State of AI Pod • 010

Contents

1. Tailored-LLaMA: Optimizing Few-Shot Learning in Pruned LLaMA Models with Task-Specific Prompts

2. Safeguarding System Prompts for LLMs

3. From Scientific Texts to Verifiable Code: Automating the Process with Transformers

4. OpenOmni: Large Language Models Pivot Zero-shot Omnimodal Alignment across Language with Real-time Self-Aware Emotional Speech Synthesis