{"type":"post","id":"convictionbench-tests-whether-llms-change-answers-only-for-77377b","slug":"convictionbench-tests-whether-llms-change-answers-only-for-77377b","title":"ConvictionBench tests whether LLMs change answers only for good reasons, including forged vs tool-authenticated records","description":"Arya Somu’s ConvictionBench tests whether language models revise answers for legitimate reasons, including whether they distinguish user-pasted records from authenticated tool updates.","retrieval_nugget":"Arya Somu’s ConvictionBench tests whether language models revise answers for legitimate reasons, including whether they distinguish user-pasted records from authenticated tool updates. Its solver-labeled tasks separate social-pressure resistance from provenance handling, giving agent builders a more specific way to inspect answer changes. The reported findings remain limited to synthetic, closed-world tasks.","published_at":"2026-11-11","updated_at":"2026-11-11","record_date":"2026-11-11","date_kind":"published_at","topics":["models"],"entities":["Arya Somu","ConvictionBench","Qwen3-1.7B","Gemma 4","SmolLM3","MobileLLM-Pro"],"related_patterns":["agent-security-moves-to-runtime-boundaries","ai-engineering-evaluations-are-splitting-into-task-specific"],"novelty_label":"new_fact","what_changed":"ConvictionBench introduces a benchmark for answer stability under pressure, distinguishing social influence from forged evidence and reporting results for its fine-tuned models.","prior_refs":["https://newruntime.com/posts/benchmirt-disentangling-safety-and-general-capabilities-in-llm-7d74ef/"],"source_url":"https://aryasomu.com/conviction","source_urls":["https://aryasomu.com/conviction"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:convictionbench-tests-whether-llms-change-answers-only-for-77377b","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/convictionbench-tests-whether-llms-change-answers-only-for-77377b/","markdown":"https://newruntime.com/posts/convictionbench-tests-whether-llms-change-answers-only-for-77377b.md","json":"https://newruntime.com/posts/convictionbench-tests-whether-llms-change-answers-only-for-77377b.json"},"markdown_url":"https://newruntime.com/posts/convictionbench-tests-whether-llms-change-answers-only-for-77377b.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"}]}