Large language models (LLMs) are increasingly being adopted in a wide range of real-world applications. Despite their impressive performance, recent studies have shown that LLMs are vulnerable to deliberately crafted adversarial prompts even when aligned via Reinforcement Learning from Human Feedback or supervised fine-tuning. While existing defense methods focus on either detecting harmful prompts or reducing the likelihood of harmful responses through various means, defending LLMs against jailbreak attacks based on the inner mechanisms of LLMs remains largely unexplored. In this work, we investigate how LLMs respond to harmful prompts and propose a novel defense method termed Layer-specific Editing (LED) to enhance the resilience of LLMs against jailbreak attacks. Through LED, we reveal that several critical safety layers exist among the early layers of LLMs. We then show that realigning these safety layers (and some selected additional layers) with the decoded safe response from identified toxic layers can significantly improve the alignment of LLMs against jailbreak attacks. Extensive experiments across various LLMs (e.g., Llama2, Mistral) show the effectiveness of LED, which effectively defends against jailbreak attacks while maintaining performance on benign prompts. Our code is available at https://github.com/ledllm/ledllm.

In this paper, we describe the system and approach used by Institute for Infocomm Research (I2R) for the IWSLT 2007 spoken language evaluation campaign. A multi-pass approach is exploited to generate and select best translation. First, we use two decoders namely the open source Moses and an in-home syntax-based decoder to generate N-best lists. Next we spawn new translation entries through a word-based n-gram language model estimated on the former N-best entries. Finally, we join the N-best lists from the previous two passes, and select the best translation by rescoring them with additional feature functions. In particular, this paper reports our effort on new translation entry generation and system combination. The performance on development and test sets are reported. The system was ranked first with respect to the BLEU measure in Chinese-to-English open data track.