# Enhancing Large Language Models Against Inductive Instructions with Dual-critique Prompting

arXiv: https://arxiv.org/abs/2305.13733 · HTML: https://mlsanity.com/paper/2305.13733/

## Key facts

- Cited by 21 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 29 references that are in the map.
- First posted on arXiv 23 May 2023.

## Builds on (most cited first, 17 of 17)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [The Curious Case of Neural Text Degeneration](https://mlsanity.com/paper/1904.09751/) (2019)
- [Know What You Don't Know: Unanswerable Questions for SQuAD](https://mlsanity.com/paper/1806.03822/) (2018)
- [GLM: General Language Model Pretraining with Autoregressive Blank Infilling](https://mlsanity.com/paper/2103.10360/) (2021)
- [Red Teaming Language Models with Language Models](https://mlsanity.com/paper/2202.03286/) (2022)
- [GLM-130B: An Open Bilingual Pre-trained Model](https://mlsanity.com/paper/2210.02414/) (2022)
- [A Comprehensive Survey on Pretrained Foundation Models: A History from BERT to ChatGPT](https://mlsanity.com/paper/2302.09419/) (2023)
- [CREPE: Open-Domain Question Answering with False Presuppositions](https://mlsanity.com/paper/2211.17257/) (2022)
- [Cue-CoT: Chain-of-thought Prompting for Responding to In-depth Dialogue Questions with LLMs](https://mlsanity.com/paper/2305.11792/) (2023)
- [(QA)2: Question Answering with Questionable Assumptions](https://mlsanity.com/paper/2212.10003/) (2022)
- [Which Linguist Invented the Lightbulb? Presupposition Verification for Question-Answering](https://mlsanity.com/paper/2101.00391/) (2021)
- [Will ChatGPT get you caught? Rethinking of Plagiarism Detection](https://mlsanity.com/paper/2302.04335/) (2023)
- [CHEF: A Pilot Chinese Dataset for Evidence-Based Fact-Checking](https://mlsanity.com/paper/2206.11863/) (2022)

## Built on by (most cited first, 21 of 21)

- [Challenges and Applications of Large Language Models](https://mlsanity.com/paper/2307.10169/) (2023)
- [Unleashing the potential of prompt engineering for large language models](https://mlsanity.com/paper/2310.14735/) (2023)
- [Foot-In-The-Door: A Multi-turn Jailbreak for LLMs](https://mlsanity.com/paper/2502.19820/) (2025)
- [Improving Diversity of Demographic Representation in Large Language Models via Collective-Critiques and Self-Voting](https://mlsanity.com/paper/2310.16523/) (2023)
- [Query-Dependent Prompt Evaluation and Optimization with Offline Inverse RL](https://mlsanity.com/paper/2309.06553/) (2023)
- [Analyzing the Inherent Response Tendency of LLMs: Real-World Instructions-Driven Jailbreak](https://mlsanity.com/paper/2312.04127/) (2023)
- [Distractor generation for multiple-choice questions with predictive prompting and large language models](https://mlsanity.com/paper/2307.16338/) (2023)
- [UAlign: Leveraging Uncertainty Estimations for Factuality Alignment on Large Language Models](https://mlsanity.com/paper/2412.11803/) (2024)
- [Could ChatGPT get an Engineering Degree? Evaluating Higher Education Vulnerability to AI Assistants](https://mlsanity.com/paper/2408.11841/) (2024)
- [Don't Take the Premise for Granted: Evaluating the Premise Critique Ability of Large Language Models](https://mlsanity.com/paper/2505.23715/) (2025)
- [What Makes a Good Natural Language Prompt?](https://mlsanity.com/paper/2506.06950/) (2025)
- [Layer Importance and Hallucination Analysis in Large Language Models via Enhanced Activation Variance-Sparsity](https://mlsanity.com/paper/2411.10069/) (2024)
- [MlingConf: A Comprehensive Study of Multilingual Confidence Estimation on Large Language Models](https://mlsanity.com/paper/2402.13606/) (2024)
- [Alignment is not sufficient to prevent large language models from generating harmful information: A psychoanalytic perspective](https://mlsanity.com/paper/2311.08487/) (2023)
- [Modeling Hierarchical Thinking in Large Reasoning Models](https://mlsanity.com/paper/2510.22437/) (2025)
- [NILE: Internal Consistency Alignment in Large Language Models](https://mlsanity.com/paper/2412.16686/) (2024)
- [Attention-Aware GNN-based Input Defense against Multi-Turn LLM Jailbreak](https://mlsanity.com/paper/2507.07146/) (2025)
- [Refining Critical Thinking in LLM Code Generation: A Faulty Premise-based Evaluation Framework](https://mlsanity.com/paper/2508.03622/) (2025)
- [Don't `Well, Actually' Me Unless You Know What You're Talking About: Weak Presupposition Verification Degrades General QA Performance](https://mlsanity.com/paper/2608.06539/) (2026)
- [MlingConf: A Comprehensive Study of Multilingual Confidence Estimation on Large Language Models](https://mlsanity.com/paper/2410.12478/) (2024)
- [From Instruction to Output: The Role of Prompting in Modern NLG](https://mlsanity.com/paper/2602.11179/) (2026)

## Abstract (arXiv)

Numerous works are proposed to align large language models (LLMs) with human intents to better fulfill instructions, ensuring they are trustful and helpful. Nevertheless, some human instructions are often malicious or misleading and following them will lead to untruthful and unsafe responses. Previous work rarely focused on understanding how LLMs manage instructions based on counterfactual premises, referred to here as textitinductive instructions, which may stem from users' false beliefs or malicious intents. In this paper, we aim to reveal the behaviors of LLMs towards textitinductive instructions and enhance their truthfulness and helpfulness accordingly. Specifically, we first introduce a benchmark of underlineInductive Inunderlinestructions (textscINDust), where the false knowledge is incorporated into instructions in multiple different styles. After extensive human and automatic evaluations, we uncovered a universal vulnerability among LLMs in processing inductive instructions. Additionally, we identified that different inductive styles affect the models' ability to identify the same underlying errors, and the complexity of the underlying assumptions also influences the model's performance. Motivated by these results, we propose textscDual-critique prompting to improve LLM robustness against inductive instructions. Our experiments demonstrate that textscDual-critique prompting significantly bolsters the robustness of a diverse array of LLMs, even when confronted with varying degrees of inductive instruction complexity and differing inductive styles.