analiza aiqbench
This research introduces XYBench, a new benchmark evaluating large language models' ability to detect user misconceptions and offer pragmatic solutions rather than literal answers. The study reveals a significant capability gap where models frequently fail to identify errors or redirect users toward the root problem, despite showing theoretical preference for such responses in multiple-choice settings. These findings highlight that robust error correction and intent alignment remain underdeveloped areas in current AI systems.
XYBench: Can LLMs Respond Pragmatically to Queries with Misconceptions?
arXiv:2609.06842v1 Announce Type: new Abstract: When non-expert users ask LLMs for assistance, their queries can often have misconceptions (e.g., "How do I parse XML with regex?"). In such cases, often referred to as the XY-problem, LLMs must identify the misconception ("regex are fragile") and meaningfully direct the user toward a pragmatic solution that will address the root problem implicit in the request ("use an XML parser"). We introduce XYBench, a benchmark of 8,115 such queries, drawn from technical (StackOverflow/StackExchange) and everyday (WikiHow and a manually-curated subset) domains. We design an evaluation paradigm that assesses model responses along three criteria grounded in cooperative response theory: (a) presence and (b) emphasis on pragmatic solutions, and (c) identification of misconceptions. Our experiments show that even the strongest LLMs predominantly answer the literal request (0.75--0.92) and far less often the intended one (0.33--0.71), while substantially lagging behind humans at identifying misconceptions (at most 63% vs. 79--90%). Further, models overwhelmingly prefer pragmatic responses in a multiple choice setting yet consistently fail to generate them. Oracle ablation experiments show that providing explicit user intent at generation time helps; however a large gap remains, suggesting pragmatic redirection is a fundamentally underdeveloped capability in current LLMs.
Title:XYBench: Can LLMs Respond Pragmatically to Queries with Misconceptions?
Abstract:When non-expert users ask LLMs for assistance, their queries can often have misconceptions (e.g., "How do I parse XML with regex?"). In such cases, often referred to as the XY-problem, LLMs must identify the misconception ("regex are fragile") and meaningfully direct the user toward a pragmatic solution that will address the root problem implicit in the request ("use an XML parser"). We introduce XYBench, a benchmark of 8,115 such queries, drawn from technical (StackOverflow/StackExchange) and everyday (WikiHow and a manually-curated subset) domains. We design an evaluation paradigm that assesses model responses along three criteria grounded in cooperative response theory: (a) presence and (b) emphasis on pragmatic solutions, and (c) identification of misconceptions. Our experiments show that even the strongest LLMs predominantly answer the literal request (0.75--0.92) and far less often the intended one (0.33--0.71), while substantially lagging behind humans at identifying misconceptions (at most 63% vs. 79--90%). Further, models overwhelmingly prefer pragmatic responses in a multiple choice setting yet consistently fail to generate them. Oracle ablation experiments show that providing explicit user intent at generation time helps; however a large gap remains, suggesting pragmatic redirection is a fundamentally underdeveloped capability in current LLMs.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2609.06842 [cs.CL]
(or arXiv:2609.06842v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.06842 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 300 cuvinte
Subiecte conexe
Presentation: Beyond Prompting: Context Engineering for Production-Grade AI
InfoQ · 2 sept. 2026
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · 13 oct. 2026
Security and Privacy Prompts in the Wild: What Users Ask LLMs and How LLMs Respond
arXiv · 17 iun. 2026
Runway’s WorldPrompt and the Engineering of Real-Time Worlds
Latent Space · 25 sept. 2026