Preserve the observed benchmark revision under a fixed tag while correcting co-author trailer formatting on the experiment branch. Runtime trees and recorded trial identities stay unchanged.
Co-authored-by: gensokyo-dev <[email protected]>
Co-authored-by: Kaoru0822 <[email protected]>
Use one policy for executor defaults, MCP admission and help. Make timeout default/maximum explicit, omit Code Mode limits from direct help, and describe the independent switch facet.
Validate 73 shared tests, both actual MCP catalogs, rejection recovery, and real zlib Evidence access after bounded guest failures. Model results remain pinned to earlier discovery contracts.
Co-authored-by: gensokyo-dev <[email protected]>
Co-authored-by: Kaoru0822 <[email protected]>
Both models retain correct selected-case answers and use one client wait. Record complete costs, timeout/output-budget recovery and mixed adapter outcomes without attributing a causal speedup.
Co-authored-by: gensokyo-dev <[email protected]>
Co-authored-by: Kaoru0822 <[email protected]>