Tool-using large language model (LLM) agents translate model outputs into consequential actions, turning indirect prompt injection into an execution-security problem. Existing defenses constrain tool calls, trajectories, or information flows, but often leave action semantics, state dependencies, and evidence relations...
Guo-Qin Xiong, Sheng-Tao Wen, Xiang Chen et al.· 0 citations
Current evaluations of multimodal unlearning typically focus on whether a model stops outputting target information. This surface-level criterion, however, masks two critical failure modes. First, knowledge suppression techniques often induce confabulation, leading the model to hallucinate facts about the unlearned sub...
Hao-Dong Chen, Sheng-Tao Wen, Dong Liang et al.· 0 citations
This work introduces two lightweight execution-consistent interfaces, Future-Sequence for TD-MPC2 and Applied-Action Feedback for DreamerV3, that correct future-action timeline mismatch between imagined and executed action sequences without modifying the pretrained world models.
Sheng-Tao Wen, Xiang Chen, Yu Tian et al.· 0 citations
SPARK is proposed, a retention-correction decomposition that first freezes the learned post-task distribution and then applies selective correction around this stable reference and achieves effective selective PII suppression while preserving strong continual-learning utility and knowledge retention across diverse sett...
VARM-Bench provides an auditable and reproducible benchmark for evaluating verifiable moderation rationales in Chinese abusive-speech moderation, and shows that strong label-level performance can conceal substantial errors in complete moderation records.
Mingyu Yuan, Sheng-Tao Wen, Ling-Bing Guo et al.· 0 citations
Recall-Anchored Distillation (RAD), a base-anchored self-distillation objective that preserves out-of-distribution generation behavior by aligning the adapted model with the original base model's soft continuation distribution on unlabeled OOD text, is introduced.
Hao-Dong Chen, Ya-Dong Wang, Sheng-Tao Wen et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.