# ICE: Intervention-Consistent Explanation Evaluation with Statistical Grounding for LLMs

arXiv: https://arxiv.org/abs/2603.18579 · HTML: https://mlsanity.com/paper/2603.18579/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 37 references that are in the map.
- First posted on arXiv 19 Mar 2026.

## Builds on (most cited first, 22 of 22)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- ["Why Should I Trust You?": Explaining the Predictions of Any Classifier](https://mlsanity.com/paper/1602.04938/) (2016)
- [Axiomatic Attribution for Deep Networks](https://mlsanity.com/paper/1703.01365/) (2017)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [No Language Left Behind: Scaling Human-Centered Machine Translation](https://mlsanity.com/paper/2207.04672/) (2022)
- [Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting](https://mlsanity.com/paper/2305.04388/) (2023)
- [Towards Faithfully Interpretable NLP Systems: How should we define and evaluate faithfulness?](https://mlsanity.com/paper/2004.03685/) (2020)
- [ERASER: A Benchmark to Evaluate Rationalized NLP Models](https://mlsanity.com/paper/1911.03429/) (2019)
- [e-SNLI: Natural Language Inference with Natural Language Explanations](https://mlsanity.com/paper/1812.01193/) (2018)
- [A Diagnostic Study of Explainability Techniques for Text Classification](https://mlsanity.com/paper/2009.13295/) (2020)
- [Towards Leaving No Indic Language Behind: Building Monolingual Corpora, Benchmark and Models for Indic Languages](https://mlsanity.com/paper/2212.05409/) (2022)
- [Are self-explanations from Large Language Models faithful?](https://mlsanity.com/paper/2401.07927/) (2024)
- [The Out-of-Distribution Problem in Explainability and Search Methods for Feature Importance Explanations](https://mlsanity.com/paper/2106.00786/) (2021)
- [On Measuring Faithfulness or Self-consistency of Natural Language Explanations](https://mlsanity.com/paper/2311.07466/) (2023)
- [F-Fidelity: A Robust Framework for Faithfulness Evaluation of Explainable AI](https://mlsanity.com/paper/2410.02970/) (2024)
- [A Causal Lens for Evaluating Faithfulness Metrics](https://mlsanity.com/paper/2502.18848/) (2025)
- [Comparing Explanation Faithfulness between Multilingual and Monolingual Fine-tuned Language Models](https://mlsanity.com/paper/2403.12809/) (2024)
- [Evaluating Input Feature Explanations through a Unified Diagnostic Evaluation Framework](https://mlsanity.com/paper/2406.15085/) (2024)
- [Dynamic Top-k Estimation Consolidates Disagreement between Feature Attribution Methods](https://mlsanity.com/paper/2310.05619/) (2023)
- [Permutation-based Hypothesis Testing for Neural Networks](https://mlsanity.com/paper/2301.11354/) (2023)
- [Learning from Sufficient Rationales: Analysing the Relationship Between Explanation Faithfulness and Token-level Regularisation Strategies](https://mlsanity.com/paper/2511.16353/) (2025)

## Abstract (arXiv)

Evaluating whether explanations faithfully reflect a model's reasoning remains an open problem. Existing benchmarks use single interventions without statistical testing, making it impossible to distinguish genuine faithfulness from chance-level performance. We show that faithfulness is not a fixed property but an operator-dependent quantity that changes with the intervention method used to measure it. We introduce ICE (Intervention-Consistent Explanation), a framework that evaluates explanations against random baselines of equal size under multiple operators. Evaluating 7 LLMs across 4 tasks with deletion and retrieval infill operators, we find that switching operators crosses the positive-evidence threshold in 18% of configurations (5 of 28 attention comparisons), with gaps reaching 44 percentage points. Randomized baselines detect anti-faithfulness (explanations worse than random) in nearly one-third of English deletion configurations, invisible without random comparisons. These patterns persist across 6 non-English languages and 2 attribution methods. The methodology generalizes to step-level chain-of-thought evaluation, where preliminary results on 3 frontier models suggest that high accuracy does not imply faithful reasoning.