Despite advances in AI alignment, large language models (LLMs) remain vulnerable to adversarial attacks or jailbreaking, in which adversaries can modify prompts to induce unwanted behavior.

They are used in our experiments section. OpenAI GPT -4 based prompts The input template for OpenAI GPT -4 is defined as: Please describe the appearance of [class_name] in endoscopic surgery, and change the description to a phrase with subject, and not use colons. The dataset consists of both training and test cases. Each video is recorded at 25 FPS and has annotations for instruments and operation phases. For EndoVis2019, the results are shown in Tab. 1, our method (input size 448) notably surpasses the competition's top performers, with +3% increase in DSC and +2% enhancement in NSD, which demonstrates the superiority of our method.

We call this dilemma the generalization problem.

Neural Information Processing Systems http://nips.cc/

As showninTable 10,adopting DIST with Pearson correlation achieveshigher accuracies compared to KD and DIST with cosine similarity, especially when the teacher and student are trained with label smoothing (the predicted probabilistic distributions would be shifted by it). The speed is tested based on our implementations on 8 NVIDIAV100GPUs. KDRKDSRRLCRDDIST [16] [30] [47] [41] 14.28 11.11 12.98 8.33 14.19 A.5 Landscapesofmatchingfunctions As discussed in our main text, the matching functions such as KL divergence and MSE are used tomatch the outputs between student and teacher inKD. DIST can improve multi-class classification consistently in various tasks such as classification, object detection, andsemantic segmentation. However,itwouldbeless-effectiveon binary classification task, as the task only contains two classes and the information in inter-class relation is limited.