Gradients: When Markets Meet Fine-tuning -- A Distributed Approach to Model Optimisation

TL;DR

This paper introduces Gradients, a decentralized, competitive approach to model fine-tuning that leverages market dynamics to outperform traditional AutoML platforms across various tasks and models.

Contribution

The paper presents a novel distributed optimization method using market-based competition among miners, significantly improving fine-tuning performance over existing AutoML solutions.

Findings

Gradients achieved a 100% win rate against major AutoML platforms.

Mean performance improvements of 42.1% over commercial AutoML systems.

Significant gains in retrieval-augmented generation and diffusion models.

Abstract

Current AutoML platforms leave substantial performance untapped. Testing 180 fine-tuning tasks across models from 70M to 70B parameters, we found that HuggingFace AutoTrain, TogetherAI, Databricks, and Google Cloud consistently produce suboptimal configurations. Gradients, built on the Bittensor network, attacks this problem through competition. Independent miners race to find optimal hyperparameters, earning rewards proportional to their models' performance. This tournament drives exploration of configuration spaces that single-strategy methods never examine. In our experiments, Gradients achieved a 100\% win rate against TogetherAI, Databricks, and Google Cloud, and beat HuggingFace AutoTrain in 82.8\% of experiments. Mean improvements reached 42.1\% against commercial platforms. Retrieval-augmented generation tasks saw 30-40\% gains; diffusion models improved 23.4\% on person-specific generation. When miners compete for rewards, they develop optimization strategies that centralized approaches overlook. These findings demonstrate that decentralized systems with economic incentives can systematically outperform traditional AutoML, suggesting market dynamics may be key to achieving superior fine-tuning results. Code is available at https://github.com/rayonlabs/G.O.D.