Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

linguistic perturbations

Linguistic perturbations are deliberate, systematic modifications applied to natural language text to evaluate and test how computational language models process and respond to altered inputs. These modifications encompass a wide spectrum of changes, including character-level alterations like typos, word-level adjustments such as synonym substitutions, syntactic restructuring, and higher-level semantic or informational shifts like introducing contradictions, ambiguity, or missing context. Within artificial intelligence and natural language processing, linguistic perturbations serve as diagnostic instruments to probe model robustness, detect behavioral failure modes, determine whether systems rely on superficial statistical correlations rather than true comprehension, and evaluate reliability and safety under non-ideal or manipulated textual conditions.

1 item

RefusalBench: Generative Evaluation of Selective Refusal in Grounded Language Models

RefusalBench: Generative Evaluation of Selective Refusal in Grounded Language Models

Aashiq Muhamed, Leonardo F. R. Ribeiro, Markus Dreyer, Virginia Smith, Mona T. Diab

OrganizationsAmazonCarnegie Mellon University

Why you should read this

Presents RefusalBench, a generative evaluation framework using 176 perturbation strategies to expose how frontier retrieval-augmented models fail to selectively refuse answering from flawed contexts, while demonstrating that this capability requires targeted training rather than increased model scale.

The ability of language models in RAG systems to selectively refuse to answer based on flawed context is critical for safety, yet remains a significant failure point. Our large-scale study reveals that even frontier models struggle in this setting, with refusal accuracy dropping below 50% on multi-document tasks, while exhibiting either dangerous overconfidence or overcaution. Static benchmarks fail to reliably evaluate this capability, as models exploit dataset-specific artifacts and memorize test instances. We introduce RefusalBench, a generative methodology that programmatically creates diagnostic test cases through controlled linguistic perturbation. Our framework employs 176 distinct perturbation strategies across six categories of informational uncertainty and three intensity levels. Evaluation of over 30 models uncovers systematic failure patterns: refusal comprises separable detection and categorization skills, and neither scale nor extended reasoning improves performance. We find that selective refusal is a trainable, alignment-sensitive capability, offering a clear path for improvement. We release two benchmarks -- RefusalBench-NQ (single document) and RefusalBench-GaRAGe (multi-document) -- and our complete generation framework to enable continued, dynamic evaluation of this critical capability.

Added

2026-09-30