7 papers
Distributionally Robust Token Optimization in RLHF
Yeping Jin, Jiaming Hu, Ioannis Ch. Paschalidis
Large Language Models (LLMs) tend to respond correctly to prompts that align well with the data they were trained and fine-tuned on. Yet, small shifts in wording, format, or langua…
Towards General Preference Alignment: Diffusion Models at Nash Equilibrium
Jiaming Hu, Jiamu Bai, Haoyu Wang +2
Reinforcement learning from human feedback (RLHF) has been popular for aligning text-to-image (T2I) diffusion models with human preferences. As a mainstream branch of RLHF, Direct…
Bridging the Gap Between Average and Discounted TD Learning
Haoxing Tian, Zaiwei Chen, Ioannis Ch. Paschalidis +1
The analysis of Temporal Difference (TD) learning in the average-reward setting faces notable theoretical difficulties because the Bellman operator is not contractive with respect…
Scaling In-Context Online Learning Capability of LLMs via Cross-Episode Meta-RL
Xiaofeng Lin, Sirou Zhu, Yilei Chen +6
Large language models (LLMs) achieve strong performance when all task-relevant information is available upfront, as in static prediction and instruction-following problems. However…
CCFC: Core & Core-Full-Core Dual-Track Defense for LLM Jailbreak Protection
Jiaming Hu, Haoyu Wang, Debarghya Mukherjee +1
Jailbreak attacks pose a serious challenge to the safe deployment of large language models (LLMs). We introduce CCFC (Core & Core-Full-Core), a dual-track, prompt-level defense fra…
Improving Mixup Calibration with Wasserstein Distributionally Robust Optimization
Jiaming Hu, Debarghya Mukherjee, Yeping Jin +1
In many real-world applications, ensuring the robustness and stability of deep neural networks (DNNs) is crucial, particularly for image classification tasks that encounter various…
Distributionally Robust Learning in Survival Analysis
Yeping Jin, Lauren Wise, Ioannis Ch. Paschalidis
We introduce an innovative approach that incorporates a Distributionally Robust Learning (DRL) approach into Cox regression to enhance the robustness and accuracy of survival predi…