LLM Advertisement based on Neuron Auctions

2026-05-08

AI Summary

This work addresses the challenge of integrating advertisements into large language models while jointly optimizing advertiser utility, platform revenue, and user experience—a balance often disrupted by existing approaches that degrade textual coherence and lack controllable intervention mechanisms. The authors propose Neuron Auctions, a novel paradigm that introduces ad auctions directly into the neuron activation space of LLMs. Leveraging mechanistic interpretability, the method identifies brand-specific feedforward neurons and exploits their near-orthogonality to enable decoupled interventions. A continuous menu-based auction mechanism is designed, incorporating a user utility penalty term to align commercial incentives with conversational naturalness. Experimental results demonstrate that this approach effectively harmonizes business objectives with user experience without compromising dialogue fluency, achieving a Pareto improvement across all stakeholders.

Abstract

As Large Language Models (LLMs) transition into conversational agents, generative advertising emerges as a crucial monetization strategy. However, embedding advertisements within unstructured LLM outputs introduces a critical trilemma: balancing advertiser payoffs, platform revenue, and user experience. Existing methods, such as prompt injection or rigid position slots, disrupt semantic coherence and lack a parametric framework for independent control, rendering rigorous mechanism design intractable. To bridge this gap, we introduce Neuron Auctions, a novel paradigm that shifts the auction object from the surface text space to the LLM's internal representations. Leveraging mechanistic interpretability, we identify brand-specific feed-forward network (FFN) neurons and demonstrate that competing brands activate within approximately orthogonal subspaces. This near-perfect independence allows us to define continuous, disentangled intervention budgets (specifically, neuron counts and amplification factors) as auctionable commodities. Building on this computational carrier, we design a continuous menu-based auction mechanism that naturally guarantees strategy-proofness and optimizes revenue for the platform. By explicitly incorporating a user utility penalty into the platform's optimization objective, our framework dynamically prices out overly aggressive interventions. Extensive experiments demonstrate that Neuron Auctions effectively preserve natural discourse quality while achieving an optimal alignment between commercial incentives and user satisfaction.

Problem

• generative advertising
• Large Language Models
• user experience
• advertiser payoffs
• platform revenue

Innovation

• Neuron Auctions
• mechanistic interpretability
• orthogonal subspaces
• continuous auction mechanism
• LLM advertising