response quality

5 papers

• CARE: Decoding-Time Safety Alignment via Rollback and Introspection Intervention
• DynamicRAG: Leveraging Outputs of Large Language Model as Feedback for Dynamic Reranking in Retrieval-Augmented Generation
• Scalable Best-of-N Selection for Large Language Models via Self-Certainty
• Transcending Cost-Quality Tradeoff in Agent Serving via Session-Awareness
• Weaver: Shrinking the Generation-Verification Gap by Scaling Compute for Verification