# Truth or Twist? Optimal Model Selection for Reliable Label Flipping Evaluation in LLM-based Counterfactuals

arXiv: https://arxiv.org/abs/2505.13972 · HTML: https://mlsanity.com/paper/2505.13972/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 30 of its 48 references that are in the map.
- First posted on arXiv 20 May 2025.

## Builds on (most cited first, 30 of 30)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [A large annotated corpus for learning natural language inference](https://mlsanity.com/paper/1508.05326/) (2015)
- [Explanation in Artificial Intelligence: Insights from the Social Sciences](https://mlsanity.com/paper/1706.07269/) (2017)
- [Learning the Difference that Makes a Difference with Counterfactually-Augmented Data](https://mlsanity.com/paper/1909.12434/) (2019)
- [Polyjuice: Generating Counterfactuals for Explaining, Evaluating, and Improving Models](https://mlsanity.com/paper/2101.00288/) (2021)
- [Measuring Association Between Labels and Free-Text Rationales](https://mlsanity.com/paper/2010.12762/) (2020)
- [Post-hoc Interpretability for Neural NLP: A Survey](https://mlsanity.com/paper/2108.04840/) (2021)
- [Preference Leakage: A Contamination Problem in LLM-as-a-judge](https://mlsanity.com/paper/2502.01534/) (2025)
- [Explaining NLP Models via Minimal Contrastive Editing (MiCE)](https://mlsanity.com/paper/2012.13985/) (2020)
- [Generate Your Counterfactuals: Towards Controlled Counterfactual Generation for Text](https://mlsanity.com/paper/2012.04698/) (2020)
- [Towards LLM-guided Causal Explainability for Black-box Text Classifiers](https://mlsanity.com/paper/2309.13340/) (2023)
- [Prompting Large Language Models for Counterfactual Generation: An Empirical Study](https://mlsanity.com/paper/2305.14791/) (2023)
- [LLMs for Generating and Evaluating Counterfactuals: A Comprehensive Study](https://mlsanity.com/paper/2405.00722/) (2024)
- [CORE: A Retrieve-then-Edit Framework for Counterfactual Data Generation](https://mlsanity.com/paper/2210.04873/) (2022)
- [CREST: A Joint Framework for Rationalization and Counterfactual Text Generation](https://mlsanity.com/paper/2305.17075/) (2023)
- [Is BERT Robust to Label Noise? A Study on Learning with Noisy Labels in Text Classification](https://mlsanity.com/paper/2204.09371/) (2022)
- [Zero-shot LLM-guided Counterfactual Generation: A Case Study on NLP Model Evaluation](https://mlsanity.com/paper/2405.04793/) (2024)
- [Enhancing LLM Robustness to Perturbed Instructions: An Empirical Study](https://mlsanity.com/paper/2504.02733/) (2025)
- [Text Counterfactuals via Latent Optimization and Shapley-Guided Search](https://mlsanity.com/paper/2110.11589/) (2021)
- [LLMCheckup: Conversational Examination of Large Language Models via Interpretability Tools and Self-Explanations](https://mlsanity.com/paper/2401.12576/) (2024)
- [Counterfactual Evaluation for Explainable AI](https://mlsanity.com/paper/2109.01962/) (2021)
- [CEval: A Benchmark for Evaluating Counterfactual Text Generation](https://mlsanity.com/paper/2404.17475/) (2024)
- [InterroLang: Exploring NLP Models and Datasets through Dialogue-based Explanations](https://mlsanity.com/paper/2310.05592/) (2023)
- [Text Classification in the LLM Era -- Where do we stand?](https://mlsanity.com/paper/2502.11830/) (2025)
- [Towards Unifying Evaluation of Counterfactual Explanations: Leveraging Large Language Models for Human-Centric Assessments](https://mlsanity.com/paper/2410.21131/) (2024)
- [FitCF: A Framework for Automatic Feature Importance-guided Counterfactual Example Generation](https://mlsanity.com/paper/2501.00777/) (2025)

## Built on by (most cited first, 5 of 5)

- [Can Large Language Models Still Explain Themselves? Investigating the Impact of Quantization on Self-Explanations](https://mlsanity.com/paper/2601.00282/) (2026)
- [Parallel Universes, Parallel Languages: A Comprehensive Study on LLM-based Multilingual Counterfactual Example Generation](https://mlsanity.com/paper/2601.00263/) (2026)
- [iFlip: Iterative Feedback-driven Counterfactual Example Refinement](https://mlsanity.com/paper/2601.01446/) (2026)
- [Explanation Generation for Contradiction Reconciliation with LLMs](https://mlsanity.com/paper/2603.22735/) (2026)
- [Causal Evidence Extraction and Triangulation in Crisis Reports using Large Language Models: A ReliefWeb-based Study](https://mlsanity.com/paper/2608.04576/) (2026)

## Abstract (arXiv)

Counterfactual examples are widely employed to enhance the performance and robustness of large language models (LLMs) through counterfactual data augmentation (CDA). However, the selection of the judge model used to evaluate label flipping, the primary metric for assessing the validity of generated counterfactuals for CDA, yields inconsistent results. To decipher this, we define four types of relationships between the counterfactual generator and judge models: being the same model, belonging to the same model family, being independent models, and having an distillation relationship. Through extensive experiments involving two state-of-the-art LLM-based methods, three datasets, four generator models, and 15 judge models, complemented by a user study (n = 90), we demonstrate that judge models with an independent, non-fine-tuned relationship to the generator model provide the most reliable label flipping evaluations. Relationships between the generator and judge models, which are closely aligned with the user study for CDA, result in better model performance and robustness. Nevertheless, we find that the gap between the most effective judge models and the results obtained from the user study remains considerably large. This suggests that a fully automated pipeline for CDA may be inadequate and requires human intervention.