved via a relatively complex, two-stage training proce…
Kandinsky Conformal Prediction: Beyond Class- and Covariate-Conditional Coverage
Konstantina Bairaktari, Jiayun Wu, Zhiwei Steven Wu
Conformal prediction is a powerful distribution-free framework for constructing prediction sets with coverage guarantees. Classical methods, such as split conformal prediction, pro…
Multi-Agent Imitation Learning: Value is Easy, Regret is Hard
Jingwu Tang, Gokul Swamy, Fei Fang +1
We study a multi-agent imitation learning (MAIL) problem where we take the perspective of a learner attempting to coordinate a group of agents based on demonstrations of an expert…
Bridging Multicalibration and Out-of-distribution Generalization Beyond Covariate Shift
Jiayun Wu, Jiashuo Liu, Peng Cui +1
We establish a new model-agnostic optimization framework for out-of-distribution generalization via multicalibration, a criterion that ensures a predictor is calibrated across a fa…
Provable Pluralistic Alignment: Multi-Party RLHF under Offline Human Feedback
Huiying Zhong, Zhun Deng, Tianwei Gao +3
Pluralistic alignment requires learning from feedback that reflects persistent and potentially conflicting stakeholder preferences while ultimately selecting a single collective po…
The Virtues of Pessimism in Inverse Reinforcement Learning
David Wu, Gokul Swamy, J. Andrew Bagnell +2
Inverse Reinforcement Learning (IRL) is a powerful framework for learning complex behaviors from expert demonstrations. However, it traditionally requires repeatedly solving a comp…
Hybrid Inverse Reinforcement Learning
Juntao Ren, Gokul Swamy, Zhiwei Steven Wu +2
The inverse reinforcement learning approach to imitation learning is a double-edged sword. On the one hand, it can enable learning from a smaller number of expert demonstrations wi…
A Minimaximalist Approach to Reinforcement Learning from Human Feedback
Gokul Swamy, Christoph Dann, Rahul Kidambi +2
We present Self-Play Preference Optimization (SPO), an algorithm for reinforcement learning from human feedback. Our approach is minimalist in that it does not require training a r…
Learning Shared Safety Constraints from Multi-task Demonstrations
Konwoo Kim, Gokul Swamy, Zuxin Liu +3
Regardless of the particular task we want them to perform in an environment, there are often shared safety constraints we want our agents to respect. For example, regardless of whe…