Foresight Arena: An On-Chain Benchmark for Evaluating AI Forecasting Agents
Maksym Nechepurenko, Pavel Shuvalov · 2026 · Preprint · SSRN / arXiv
Abstract
Evaluating the forecasting ability of AI agents requires environments resistant to overfitting, free from centralized trust, and grounded in incentive-compatible scoring. We introduce Foresight Arena, a permissionless on-chain benchmark for AI forecasting agents on real-world prediction markets. Agents submit probabilistic forecasts on binary Polymarket markets via a commit–reveal protocol enforced by Solidity smart contracts on Polygon PoS; outcomes resolve through the Gnosis Conditional Token Framework. Performance is measured by the Brier Score and an Alpha Score that isolates predictive edge over market consensus. We provide closed-form variance for per-market Alpha, a Murphy-decomposition connection, a power analysis, and a deterministic seed-controlled simulation. Live benchmark results are reserved for a future revision; the contracts and evaluation infrastructure are open-source.
Related tracks
Cite this work
@online{nechepurenko_shuvalov2026_foresightarena,
author = {Nechepurenko, Maksym and Shuvalov, Pavel},
title = {Foresight Arena: An On-Chain Benchmark for Evaluating AI Forecasting Agents},
date = {2026-04-29}, doi = {10.2139/ssrn.6674059}, url = {https://ssrn.com/abstract=6674059},
eprint = {2605.00420}, eprinttype = {arxiv}, pubstate = {preprint}
}