Measuring Stability and Failure Behavior in Language Models Under Structured Perturbations

• URL: http://arxiv.org/abs/2608.22138v1
• Date: Sat, 22 Aug 2026 23:46:11 GMT
• Title: Measuring Stability and Failure Behavior in Language Models Under Structured Perturbations
• Abstract summary: We present a graded, multi-family, failure-aware framework for stress-testing reasoning models.<n>It perturbs each problem along a multi-level severity ladder across seven families.<n>Every test is validity-gated and labeled by its measured severity.
• Score: 0.5076419064097734
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Language models are usually judged by a single accuracy score, which does not reveal how their performance degrades as inputs are perturbed. We present a graded, multi-family, failure-aware framework for stress-testing reasoning models. It perturbs each problem along a multi-level severity ladder across seven families: six that preserve the answer, paraphrase, input noise, formatting, irrelevant context, context load, and conflicting instructions, and a Knowledge Boundary family that removes answerability so that refusal becomes the correct response. Every test is validity-gated and labeled by its measured severity, and each model is summarized by per-level Accuracy, a magnitude-weighted Stability, and a per-family Collapse Point defined relative to the model's own baseline. Instantiated on the same 100 seed problems used by GSM-Symbolic, expanded into 4,473 gated tests and run on four models spanning capability tiers, the framework exposes structure that an aggregate score hides: the level at which a model fails is family-specific rather than global, and two stressors expose consistent weaknesses across all models: conflicting instructions and questions built on an impossible premise. Recognition of unanswerability is otherwise uneven, reliable on missing information and fabricated evidence but weak on impossible premises. These failure points are invisible to standard accuracy reporting.

Related papers

• The Illusion of Robustness: Aggregate Accuracy Hides Prediction Flips under Task-Irrelevant Context [76.36635039699112]
  Large language models (LLMs) are increasingly deployed in context-rich settings where task inputs are often accompanied by long, partially irrelevant context.<n>We find that state-of-the-art models often appear robust to task-irrelevant context at the aggregate level.<n>This aggregate stability, however, masks significant per-example instability.
  arXiv  Detail & Related papers  (2026-07-14T17:01:12Z)
• Where Instruction Hierarchy Breaks: Diagnosing and Repairing Failures in Reasoning Language Models [15.752705487977861]
  We introduce a diagnostic framework that localizes instruction hierarchy failures into instruction identification, conflict resolution, and response realization.<n>We evaluate three reasoning models--Gemma-4-31B-IT, Qwen3.6-35B-A3B, and Claude Sonnet 4.6--on long-context adaptations of IHEval and IHChallenge.<n>The strongest monitor reduces rule-following non-compliance by 81-99%, with GPT-5.3 reductions of 86% under static attacks and 45% under adaptive attacks.
  arXiv  Detail & Related papers  (2026-06-05T19:36:48Z)
• LLMs Show No Signs Of Individuated Metacognition [0.023227405857540805]
  We decompose binary confidence judgements from 20 Large Language Models.<n>We ask whether two models that differ in confidence also differ in performance.<n>We find no evidence for significant verbalised individuated metacognition in any tested domain.
  arXiv  Detail & Related papers  (2026-05-22T23:54:33Z)
• Seirênes: Adversarial Self-Play with Evolving Distractions for LLM Reasoning [56.48520300004217]
  We present Seirnes, a self-play RL framework that transforms contextual interference into an internal training signal.<n>A single model is trained to both construct plausible yet distracting contexts that expose its own reasoning blind spots.<n>By pitting these competing objectives against each other, Seirnes compels the model to move beyond superficial pattern matching.
  arXiv  Detail & Related papers  (2026-05-12T06:58:35Z)
• What Single-Prompt Accuracy Misses: A Multi-Variant Reliability Audit of Language Models [0.0]
  Single-prompt accuracy is the dominant way to benchmark language models, but it can miss reliability failures that matter.<n>We evaluate a 15-model open-weight corpus, with the main reliability analyses focused on 10 instruct models across five classification and reasoning benchmarks.<n>We find three broad results. First, evaluation design can materially change the conclusion.<n>Second, confidence signals are fragile. On MMLU-Pro, every primary model verbally reports confidence substantially above both its accuracy and its token-probability confidence on the same rows, and verbal parse rate can collapse for a single model on a single prompt variant.
  arXiv  Detail & Related papers  (2026-05-03T20:05:08Z)
• Sensitivity Uncertainty Alignment in Large Language Models [0.0]
  We propose a framework for analyzing failures of large language models under adversarial and ambiguous inputs.<n>We define a scalar score, SUA_theta(x), capturing the difference between distributional sensitivity and predictive entropy.<n>We introduce SUA-TR, a training method combining consistency regularization and entropy alignment, along with an abstention rule for safer inference.
  arXiv  Detail & Related papers  (2026-04-21T17:53:12Z)
• On Calibration of Large Language Models: From Response To Capability [66.59139960234326]
  Large language models (LLMs) are widely deployed as general-purpose problem solvers.<n>We introduce capability calibration, which targets the model's expected accuracy on a query.<n>Our results demonstrate that capability-calibrated confidence improves pass@$k$ prediction and inference budget allocation.
  arXiv  Detail & Related papers  (2026-02-14T01:07:45Z)
• Beyond Memorization: Testing LLM Reasoning on Unseen Theory of Computation Tasks [8.210112631285666]
  Large language models (LLMs) have demonstrated strong performance on formal language tasks.<n>We introduce a benchmark for deterministic finite automata (DFA) construction from regular languages.<n>We show that models achieve perfect accuracy on factual questions and 84-90% on seen tasks, but accuracy drops sharply on unseen problems.
  arXiv  Detail & Related papers  (2026-01-19T21:00:31Z)
• Encyclo-K: Evaluating LLMs with Dynamically Composed Knowledge Statements [78.87065404966002]
  Existing benchmarks predominantly curate questions at the question level.<n>We propose Encyclo-K, a statement-based benchmark that rethinks benchmark construction from the ground up.
  arXiv  Detail & Related papers  (2025-12-31T13:55:54Z)
• SVeritas: Benchmark for Robust Speaker Verification under Diverse Conditions [54.34001921326444]
  Speaker verification (SV) models are increasingly integrated into security, personalization, and access control systems.<n>Existing benchmarks evaluate only subsets of these conditions, missing others entirely.<n>We introduce SVeritas, a comprehensive Speaker Verification tasks benchmark suite, assessing SV systems under stressors like recording duration, spontaneity, content, noise, microphone distance, reverberation, channel mismatches, audio bandwidth, codecs, speaker age, and susceptibility to spoofing and adversarial attacks.
  arXiv  Detail & Related papers  (2025-09-21T14:11:16Z)
• Can Large Language Models Always Solve Easy Problems if They Can Solve Harder Ones? [65.43882564649721]
  Large language models (LLMs) have demonstrated impressive capabilities, but still suffer from inconsistency issues. We develop the ConsisEval benchmark, where each entry comprises a pair of questions with a strict order of difficulty. We analyze the potential for improvement in consistency by relative consistency score.
  arXiv  Detail & Related papers  (2024-06-18T17:25:47Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.