Glossary

Terms

Deep research agent

A deep research agent is an AI system that plans a broad question, runs many searches and reads, and writes a cited report.

DRACO

DRACO is a benchmark of 100 deep research tasks from Perplexity and Harvard that scores reports against expert rubrics.

BrowseComp

BrowseComp is an OpenAI benchmark of 1,266 hard-to-find facts that tests whether an agent can keep searching until it finds a short, checkable answer.

LLM judge

An LLM judge is a language model that grades another model's output against a set of criteria.

Rubric

A rubric is a list of specific, checkable criteria that a good answer should meet.

Factual accuracy

Factual accuracy is whether each claim in a report matches what its cited source actually says.

Multi-step search

Multi-step search is a loop in which an agent uses the result of one search to decide what to search next.

Research chapter

A research chapter is one section of a deep research plan, covering a single sub-question and its evidence.

Cite this entry