paper-with-me

Papers

TorchOpera: A Compound AI System for LLM Safety

2024-06-16 · Shanshan Han, Zijian Hu, Alay Dilipbhai Shah, Han Jin, Yuhang Yao, Dimitris Stripelis, Zhaozhuo Xu, Chaoyang He

We introduce TorchOpera, a compound AI system for enhancing the safety and quality of prompts and responses for Large Language Models. TorchOpera ensures that all user prompts are safe, contextually grounded, and effectively processed, while enhancing LLM responses to be relevant and high quality. TorchOpera utilizes the vector database for contextual grounding, rule-based wrappers for flexible modifications, and specialized mechanisms for detecting and adjusting unsafe or incorrect content. We also provide a view of the compound AI system to reduce the computational cost. Extensive experiments show that TorchOpera ensures the safety, reliability, and applicability of LLMs in real-world settings while maintaining the efficiency of LLM responses.

📄 PDF Abstract BibTeX arXiv:2406.10847

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Safety Certification is Classification

2026-05-07 · Oliver Schön, Licio Romao, Sadegh Soudjani arxiv

The goal of this paper is certifying safety of dynamical systems subject to uncertainty. Existing approaches use trajectory data to estimate transition probabilities, and compute safety probabilities recursively via dyna…

Breaking the Epistemic Trap: Active Perception Under Compound Uncertainty

2026-05-26 · Chayan Banerjee, Ethan Goan arxiv

Deploying reinforcement learning in safety critical domains, from autonomous vehicles to medical decision support, is constrained by failures arising when systems encounter unfamiliar conditions. We argue that the fundam…

Reinforcement LearningAutonomous VehiclesDecision Making

Real-Time Adaptive Safety-Critical Control with Gaussian Processes in High-Order Uncertain Models

2024-02-29 · Yu Zhang, long wen, Xiangtong Yao, Zhenshan Bing 외

This paper presents an adaptive online learning framework for systems with uncertain parameters to ensure safety-critical control in non-stationary environments. Our approach consists of two phases. The initial phase is …

Computational EfficiencyGaussian Processes

Rigorous Evaluation of Predictive Toxicity Models by Multi-Objective Optimization of Reference Compound Lists Using Genetic Algorithms

2025-05-11 · Yohei Ohto, Tadahaya Mizuno, Yasuhiro Yoshikai, Hiromi Fujimoto 외

In pharmaceutical safety assessments, validation studies are essential for evaluating the predictive performance and reliability of alternative methods prior to regulatory acceptance. Typically, these studies utilize ref…

Diversity

One Anchor for All: Unified Multilingual and Multimodal Safety Alignment for LVLMs

2026-07-30 · Enyi Shi, Fei Shen, Chuancheng Shi, Linxia Zhu 외 arxiv

As large vision-language models (LVLMs) are deployed globally, the combination of multilingual instructions and visual information makes malicious attacks more covert and sophisticated than ever before. However, existing…