{"type":"post","id":"structured-prompts-improve-evaluation-of-language-models-f6c42b","slug":"structured-prompts-improve-evaluation-of-language-models-f6c42b","title":"Structured Prompts Improve Evaluation of Language Models","description":"A study combining DSPy’s structured prompting methods with HELM finds that prompt choice can materially change language-model benchmark results.","retrieval_nugget":"A study combining DSPy’s structured prompting methods with HELM finds that prompt choice can materially change language-model benchmark results. Across six models and seven benchmarks, structured prompting improved scores by 6% on average and shifted rankings on five benchmarks, making a single static prompt an unreliable proxy for model capability alone.","published_at":"2026-11-06","updated_at":"2026-11-06","record_date":"2026-11-06","date_kind":"published_at","topics":["models"],"entities":["DSPy","HELM","Stanford University","Center for Research on Foundation Models"],"related_patterns":["ai-engineering-evaluations-are-splitting-into-task-specific"],"novelty_label":"new_fact","what_changed":"A new DSPy+HELM study finds structured prompting improves scores by 6% on average and shifts rankings on 5 of 7 benchmarks; earlier priors cover a DSPy release and its use in Dropbox evaluations.","prior_refs":["https://newruntime.com/posts/dspy-3-4-0-out-with-1-native-4439f5/"],"source_url":"https://arxiv.org/abs/2511.20836","source_urls":["https://arxiv.org/abs/2511.20836"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:structured-prompts-improve-evaluation-of-language-models-f6c42b","status":"published","visuals":[],"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/structured-prompts-improve-evaluation-of-language-models-f6c42b/","markdown":"https://newruntime.com/posts/structured-prompts-improve-evaluation-of-language-models-f6c42b.md","json":"https://newruntime.com/posts/structured-prompts-improve-evaluation-of-language-models-f6c42b.json"},"markdown_url":"https://newruntime.com/posts/structured-prompts-improve-evaluation-of-language-models-f6c42b.md","next_reads":[{"type":"topic","path":"/topics/models/","url":"https://newruntime.com/topics/models/","title":"Models","reason":"Explore the models topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/models.md","json_url":"https://newruntime.com/topics/models.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/sources/","url":"https://newruntime.com/sources/","title":"Sources","reason":"See the original sources this archive cites.","media_type":"text/html"}]}