{"slug": "the-wording-effect-quantifying-two-way-drift-in-llm-benchmark-performance", "title": "The Wording Effect: Quantifying Two-Way Drift in LLM Benchmark Performance", "summary": "A new study introducing BenchDrift shows that rephrasing benchmark problems while keeping meaning and answer fixed flips LLM correctness in both directions across eight models and three benchmarks (GSM8K, MMLU, MATH-Hard). The authors find that phrasing sensitivity does not fade as models improve but changes sign: weak models gain more from rephrasing than they lose, while strong models lose far more than they gain, making the best models the most wording-dependent. The study also finds that models largely agree on which rephrasings cost the most correct answers, indicating fragility belongs to the rephrasing rather than the model.", "body_md": "# Computer Science > Computation and Language\n\n[Submitted on 12 Aug 2026]\n\n# Title:The Wording Effect: Quantifying Two-Way Drift in LLM Benchmark Performance\n\n[View PDF](/pdf/2608.11694)\n\n[HTML (experimental)](https://arxiv.org/html/2608.11694v1)\n\nAbstract:A benchmark score comes from a single phrasing of each problem. That single phrasing is treated as if it stood for the whole space of ways the same problem could be asked, but it does not. We show that rephrasing a problem while keeping its meaning and answer fixed routinely flips a model's answer in both directions, so some failures become successes and some successes become failures. We call this drift. BenchDrift generates meaning-preserving variations of benchmark problems along four axes, namely linguistic, referential, pragmatic, and structural, and measures how often, and why, correctness flips under each. Across eight models and three benchmarks (GSM8K, MMLU, MATH-Hard), we observe that drift is large in both directions. Two findings stand out. First, phrasing sensitivity does not fade as models get better. Instead, it changes sign. Weak models gain more from rephrasing than they lose, while strong models lose far more than they gain. We find that the best models on a benchmark are therefore the ones whose scores depend most on the wording they happened to be given. Second, the models largely agree on which rephrasings cost the most correct answers even though they differ in how much they drift, so fragility belongs to the rephrasing and not to the model. Furthermore, rephrasing breaks answers a model was confident about, whether the problem is made shorter or longer. Code and Data:[this https URL]\n\n### References & Citations\n\nLoading...\n\n# Bibliographic and Citation Tools\n\nBibliographic Explorer\n\n*(*[What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))\nConnected Papers\n\n*(*[What is Connected Papers?](https://www.connectedpapers.com/about))\nLitmaps\n\n*(*[What is Litmaps?](https://www.litmaps.co/))\nscite Smart Citations\n\n*(*[What are Smart Citations?](https://www.scite.ai/))# Code, Data and Media Associated with this Article\n\nalphaXiv\n\n*(*[What is alphaXiv?](https://alphaxiv.org/))\nCatalyzeX Code Finder for Papers\n\n*(*[What is CatalyzeX?](https://www.catalyzex.com))\nDagsHub\n\n*(*[What is DagsHub?](https://dagshub.com/))\nGotit.pub\n\n*(*[What is GotitPub?](http://gotit.pub/faq))\nHugging Face\n\n*(*[What is Huggingface?](https://huggingface.co/huggingface))\nScienceCast\n\n*(*[What is ScienceCast?](https://sciencecast.org/welcome))# Demos\n\n# Recommenders and Search Tools\n\nInfluence Flower\n\n*(*[What are Influence Flowers?](https://influencemap.cmlab.dev/))\nCORE Recommender\n\n*(*[What is CORE?](https://core.ac.uk/services/recommender))# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth 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.\n\nHave an idea for a project that will add value for arXiv's community? [ Learn more about arXivLabs](https://info.arxiv.org/labs/index.html).", "url": "https://wpnews.pro/news/the-wording-effect-quantifying-two-way-drift-in-llm-benchmark-performance", "canonical_source": "https://arxiv.org/abs/2608.11694", "published_at": "2026-08-13 07:07:07+00:00", "updated_at": "2026-08-13 07:40:55.163646+00:00", "lang": "en", "topics": ["large-language-models", "ai-research"], "entities": ["BenchDrift", "GSM8K", "MMLU", "MATH-Hard"], "alternates": {"html": "https://wpnews.pro/news/the-wording-effect-quantifying-two-way-drift-in-llm-benchmark-performance", "markdown": "https://wpnews.pro/news/the-wording-effect-quantifying-two-way-drift-in-llm-benchmark-performance.md", "text": "https://wpnews.pro/news/the-wording-effect-quantifying-two-way-drift-in-llm-benchmark-performance.txt", "jsonld": "https://wpnews.pro/news/the-wording-effect-quantifying-two-way-drift-in-llm-benchmark-performance.jsonld"}}