Qwen3.6-35B text-protocol re-execution counts are an upper bound because many were booking-line citations after a correct refusal, a behavior absent in the tool harness; the tool harness used tools that never fail, tau2-bench covered only airline with an internal simulator, and PM-Bench covered one released week. · CiteArk
Not assessedNo independent reproduction scheduledLimitationclaim-limitation-protocol-transfer
Qwen3.6-35B text-protocol re-execution counts are an upper bound because many were booking-line citations after a correct refusal, a behavior absent in the tool harness; the tool harness used tools that never fail, tau2-bench covered only airline with an internal simulator, and PM-Bench covered one released week.
Source: paper:PDF p. 9, Section 6
Reported and observed measurements
No structured measurement is attached to this Claim.
Assessments (0)
No immutable Assessment has been published for this Claim yet.
