Summary
Quant-Arena (arXiv:2610.30113) is a benchmark and public leaderboard for evaluating whether large language models can complete the full quantitative research loop: proposing market hypotheses, implementing executable stock-scoring code, backtesting, and producing strategies that generalize beyond the research period. Built on real A-share market data, models develop features and scoring code using feedback only from 2022-2023 data, then freeze a single strategy that is evaluated on sealed 2024 data. A cash-account evaluator applies realistic trading constraints and costs, including market impact estimated from a simulated order book. Scores are primarily based on realized excess returns, complemented by signal quality and feature diversity. The authors evaluated nine frontier models; GPT-6 Astra achieved the highest score, yet no model outperformed an equal-weight market baseline on the 2024 out-of-sample data. Quant-Arena accepts open submissions from any model or agent framework.
Paper Overview
Research Area: NLP
Authors: Houde Qian, Zheng Jiang, Wenyi Li, Chenxi Xu, Xinjie Lin, Yuzhao Peng, Ruoyu Han, Youjie Zheng, Kaisen Yang, Qinhuai Na, You Fu, Calvin Xiao
Published: 2026-10-21
arXiv: 2610.30113
Abstract
Quantitative research is an iterative loop: a researcher proposes a market hypothesis, implements it as an executable signal, tests it on historical data, and decides what to keep. Existing benchmarks score formulas, code, or individual trading decisions, leaving open whether large language models (LLMs) can complete this loop and produce strategies that work beyond the research period.
The authors introduce Quant-Arena, a benchmark and public leaderboard for end-to-end quantitative research on real A-share market data:
- A model constructs features and stock-scoring code, receives feedback only on 2022-2023 data, and freezes one strategy for assessment on the sealed year 2024.
- A cash-account evaluator applies trading constraints and costs, including market impact estimated from a simulated order book.
- Scores are dominated by realized excess returns, supplemented by signal quality and the diversity of features used by the strategy.
The paper (per the Chinese summary) evaluated nine frontier models, with GPT-6 Astra scoring highest; however,
no model beat the equal-weight market in 2024. Quant-Arena is open to submissions from any model or agent framework.
*Auto-collected on 2026-10-23.*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178636344