CRITIC is a framework that enables large language models to self-correct their outputs by interacting with external tools such as search engines and code interpreters. Introduced by Gou et al. (2023), CRITIC implements a verify-then-revise loop inspired by human critical thinking: generate an initial response, use tools to critique it, then revise based on evidence.

A fundamental limitation of LLMs is their tendency to generate plausible but incorrect outputs – hallucinated facts, faulty reasoning, or toxic content. Prior self-correction approaches that rely solely on the LLM's own judgment yield minimal improvement. CRITIC addresses this by grounding correction in external tool feedback, providing objective evidence for revision.

The key insight: LLM-only self-critique yields marginal gains (-0.03 to +2.33 F1), while tool-augmented critique produces substantial improvements (+7.0 to +7.7 F1).

At each iteration <latex>i</latex>, the revised output is sampled from:

<latex>\hat{y}_{i+1} \sim P_M(\cdot\, |\, \wp \oplus x \oplus \hat{y}_i \oplus c_i)</latex>

where <latex>\wp</latex> is the task prompt, <latex>x</latex> is the input, <latex>\hat{y}_i</latex> is the current output, and <latex>c_i</latex> is the tool-generated critique. The process uses in-context learning for general applicability – no task-specific fine-tuning required.

CRITIC was evaluated across three diverse task categories:

Question Answering:

• +7.7 F1 improvement over baselines across three QA tasks
• Beats rejection sampling by 4.5-3.3 EM
• CoT initial: 74.7 → 83.7 and 73.5 → 86.6 after tool-verified revision

Mathematical Reasoning:

• +7.0% absolute accuracy gains across three math tasks
• Code interpreter catches computational errors that self-reflection misses

Toxicity Reduction:

• 79.2% reduction in toxicity probability
• Matches supervised state-of-the-art without any training
• Preserves fluency and diversity of generated text

# CRITIC-style verify-then-revise loop
import openai
 
def critic_loop(query, max_iterations=3):
    # Step 1: Generate initial response
    response = openai.chat.completions.create(
        model='gpt-4',
        messages=[{'role': 'user', 'content': query}]
    ).choices[0].message.content
 
    for i in range(max_iterations):
        # Step 2: Verify with external tool (e.g., search)
        critique = verify_with_tool(response, query)
        if critique['is_correct']:
            break
        # Step 3: Revise based on critique
        revision_prompt = (
            'Original question: ' + query + '\n'
            'Your answer: ' + response + '\n'
            'Tool feedback: ' + critique['evidence'] + '\n'
            'Please revise your answer based on this feedback.'
        )
        response = openai.chat.completions.create(
            model='gpt-4',
            messages=[{'role': 'user', 'content': revision_prompt}]
        ).choices[0].message.content
    return response

• Gou et al. (2023) - CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing
• ToRA: Tool-Integrated Reasoning Agents (same research group)
• Self-Refine: Iterative Refinement with Self-Feedback

• ToRA: Tool-Integrated Reasoning
• Reflexion: Verbal Reinforcement Learning
• Self-Correction in LLM Agents