# Breaking Free from MMI: A New Frontier in Rationalization by Probing Input Utilization

arXiv: https://arxiv.org/abs/2503.06202 · HTML: https://mlsanity.com/paper/2503.06202/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 46 of its 62 references that are in the map.
- First posted on arXiv 8 Mar 2025.

## Builds on (most cited first, 30 of 46)

- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Survey of Hallucination in Natural Language Generation](https://mlsanity.com/paper/2202.03629/) (2022)
- [Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization](https://mlsanity.com/paper/1911.08731/) (2019)
- [Intrinsic Dimensionality Explains the Effectiveness of Language Model Fine-Tuning](https://mlsanity.com/paper/2012.13255/) (2020)
- [LESS: Selecting Influential Data for Targeted Instruction Tuning](https://mlsanity.com/paper/2402.04333/) (2024)
- [ERASER: A Benchmark to Evaluate Rationalized NLP Models](https://mlsanity.com/paper/1911.03429/) (2019)
- [Is ChatGPT a General-Purpose Natural Language Processing Task Solver?](https://mlsanity.com/paper/2302.06476/) (2023)
- [NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models](https://mlsanity.com/paper/2405.17428/) (2024)
- [Git Re-Basin: Merging Models modulo Permutation Symmetries](https://mlsanity.com/paper/2209.04836/) (2022)
- [Backdoor Learning: A Survey](https://mlsanity.com/paper/2007.08745/) (2020)
- [Parameterized Explainer for Graph Neural Network](https://mlsanity.com/paper/2011.04573/) (2020)
- [Causal Reasoning and Large Language Models: Opening a New Frontier for Causality](https://mlsanity.com/paper/2305.00050/) (2023)
- [TrustLLM: Trustworthiness in Large Language Models](https://mlsanity.com/paper/2401.05561/) (2024)
- [Rationalizing Neural Predictions](https://mlsanity.com/paper/1606.04155/) (2016)
- [A Comprehensive Capability Analysis of GPT-3 and GPT-3.5 Series Models](https://mlsanity.com/paper/2303.10420/) (2023)
- [Discovering Invariant Rationales for Graph Neural Networks](https://mlsanity.com/paper/2201.12872/) (2022)
- [Invariant Rationalization](https://mlsanity.com/paper/2003.09772/) (2020)
- [Evaluating ChatGPT's Information Extraction Capabilities: An Assessment of Performance, Explainability, Calibration, and Faithfulness](https://mlsanity.com/paper/2304.11633/) (2023)
- [Interpretable Neural Predictions with Differentiable Binary Variables](https://mlsanity.com/paper/1905.08160/) (2019)
- [Learning to Faithfully Rationalize by Construction](https://mlsanity.com/paper/2005.00115/) (2020)
- [An Interpretability Illusion for BERT](https://mlsanity.com/paper/2104.07143/) (2021)
- [Rethinking Cooperative Rationalization: Introspective Extraction and Complement Control](https://mlsanity.com/paper/1910.13294/) (2019)
- [Leakage-Adjusted Simulatability: Can Models Generate Non-Trivial Explanations of Their Behavior in Natural Language?](https://mlsanity.com/paper/2010.04119/) (2020)
- [An Information Bottleneck Approach for Controlling Conciseness in Rationale Extraction](https://mlsanity.com/paper/2005.00652/) (2020)
- [Deriving Machine Attention from Human Rationales](https://mlsanity.com/paper/1808.09367/) (2018)
- [SelfExplain: A Self-Explaining Architecture for Neural Text Classifiers](https://mlsanity.com/paper/2103.12279/) (2021)
- [Understanding Interlocking Dynamics of Cooperative Rationalization](https://mlsanity.com/paper/2110.13880/) (2021)
- [Can Rationalization Improve Robustness?](https://mlsanity.com/paper/2204.11790/) (2022)
- [UNIREX: A Unified Learning Framework for Language Model Rationale Extraction](https://mlsanity.com/paper/2112.08802/) (2021)
- [Interpretable Data-Based Explanations for Fairness Debugging](https://mlsanity.com/paper/2112.09745/) (2021)

## Built on by (most cited first, 6 of 6)

- [Beyond Zero Initialization: Investigating the Impact of Non-Zero Initialization on LoRA Fine-Tuning Dynamics](https://mlsanity.com/paper/2505.23194/) (2025)
- [MTL-UE: Learning to Learn Nothing for Multi-Task Learning](https://mlsanity.com/paper/2505.05279/) (2025)
- [Adversarial Cooperative Rationalization: The Risk of Spurious Correlations in Even Clean Datasets](https://mlsanity.com/paper/2505.02118/) (2025)
- [The Panaceas for Improving Low-Rank Decomposition in Communication-Efficient Federated Learning](https://mlsanity.com/paper/2505.23176/) (2025)
- [Quantifying Distributional Invariance in Causal Subgraph for IRM-Free Graph Generalization](https://mlsanity.com/paper/2510.20295/) (2025)
- [TopInG: Topologically Interpretable Graph Learning via Persistent Rationale Filtration](https://mlsanity.com/paper/2510.05102/) (2025)

## Abstract (arXiv)

Extracting a small subset of crucial rationales from the full input is a key problem in explainability research. The most widely used fundamental criterion for rationale extraction is the maximum mutual information (MMI) criterion. In this paper, we first demonstrate that MMI suffers from diminishing marginal returns. Once part of the rationale has been identified, finding the remaining portions contributes only marginally to increasing the mutual information, making it difficult to use MMI to locate the rest. In contrast to MMI that aims to reproduce the prediction, we seek to identify the parts of the input that the network can actually utilize. This is achieved by comparing how different rationale candidates match the capability space of the weight matrix. The weight matrix of a neural network is typically low-rank, meaning that the linear combinations of its column vectors can only cover part of the directions in a high-dimensional space (high-dimension: the dimensions of an input vector). If an input is fully utilized by the network, {it generally matches these directions (e.g., a portion of a hypersphere), resulting in a representation with a high norm. Conversely, if an input primarily falls outside (orthogonal to) these directions}, its representation norm will approach zero, behaving like noise that the network cannot effectively utilize. Building on this, we propose using the norms of rationale candidates as an alternative objective to MMI. Through experiments on four text classification datasets and one graph classification dataset using three network architectures (GRUs, BERT, and GCN), we show that our method outperforms MMI and its improved variants in identifying better rationales. We also compare our method with a representative LLM (llama-3.1-8b-instruct) and find that our simple method gets comparable results to it and can sometimes even outperform it.