How AI systems really think
How do AI systems arrive at an answer—and what makes that answer trustworthy? In this webinar, we’ll explore how AI uses context, follows chains of verification, and checks information step by step to produce more accurate and reliable results.
What We’ll Explore
This webinar examines two important processes behind AI-generated answers: how an assistant constructs the context it uses to understand a request, and how structured verification can help it identify and correct factual mistakes.
1. Context
Prompt ≠ Input: Prompt-and-Context Construction Behind AI Assistant Behavior
Presented by Tamar Peretz — Gen AI Researcher
This talk focuses on how visible conversation prompts are turned into model context before inference. In modern AI assistants, the latest user message is not passed to the model in isolation. It is combined with prior conversation turns, selected context, system and product instructions, retrieved material, tool outputs, memory signals, and operational constraints. In practice, this construction step determines which signals receive precedence and what the model actually conditions on when producing an answer.
2. Chain of Verification
Reducing Hallucinations Through Structured Self-Checking
Presented by Shimon Moyal — AI Data Engineer
Chain of Verification is an artificial intelligence prompting method designed to reduce hallucinations and improve the factual reliability of model-generated responses. Instead of treating its first answer as final, the model creates an initial draft, plans verification questions to check the claims it has made, answers those questions independently to avoid being influenced by the original response, and then produces a revised final answer. This structured process separates generation from verification, helping the model identify and correct mistakes before responding and produce more accurate and dependable outputs.
What You’ll Take Away
Why the prompt you write is only one part of the information an AI assistant receives.
How conversation history, instructions, retrieved information, tools, and other signals shape an AI-generated answer.
How Chain of Verification separates initial generation from factual checking.
Why independent verification can reduce hallucinations and improve reliability.
A clearer understanding of what happens between asking an AI system a question and receiving its final response.
Join Tamar Peretz and Shimon Moyal for a closer look behind the interface—and discover how context construction and structured verification shape the answers AI systems produce.