While model and agent evaluation are inextricably linked, their technical benchmarks and metrics for success are fundamentally different.
Mastering Agentic Techniques: AI Agent Evaluation | NVIDIA Technical Blog… The deep research agent uses the create deep agent factory from LangChain’s deepagents library. from deepagents import create deep agent return create deep agent model=self.llm provider.get LLMRole.ORCHESTRATOR , system prompt=orchestrator prompt, tools=self.tools, subagents=self.subagents, middlew… …
… To learn about customizing AI agents, see Mastering Agentic Techniques: AI Agent Customization . What’s the difference between evaluating an AI model and evaluating an AI agent? …
… See NeMo Agent Toolkit in Action Create Your Own AI Agent 18:31 Benchmarking and Optimizing AI Agents 11:04 How To Develop Teams of AI Agents 23:06 Optimize Your AI Agent Workflows 05:29 How NeMo Agent Toolkit Works NVIDIA NeMo Agent Toolkit provides unified monitoring and optimization for AI agent… …
… An agent with persistent shell access, live credentials, the ability to rewrite its own tooling, and six hours of accumulated context running against your internal APIs is a fundamentally different threat model. …
… AI-Q 커스터마이징: 워크플로우, 추적 및 모델 설정 configs/config web docker.yml 파일을 열어보세요. 이 파일은 LLM, 도구, 에이전트, 워크플로우 구성을 모두 제어하는 핵심 설정 파일입니다. llms 섹션에서는 사용할 모델들을 선언합니다. 여기서 특히 enable thinking 플래그에 주목할 만한데, 이는 Nemotron의 생각의 사슬 Chain-of-Thought 추론 기능을 활성화하거나 비활성화하는 스위치입니다. 다음 예시는 서로 다른 역할을 가진 세 가지 LLM을 정의하고 있습니다. llms… …
… Discuss 0 Discuss 0 Tags Agentic AI / Generative AI | General | Blueprint | NeMo | NIM | Beginner Technical | News | Agent toolkit | AI Agent | AI Data Platform | AI-Ready Data | featured | GTC Paris 2025 | LLMs | Retrieval Augmented Generation RAG About the Authors About Nicola Sessions Nicola Ses… …
… May orchestrate sub-agents that tackle subtasks. Typically, the primary agent is powered by the smartest model and talks directly with the user. Sub-agents: Spawned by the primary agent to handle narrower tasks, with ability to self-manage their context windows like the primary agent. …
… Benjamin also works on accelerating software development velocity through the design and deployment of frontier coding agent systems. Before NVIDIA, Benjamin was previously Head of Agents at Groq, where he led their research agent line Compound. …
… A pipeline built for research, not adapted to it Agent harnesses are designed around orchestration. When agents handle research without a dedicated research backend, a general-purpose agent or sub-agent changes a research workflow. …
… This post explains nine techniques for customizing AI agents, along with criteria for selecting the right techniques for your use case. To learn about evaluating AI agents, see Mastering Agentic Techniques: AI Agent Evaluation . Why is it necessary to customize an AI agent? …