VERBA: Verbalizing Model Differences Using Large Language Models

2025/07/03 by Doda, Shravan, Javaji, Shashidhar Reddy, Zhu, Zining
#FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2507.02241

Abstract

In the current machine learning landscape, we face a "model lake" phenomenon: Given a task, there is a proliferation of trained models with similar performances despite different behavior. For model users attempting to navigate and select from the models, documentation comparing model pairs is helpful. However, for every N models there could be O(N²) pairwise comparisons, a number prohibitive for the model developers to manually perform pairwise comparisons and prepare documentations. To facilitate fine-grained pairwise comparisons among models, we introduced VERBA. Our approach leverages a large language model (LLM) to generate verbalizations of model differences by sampling from the two models. We established a protocol that evaluates the informativeness of the verbalizations via simulation. We also assembled a suite with a diverse set of commonly used machine learning models as a benchmark. For a pair of decision tree models with up to 5% performance difference but 20-25% behavioral differences, VERBA effectively verbalizes their variations with up to 80% overall accuracy. When we included the models' structural information, the verbalization's accuracy further improved to 90%. VERBA opens up new research avenues for improving the transparency and comparability of machine learning models in a post-hoc manner.

Citations

• CoSy: Evaluating Textual Explanations of Neurons
• LLM-Generated Black-box Explanations Can Be Adversarially Helpful
• Automatic Generation of Model and Data Cards: A Step Towards Responsible AI
• Towards Generating Informative Textual Description for Neurons in Language Models
• Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models
• Rigorously Assessing Natural Language Explanations of Neurons
• Explaining black box text modules in natural language with language models
• Eliciting Latent Predictions from Transformers with the Tuned Lens
• How Good Are GPT Models at Machine Translation? A Comprehensive Evaluation
• Data Cards: Purposeful and Transparent Dataset Documentation for Responsible AI
• Describing Differences between Text Distributions with Natural Language
• Natural Language Descriptions of Deep Visual Features
• Ranking and Tuning Pre-trained Models: A New Paradigm for Exploiting Model Hubs
• Compositional Explanations of Neurons
• Language Models are Few-Shot Learners
• Model Cards for Model Reporting

Related

• Model-diff: A Tool for Comparative Study of Language Models in the Input Space
• Verbosity Bias in Preference Labeling by Large Language Models
• BehaviorBox: Automated Discovery of Fine-Grained Performance Differences Between Language Models
• Discovering Language Model Behaviors with Model-Written Evaluations
• Verbosity ≠ Veracity: Demystify Verbosity Compensation Behavior of Large Language Models