analiza aiqbench
The paper introduces a benchmark that tests whether LLM sales agents act on the policies they can state, scoring tool-verified actions rather than rule recital or task completion. Its headline result is a large gap between static policy compliance (roughly 90 to 99%) and grounded action compliance (3 to 17%) across eight backbones, though the authors note the gap also reflects safe-termination requirements, not pressure alone. Readers building or evaluating deployed agents should note that committed-action metrics can reveal failures that standard policy QA misses.
R2A-Sales: A Policy-Atom-Matched Benchmark for Rule-to-Action Gaps in Goal-Directed Sales Agents
De la arXiv
Text integralarXiv:2610.34185v1 Announce Type: new Abstract: LLM-based sales agents are often evaluated on static policy questions or task success, but deployment asks a sharper question: when a model can state a rule, will it choose compliant, tool-grounded actions under commercial pressure? We introduce R2A-Sales (Rule-to-Action), a policy-atom-matched benchmark for this question. Each item pairs a static probe with a multi-turn sales scenario sharing the same versioned policy atom, facts, evidence, behavioral constraints, and tool contract, so differences cannot be attributed to policy-content mismatch.
In the interactive arm, a code-owned ledger verifies committed tool effects and terminal states; unsupported action claims receive no credit. We report static compliance $K$, grounded action compliance $A=\Pr(\text{safe terminal}\land\text{no hard violation}\land\text{non-echo})$, and the matched gap $G=K-A$. The frozen release spans three controlled Chinese-language sales domains, 24 policy atoms, 42 semantic cores, and 2,016 matched pairs. The confirmatory study audits 21,888 runs over 19 cores, covering eight backbones plus four Qwen3-8B prompt treatments (12 configurations).
Static compliance is near ceiling $(K=89.9\%\text{--}99.1\%)$, yet eight-backbone action compliance is only $A=3.0\%\text{--}17.3\%$, giving $G=80.4\text{--}92.5$ percentage points. Since $A$ also requires safe termination, $G$ upper-bounds rule-to-action degradation rather than isolating pressure alone. No prompt-only treatment dominates all metrics. R2A-Sales shows that sales-agent reliability must be measured at committed action, not rule recital or task completion alone.
Abstract:LLM-based sales agents are often evaluated on static policy questions or task success, but deployment asks a sharper question: when a model can state a rule, will it choose compliant, tool-grounded actions under commercial pressure? We introduce R2A-Sales (Rule-to-Action), a policy-atom-matched benchmark for this question. Each item pairs a static probe with a multi-turn sales scenario sharing the same versioned policy atom, facts, evidence, behavioral constraints, and tool contract, so differences cannot be attributed to policy-content mismatch.
In the interactive arm, a code-owned ledger verifies committed tool effects and terminal states; unsupported action claims receive no credit. We report static compliance $K$, grounded action compliance $A=\Pr(\text{safe terminal}\land\text{no hard violation}\land\text{non-echo})$, and the matched gap $G=K-A$. The frozen release spans three controlled Chinese-language sales domains, 24 policy atoms, 42 semantic cores, and 2,016 matched pairs. The confirmatory study audits 21,888 runs over 19 cores, covering eight backbones plus four Qwen3-8B prompt treatments (12 configurations).
Static compliance is near ceiling $(K=89.9\%\text{--}99.1\%)$, yet eight-backbone action compliance is only $A=3.0\%\text{--}17.3\%$, giving $G=80.4\text{--}92.5$ percentage points. Since $A$ also requires safe termination, $G$ upper-bounds rule-to-action degradation rather than isolating pressure alone. No prompt-only treatment dominates all metrics. R2A-Sales shows that sales-agent reliability must be measured at committed action, not rule recital or task completion alone.
25 pages, including 16 pages of supplementary material. Both authors contributed equally
Artificial Intelligence (cs.AI)
arXiv:2610.34185 [cs.AI]
(or arXiv:2610.34185v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.34185 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 314 cuvinte