8 papers
BERAG: Bayesian Ensemble Retrieval-Augmented Generation for Knowledge-based Visual Question Answering
Jinghong Chen, Jingbiao Mei, Guangyu Yang +1
A common approach to question answering with retrieval-augmented generation (RAG) is to concatenate documents into a single context and pass it to a language model to generate an a…
According to Me: Long-Term Personalized Referential Memory QA
Jingbiao Mei, Jinghong Chen, Guangyu Yang +3
Personalized AI assistants must recall and reason over long-term user memory, which naturally spans multiple modalities and sources such as images, videos, and emails. However, exi…
ExPO-HM: Learning to Explain-then-Detect for Hateful Meme Detection
Jingbiao Mei, Mingsheng Sun, Jinghong Chen +4
Hateful memes have emerged as a particularly challenging form of online abuse, motivating the development of automated detection systems. Most prior approaches rely on direct detec…
CycleVLA: Proactive Self-Correcting Vision-Language-Action Models via Subtask Backtracking and Minimum Bayes Risk Decoding
Chenyang Ma, Guangyu Yang, Kai Lu +6
Current work on robot failure detection and correction typically operates in a post hoc manner, analyzing errors and applying corrections only after failures occur. This work intro…
On Extending Direct Preference Optimization to Accommodate Ties
Jinghong Chen, Guangyu Yang, Weizhe Lin +2
We derive and investigate two DPO variants that explicitly model the possibility of declaring a tie in pair-wise comparisons. We replace the Bradley-Terry model in DPO with two wel…
Retrieval-Augmented Defense: Adaptive and Controllable Jailbreak Prevention for Large Language Models
Guangyu Yang, Jinghong Chen, Jingbiao Mei +2
Large Language Models (LLMs) remain vulnerable to jailbreak attacks, which attempt to elicit harmful responses from LLMs. The evolving nature and diversity of these attacks pose ma…