Retrieval used as the task record failed almost every episode through omissions; verbatim top-k recall added to the full transcript helped but stayed below directives, Mem0 augmentation did not help, and an unverified agent-written ledger outperformed the always-accurate displayed checklist while still leaking re-executions.

arm: raw transcript · brief: amended · model: Qwen3-235B · steps: 10 · density: 0.15 · reasoning: no-think

arm: raw transcript · brief: amended · model: Qwen3-235B · steps: 10 · density: 0.15 · reasoning: no-think

arm: raw transcript · brief: amended · model: Qwen3-235B · steps: 10 · density: 0.15 · reasoning: no-think

arm: raw transcript · brief: amended · model: Qwen3-235B · steps: 15 · density: 0.3 · reasoning: no-think

arm: raw transcript · brief: amended · model: Qwen3-235B · steps: 15 · density: 0.3 · reasoning: no-think

arm: raw transcript · brief: amended · model: Qwen3-235B · steps: 15 · density: 0.3 · reasoning: no-think

arm: top-k retrieval, replace · brief: amended · model: Qwen3-235B · steps: 10 · density: 0.15 · reasoning: no-think

arm: top-k retrieval, replace · brief: amended · model: Qwen3-235B · steps: 10 · density: 0.15 · reasoning: no-think

arm: top-k retrieval, replace · brief: amended · model: Qwen3-235B · steps: 10 · density: 0.15 · reasoning: no-think

arm: top-k retrieval, replace · brief: amended · model: Qwen3-235B · steps: 10 · density: 0.15 · reasoning: no-think

arm: top-k retrieval, replace · brief: amended · model: Qwen3-235B · steps: 15 · density: 0.3 · reasoning: no-think

arm: top-k retrieval, replace · brief: amended · model: Qwen3-235B · steps: 15 · density: 0.3 · reasoning: no-think