← Search

Will Cai

2 accepted papers

2025

Improving LLM Safety Alignment with Dual-Objective Optimization

ICML 2025poster

Existing training-time safety alignment techniques for large language models (LLMs) remain vulnerable to jailbreak attacks. Direct preference optimization (DPO), a widely deployed alignment method, exhibits limitations in both experimental and theoretical contexts as its loss function proves subopti…

2025

Scaling Trends for Data Poisoning in LLMs

AAAI 2025technical

LLMs produce harmful and undesirable behavior when trained on datasets containing even a small fraction of poisoned data. We demonstrate that GPT models remain vulnerable to fine-tuning on poisoned data, even when safeguarded by moderation systems. Given the persistence of data poisoning vulnerabili…