Deepchecks ORION: SOTA for Detecting Hallucination in LLM Workflows

Description

In this webinar, we talked about ORION, our new SOTA evaluation models designed to detect hallucinations in LLM applications, RAG pipelines, & Agentic workflows.

Philip Tannor & Assaf Gerner from the Deepchecks team covered how ORION performs claim-level evaluations and how the system identifies failure modes in LLM applications.

Topics that were covered:

✅ Claim-level evaluation for precision

✅ Detecting hallucinations, planning breakdowns, and tool misuse

✅ Deepchecks’ “Swarm of Evaluation Agents” approach

✅ Integrations with LangChain, LlamaIndex, CrewAI, AutogenAI, AWS AI SageMaker, and more

Recent blogs:

LLMOps.Space LLMOps.Space

Deepchecks is a founding member of LLMOps.Space, a global community for LLM
practitioners. The community focuses on LLMOps-related content, discussions, and
events. Join thousands of practitioners on our Discord.
Join Discord ServerJoin Discord Server