4 papers
Uncertainty-Penalized Direct Preference Optimization
Sam Houliston, Alizée Pace, Alexander Immer +1
Aligning Large Language Models (LLMs) to human preferences in content, style, and presentation is challenging, in part because preferences are varied, context-dependent, and someti…
Preference Elicitation for Offline Reinforcement Learning
Alizée Pace, Bernhard Schölkopf, Gunnar Rätsch +1
Applying reinforcement learning (RL) to real-world problems is often made challenging by the inability to interact with the environment and the difficulty of designing reward funct…
West-of-N: Synthetic Preferences for Self-Improving Reward Models
Alizée Pace, Jonathan Mallinson, Eric Malmi +2
The success of reinforcement learning from human feedback (RLHF) in language model alignment is strongly dependent on the quality of the underlying reward model. In this paper, we…
On the Importance of Step-wise Embeddings for Heterogeneous Clinical Time-Series
Rita Kuznetsova, Alizée Pace, Manuel Burger +2
Recent advances in deep learning architectures for sequence modeling have not fully transferred to tasks handling time-series from electronic health records. In particular, in prob…