{"type":"post","id":"mid-harness-the-source-of-the-terminalbench-lite-228a6b","slug":"mid-harness-the-source-of-the-terminalbench-lite-228a6b","title":"Mid-Harness, the source of the TerminalBench-Lite result, changes how the harness selects actions.","description":"A paper introducing Mid-Harness tests whether terminal agents can improve by sampling several possible actions and verifying them before execution.","retrieval_nugget":"A paper introducing Mid-Harness tests whether terminal agents can improve by sampling several possible actions and verifying them before execution. On TerminalBench-Lite, the authors report Pass@1 rising from 50.00% to 68.03% when a GPT-5.6 Sol verifier selects among eight actions generated by TMAX-9B. The result makes action selection a measurable harness-design question.","published_at":"2026-11-12","updated_at":"2026-11-12","record_date":"2026-11-12","date_kind":"published_at","topics":["agent-harness","harness-engineering","verification"],"entities":["Mid-Harness","TerminalBench-Lite","TMAX-9B","GPT-5.6 Sol","Minki Kang"],"related_patterns":["harness-architecture-outlives-model-choice"],"novelty_label":"new_fact","what_changed":"Mid-Harness reports that verifying eight sampled actions raises TerminalBench-Lite Pass@1 from 50.00% to 68.03%, showing how action selection can improve terminal-agent reliability.","prior_refs":["https://newruntime.com/posts/our-harnesstax-analysis-examines-how-harness-choice-changes-80954d/"],"source_url":"https://arxiv.org/abs/2609.39982","source_urls":["https://arxiv.org/abs/2609.39982"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:mid-harness-the-source-of-the-terminalbench-lite-228a6b","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/mid-harness-the-source-of-the-terminalbench-lite-228a6b/","markdown":"https://newruntime.com/posts/mid-harness-the-source-of-the-terminalbench-lite-228a6b.md","json":"https://newruntime.com/posts/mid-harness-the-source-of-the-terminalbench-lite-228a6b.json"},"markdown_url":"https://newruntime.com/posts/mid-harness-the-source-of-the-terminalbench-lite-228a6b.md","next_reads":[{"type":"topic","path":"/topics/agent-harness/","url":"https://newruntime.com/topics/agent-harness/","title":"Agent Harness","reason":"Explore the agent-harness topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/agent-harness.md","json_url":"https://newruntime.com/topics/agent-harness.json"},{"type":"topic","path":"/topics/harness-engineering/","url":"https://newruntime.com/topics/harness-engineering/","title":"Harness Engineering","reason":"Explore the harness-engineering topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/harness-engineering.md","json_url":"https://newruntime.com/topics/harness-engineering.json"},{"type":"topic","path":"/topics/verification/","url":"https://newruntime.com/topics/verification/","title":"Verification","reason":"Explore the verification topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/verification.md","json_url":"https://newruntime.com/topics/verification.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"}]}