4 papers · 1 filter
f-GRPO and Beyond: Divergence-Based Reinforcement Learning Algorithms for General LLM Alignment
Rajdeep Haldar, Lantao Mei, Guang Lin +2
Recent work shows that preference alignment objectives can be interpreted as divergence estimators between aligned (preferred) & unaligned (less-preferred) distributions, yielding…
LLM Safety Alignment is Divergence Estimation in Disguise
Rajdeep Haldar, Ziyi Wang, Qifan Song +2
We present a theoretical framework showing that popular LLM alignment methods, including RLHF and its variants, can be understood as divergence estimators between aligned (safe or…
Effect of Ambient-Intrinsic Dimension Gap on Adversarial Vulnerability
Rajdeep Haldar, Yue Xing, Qifan Song
The existence of adversarial attacks on machine learning models imperceptible to a human is still quite a mystery from a theoretical perspective. In this work, we introduce two not…
On Neural Network approximation of ideal adversarial attack and convergence of adversarial training
Rajdeep Haldar, Qifan Song
Adversarial attacks are usually expressed in terms of a gradient-based operation on the input data and model, this results in heavy computations every time an attack is generated.…