protologue

CRITIC

Also called tool-interactive critiquing.

CRITIC is a self-correction framework in which a model checks its output using external tools, such as web search or a code interpreter, critiques it based on what the tools return, and then revises it.

Description

Grounding the critique in tool feedback addresses the main weakness of intrinsic self-correction. The original work showed improvements on question answering, program synthesis, and reducing toxic output, and the same verify-then-revise loop is common in agent systems.

Sources

  1. Gou et al. (2023). CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing.

Cite this entry

Protologue. (2026). CRITIC. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0119). https://protologue.com/t/critic/

BibTeX
@misc{protologue_critic,
  title = {CRITIC},
  author = {{Protologue}},
  year = {2026},
  howpublished = {Protologue: A Taxonomy of Prompting and LLM Techniques, v1.4.0},
  note = {Entry PTL-0119},
  url = {https://protologue.com/t/critic/}
}

Markdown JSON