EcoAlign: An Economically Rational Framework for Efficient LVLM Alignment

TL;DR

EcoAlign introduces an inference-time, economically rational framework for LVLM alignment that balances safety, utility, and cost by dynamically expanding thought graphs and scoring actions, improving safety and efficiency.

Contribution

EcoAlign presents a novel, inference-time alignment method that models LVLMs as boundedly rational agents, optimizing safety and utility within computational budgets.

Findings

EcoAlign matches or surpasses state-of-the-art safety and utility.

EcoAlign achieves lower computational costs compared to existing methods.

EcoAlign effectively prevents deception through the weakest-link safety enforcement.

Abstract

Large Vision-Language Models (LVLMs) exhibit powerful reasoning capabilities but suffer sophisticated jailbreak vulnerabilities. Fundamentally, aligning LVLMs is not just a safety challenge but a problem of economic efficiency. Current alignment methods struggle with the trade-off between safety, utility, and operational costs. Critically, a focus solely on final outputs (process-blindness) wastes significant computational budget on unsafe deliberation. This flaw allows harmful reasoning to be disguised with benign justifications, thereby circumventing simple additive safety scores. To address this, we propose EcoAlign, an inference-time framework that reframes alignment as an economically rational search by treating the LVLM as a boundedly rational agent. EcoAlign incrementally expands a thought graph and scores actions using a forward-looking function (analogous to net present value) that dynamically weighs expected safety, utility, and cost against the remaining budget. To prevent deception, path safety is enforced via the weakest-link principle. Extensive experiments across 3 closed-source and 2 open-source models on 6 datasets show that EcoAlign matches or surpasses state-of-the-art safety and utility at a lower computational cost, thereby offering a principled, economical pathway to robust LVLM alignment.