Victor Veitch

Chasing the Tail: Effective Rubric-based Reward Modeling for Large Language Model Post-Training

Expected Reward Prediction, with Applications to Model Routing

Incorporating Hierarchical Semantics in Sparse Autoencoder Architectures

Let it Calm: Exploratory Annealed Decoding for Verifiable Reinforcement Learning

SemanticDPP: Efficient Uncertainty Quantification in LLMs

On the Significance of Softmax Geometry: Interpretability and Token Decoding

The Geometry of Categorical and Hierarchical Concepts in Large Language Models

RATE: Score Reward Models with Imperfect Rewrites of Rewrites

RATE: Causal Explainability of Reward Models with Imperfect Counterfactuals