On paired tau2-bench airline runs, a compiled tool-dispatch gate raised Qwen3-235B pass^1 from 0.39 to 0.54 and eliminated executed rule violations, but changed little for Qwen3.6-35B because that model rarely attempted state-decidable violations; state shown and overridable notices lay between on pass^1 but leaked violations. · CiteArk

On paired tau2-bench airline runs, a compiled tool-dispatch gate raised Qwen3-235B pass^1 from 0.39 to 0.54 and eliminated executed rule violations, but changed little for Qwen3.6-35B because that model rarely attempted state-decidable violations; state shown and overridable notices lay between on pass^1 but leaked violations. · CiteArk

测量结果

已评估 0/87 项

获得支持 0有争议或冲突 0尚未评估 87

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
arm: ungated · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	0.39 proportion	—	尚未评估
arm: ungated · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	0.54 proportion	—	尚未评估
arm: ungated · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	9 episodes	—	尚未评估
arm: ungated · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	35 episodes	—	尚未评估
arm: ungated · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	10 episodes	—	尚未评估
arm: ungated · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	26 writes	—	尚未评估
arm: ledger (state shown) · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	0.52 proportion	—	尚未评估
arm: ledger (state shown) · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	0.38 proportion	—	尚未评估
arm: ledger (state shown) · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	14 episodes	—	尚未评估
arm: ledger (state shown) · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	14 episodes	—	尚未评估
arm: ledger (state shown) · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	10 episodes	—	尚未评估
arm: ledger (state shown) · model: Qwen3-235B · domain: tau2-bench airline · episodes: 100	4 writes	—	尚未评估