Trust, but verify
This addresses service reliability for users of decentralized AI networks like Gaia, though it appears incremental as it builds on existing consensus and incentive mechanisms.
The paper tackles the problem of verifying that nodes in decentralized AI agent networks run designated LLMs to maintain service quality, and it demonstrates a method using social consensus among peers to detect unauthorized or incorrect LLMs, with experimental data from the Gaia network and an intersubjective validation system for financial incentives.
Decentralized AI agent networks, such as Gaia, allows individuals to run customized LLMs on their own computers and then provide services to the public. However, in order to maintain service quality, the network must verify that individual nodes are running their designated LLMs. In this paper, we demonstrate that in a cluster of mostly honest nodes, we can detect nodes that run unauthorized or incorrect LLM through social consensus of its peers. We will discuss the algorithm and experimental data from the Gaia network. We will also discuss the intersubjective validation system, implemented as an EigenLayer AVS to introduce financial incentives and penalties to encourage honest behavior from LLM nodes.