Paying for Honesty Without Knowing the Truth: Reputation-Penalty Design for LLM Marketplace Agents
For platforms hosting LLM merchants, this provides a practical mechanism to curb dishonesty without needing ground truth, with demonstrated welfare improvements.
The paper addresses the problem of LLM agents fabricating product attributes in listings and proposes CARP, a reputation-penalty mechanism that works without ground truth, plus SPARC, a code-gated reflection mechanism. CARP suppresses sales of low-rated liars and, with SPARC, closes most of the consumer-welfare gap relative to a perfect-information oracle, achieving the best welfare among compared policies.
LLM agents increasingly act as autonomous merchants that write their own product listings, and under competitive pressure, they fabricate attributes to win sales. Even under instructions to be honest, they fabricate attributes in a majority of listings across models. A platform's obvious remedy---verifying each claim against the truth---is unavailable, because it observes only a noisy, biased complaint signal, never the ground truth. We design CARP, a reputation-penalty mechanism with a deadband that forgives complaint noise and a state-dependent severity that counters reputation-driven detection erosion. CARP requires no product-level ground truth and is robust to strategic gaming. CARP protects consumers by suppressing the sales volume of low-rated liars while sparing honest sellers. Paired with SPARC, it closes most of the consumer-welfare gap relative to a perfect-information oracle, without ever accessing the truth. It also achieves the best welfare of the policies we compare. We further show that this felt penalty becomes behaviorally binding through SPARC, a byte-clean code-gated reflection mechanism: LLM merchants fabricate when lying is free but restrain themselves when fabrication costs them sales, a self-interested response rather than compliance. We trace this distinction to penalty-gated self-correction reasoning, and observe the binding across models, with supporting confidence intervals.