StabilityBench: Benchmarking Instability in LLMs

• URL: http://arxiv.org/abs/2607.20558v1
• Date: Fri, 17 Jul 2026 16:09:37 GMT
• Title: StabilityBench: Benchmarking Instability in LLMs
• Abstract summary: We propose StabilityBench, a principled benchmark operator that transforms single-turn benchmark queries into multi-turn interaction histories.<n>We apply StabilityBench to four benchmarks spanning mathematical reasoning, health question-answering and safety, and evaluate nine large language models.<n>Our results show that model performance is consistently unstable under these injections, with considerable performance degradations on three out of four benchmarks studied.
• Score: 5.633712097504611
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: AI Assistants are increasingly deployed in high-stakes settings, such as healthcare or government services. Yet their real-world behavior remains poorly understood due to strong context dependence. Current evaluation protocols follow a defense-in-depth paradigm with compounding layers of safeguards, ranging from traditional benchmarks to live or adversarial testing. Such benchmarks remain largely static and single-turn, limiting their ability to capture real-world variability in conversational settings. We propose StabilityBench, a principled, general and model-agnostic benchmark operator that transforms single-turn benchmark queries into multi-turn interaction histories. StabilityBench augments existing benchmarks by injecting realistic user simulations, through demographic proxies or sycophantic baits, while preserving original task intent. We apply StabilityBench to four benchmarks spanning mathematical reasoning, health question-answering and safety, and evaluate nine large language models under these conditions. Our results show that model performance is consistently unstable under these injections, with considerable performance degradations on three out of four benchmarks studied. These highlight important limitations of static evaluations and motivate more realistic evaluation settings. To this end, we propose StabilityBench-Mini: a size-preserving variant of StabilityBench that samples across diversification axes, enabling more realistic evaluation without increasing costs.

Related papers

• ObsDriveBench: Benchmarking Multimodal Understanding under Adverse Weather with Observability Awareness [65.3988208273595]
  We introduce textbfObsDriveBench, a real-world multi-modal benchmark for adverse-weather autonomous driving.<n>Our benchmark is designed with three capability dimensions: textbfobservability awareness, textbfspatial reliability, and textbfrisk-aware decision-making.<n>Experiments reveal consistent performance degradation of existing vision-language models.
  arXiv  Detail & Related papers  (2026-07-26T08:13:46Z)
• Statistical Inference for Misspecified Contextual Bandits [6.178061357164435]
  We show a previously overlooked issue: standard algorithms may fail to stabilize under misspecified working models.<n>We develop an inverse-probability-weighted Z-estimation framework for a broad class of marginal moment targets.<n>We establish sufficient conditions for scaled inverse-propensity convergence for several policy classes.
  arXiv  Detail & Related papers  (2026-06-21T18:57:13Z)
• AgentNoiseBench: Benchmarking Robustness of Tool-Using LLM Agents Under Noisy Condition [72.24180896265192]
  We introduce AgentNoiseBench, a framework for evaluating robustness of agentic models under noisy environments.<n>We first conduct an in-depth analysis of biases and uncertainties in real-world scenarios.<n>We then categorize environmental noise into two primary types: user-noise and tool-noise.<n>Building on this analysis, we develop an automated pipeline that injects controllable noise into existing agent-centric benchmarks.
  arXiv  Detail & Related papers  (2026-02-11T20:33:10Z)
• Not All Preferences Are Created Equal: Stability-Aware and Gradient-Efficient Alignment for Reasoning Models [52.48582333951919]
  We propose a dynamic framework designed to enhance alignment reliability by maximizing the Signal-to-Noise Ratio of policy updates.<n>SAGE (Stability-Aware Gradient Efficiency) integrates a coarse-grained curriculum mechanism that refreshes candidate pools based on model competence.<n> Experiments on multiple mathematical reasoning benchmarks demonstrate that SAGE significantly accelerates convergence and outperforms static baselines.
  arXiv  Detail & Related papers  (2026-02-01T12:56:10Z)
• ReasonBENCH: Benchmarking the (In)Stability of LLM Reasoning [2.1461777157838724]
  We introduce ReasonBENCH, the first benchmark designed to quantify the underlying instability in large language models (LLMs) reasoning.<n>Across tasks from different domains, we find that the vast majority of reasoning strategies and models exhibit high instability.<n>We further analyze the impact of prompts, model families, and scale on the trade-off between solve rate and stability.
  arXiv  Detail & Related papers  (2025-12-08T18:26:58Z)
• 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)
• Prompt Stability in Code LLMs: Measuring Sensitivity across Emotion- and Personality-Driven Variations [40.12950482269347]
  We present PromptSE, a framework that creates semantically equivalent prompt variants with emotion and personality templates.<n>Our study shows that performance and stability behave as largely decoupled optimization objectives.<n>PromptSE enables practitioners to quantify performance stability trade offs for deployment and model selection.
  arXiv  Detail & Related papers  (2025-09-17T04:17:42Z)
• CNS-Bench: Benchmarking Image Classifier Robustness Under Continuous Nuisance Shifts [78.79936076607373]
  We introduce CNS-Bench, a Continuous Nuisance Shift Benchmark to quantify robustness of image classifiers for continuous and realistic nuisance shifts.<n>We propose a filtering mechanism that outperforms previous methods, thereby enabling reliable benchmarking with generative models.
  arXiv  Detail & Related papers  (2025-07-23T16:15:48Z)
• A Controllable Examination for Long-Context Language Models [62.845852724511964]
  This study introduces $textbfLongBioBench, a benchmark for evaluating long-context language models.<n>We show that most models still exhibit deficiencies in semantic understanding and elementary reasoning over retrieved results.<n>Our further analysis indicates some design choices employed by existing synthetic benchmarks, such as contextual non-coherence.
  arXiv  Detail & Related papers  (2025-06-03T14:23:06Z)
• A Comparative Analysis on Ethical Benchmarking in Large Language Models [0.0]
  This work contributes to the field of Machine Ethics (ME) benchmarking, which develops tests to assess whether intelligent systems accurately represent human values and act accordingly. We identify three major issues with current ME benchmarks: limited ecological validity due to unrealistic ethical dilemmas, unstructured question generation without clear inclusion/exclusion criteria, and a lack of scalability due to reliance on human annotations. We introduce two new ME benchmarks: the Triage Benchmark and the Medical Law (MedLaw) Benchmark, both featuring real-world ethical dilemmas from the medical domain.
  arXiv  Detail & Related papers  (2024-10-11T05:05:21Z)
• Position: AI Evaluation Should Learn from How We Test Humans [65.36614996495983]
  We argue that psychometrics, a theory originating in the 20th century for human assessment, could be a powerful solution to the challenges in today's AI evaluations.
  arXiv  Detail & Related papers  (2023-06-18T09:54:33Z)

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.