Shi Fu

Theoretical Guarantees for Iterative Alignment of Self-Rewarding Language Models

Towards a Theoretical Understanding of In-context Learning: Stability and Non-I.I.D Generalisation

The State of Reinforcement Finetuning for Transformer-based Agents

SeWA: Selective Weight Average via Probabilistic Masking

Self-Verification Provably Prevents Model Collapse in Recursive Synthetic Training

A Theoretical Perspective: How to Prevent Model Collapse in Self-consuming Training Loops