Computational Economics in Large Language Models: Exploring Model Behavior and Incentive Design under Resource Constraints

TL;DR

This paper introduces a 'computational economics' framework for large language models, demonstrating how resource constraints influence model behavior and proposing an incentive-based training method to improve efficiency and interpretability.

Contribution

It presents a novel economic perspective on LLM resource allocation and develops a training paradigm that enhances efficiency and transparency under computational constraints.

Findings

Models reallocate attention to high-value tokens under scarcity

Proposed method reduces FLOPS by ~40% at similar accuracy

Models show more interpretable attention patterns

Abstract

Large language models (LLMs) are limited by substantial computational cost. We introduce a "computational economics" framework that treats an LLM as an internal economy of resource-constrained agents (attention heads and neuron blocks) that must allocate scarce computation to maximize task utility. First, we show empirically that when computation is scarce, standard LLMs reallocate attention toward high-value tokens while preserving accuracy. Building on this observation, we propose an incentive-driven training paradigm that augments the task loss with a differentiable computation cost term, encouraging sparse and efficient activations. On GLUE (MNLI, STS-B, CoLA) and WikiText-103, the method yields a family of models that trace a Pareto frontier and consistently dominate post-hoc pruning; for a similar accuracy we obtain roughly a forty percent reduction in FLOPS and lower latency, together with more interpretable attention patterns. These results indicate that economic principles offer a principled route to designing efficient, adaptive, and more transparent LLMs under strict resource constraints.