CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing
By Zhibin Gou, Zhihong Shao, Yeyun Gong, Yelong Shen, Yujiu Yang, Nan Duan, Weizhu Chen. 2023
809 citations recorded by the source corpus.
Recent developments in large language models (LLMs) have been impressive. However, these models sometimes show inconsistencies and problematic behavior, such as hallucinating facts, generating flawed code, or creating offensive and toxic content. Unlike these models, humans typically utilize external tools to cross-check and refine their initial content, like using a search engine for fact-checking, or a code interpreter for debugging. Inspired by this observation, we introduce a framework called CRITIC that allows LLMs, which are essentially"black boxes"to validate and progressively amend their own outputs in a manner similar to human interaction with tools. More specifically, starting with an initial output, CRITIC interacts with appropriate tools to evaluate certain aspects of the text, and then revises the output based on the feedback obtained during this validation process. Comprehensive evaluations involving free-form question answering, mathematical program synthesis, and toxicity reduction demonstrate that CRITIC consistently enhances the performance of LLMs. Meanwhile, our research highlights the crucial importance of external feedback in promoting the ongoing self-improvement of LLMs.
Subfields: Artificial Intelligence, Information Systems, Large Language Models, LLM Agents & Tool Use
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Jianfeng Gao
- Mohit Bansal
- Furu Wei
- Luke S. Zettlemoyer
- Percy Liang
- Hannaneh Hajishirzi
- Yejin Choi
- Graham Neubig
In Information Systems
In Large Language Models
In LLM Agents & Tool Use
Citations in the corpus
- Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements (2023, background): “And for LMs, they can generate self-feedback (Madaan et al., 2023) or utilize a given feedback (Gou et al., 2023), like from classifiers, to self-correct the unsafe response.”
- Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks (2022, methodology): “These methods propose to enhance LLMs’ capabilities to solve math problems with PoT. self-critic (Gou et al., 2023) and self-eval (Xie et al., 2021) both adopt self-evaluation to enhance the robustness of the generated program. plan-and-solve (Wang et al., 2023a) instead adopt more detailed…”
- Understanding the planning of LLM agents: A survey (2024, methodology): “CRITIC [Gou et al. , 2023] uses external tools like Knowledge Bases and Search Engines to validate LLM-generated actions.”
- ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem Solving (2023, methodology): “Recent work demonstrates the benefits of integrating retrievers (Borgeaud et al., 2022; Shao et al., 2023b), search engines (Nakano et al., 2021), and multi-tool approaches (Schick et al., 2023; Paranjape et al., 2023; Gou et al., 2023) to improve generation.”
- Unleashing the Emergent Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration (2023, methodology): “…LLMs to solve tasks step by step or iteratively revise their answers (Wei et al., 2023; Kojima et al., 2022; Zhang et al., 2022; Fu et al., 2022; Xue et al., 2023; Yao et al., 2023; Madaan et al., 2023; Shinn et al., 2023; Gou et al., 2023; (a) Visualization of the SPP-identified personas.”
- A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions (2023, neutral): “This feedback typically comes from external knowledge sources through retrieval processes (He et al., 2023; Gou et al., 2023), interactive debates with other LLMs (Du et al., 2023; Cohen et al., 2023), or guidance from external evaluation metrics (Lei et al., 2023; Khalifa et al., 2023).”
- Tool learning with large language models: a survey (2024, neutral): “We categorize the latest advancements in this stage into two types: direct insertion methods and information integration methods.”
- Large Language Models Cannot Self-Correct Reasoning Yet (2023, background): “Furthermore, several related works have already presented findings consistent with our observation: that self-correction, when devoid of external feedback, serves as a relatively weak baseline (Gou et al., 2023; Zhou et al., 2023a).”
- Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate (2023, background): “, 2023) but still struggle on complex reasoning tasks (Zhu et al., 2022; Gou et al., 2023), which drives the research on cognitive behaviors of LLMs to explore human-like problem-solving strategies.”
- MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning (2023, positive): “Self-critic (Gou et al., 2023) and self-eval (Xie et al., 2022) both adopt self-evaluation to enhance the robustness of the generated program.”
- Understanding the Effects of Iterative Prompting on Truthfulness (2024, methodology): “Despite these challenges, advancements in self-correction and validation techniques [28, 29], coupled with frameworks for self-alignment [30], reflect the comprehensive approaches being explored to bolster the performance and reliability of LLMs.”
- Agentic Retrieval-Augmented Generation: A Survey on Agentic RAG (2025, positive): “Reflection has demonstrated significant performance improvements in studies like Self-Refine [28], Reflexion [29], and CRITIC [24].”
- Enhancing Retrieval-Augmented Large Language Models with Iterative Retrieval-Generation Synergy (2023, methodology): “To alleviate these issues, there is a growing trend of augmenting LLMs with tools (Mialon et al., 2023; Gou et al., 2023), e.g., a code interpreter (Gao et al., 2022b; Shao et al., 2023) or a search engine (Nakano et al., 2021), in an attempt to offload sub-tasks to more qualified experts, or to…”