AutoXiv
ExploreAskLearnReproduceMarketplaceAlpha MachineDocs
Submit a paper
AutoXiv

An open preprint server for the AI era. Structured overviews, pinned code, and reproduction records on every paper.

Platform

  • Explore papers
  • Browse by category
  • Learn
  • AI results tracker
  • Reproduction agent
  • Bounty marketplace
  • Alpha Machine
  • Submit a paper

Developers

  • MCP server
  • API keys
  • Open calls
  • Fast read

AutoXiv

  • About
  • For labs and companies
  • Transparency stats
  • Changelog
  • RSS feed
  • Terms
  • Privacy
  • X / Twitter
© 2026 AutoXiv · Open preprint infrastructureFree to submit · No APCs · Agent-readable over MCP
Alpha Machine

From paper to strategy scaffold

Screen a paper for tradable signals, find existing code, and draft a long-short strategy spec with a backtest scaffold — as a conversation. Most papers carry no alpha, and the screen says so. Research tooling, not investment advice.

What do you want to research?

or paste an arXiv link in the text
screen for alpha · find code · design the strategy · scaffold the backtest · run it in a sandbox — deliverables as .py / .ipynb / .md
Market context
charts by TradingView ↗

Your runs

Loading…

How it screens

DIRECTPaper is about return prediction — signal lifts nearly as-is.
INDIRECTMethod maps to markets through a concrete bridge.
INFRASTRUCTUREImproves the pipeline; carries no signal of its own.
NO ALPHANo defensible path to a trading signal.
The 0–100 score is a single model judgment against fixed rubric bands — direct 55–85 · indirect 25–55 · infrastructure 5–25 · none 0–10 — estimating whether a competent quant could extract net-of-costs alpha within ~3 months. A viability screen, not an audited backtest.

Disclaimer

Alpha Machine outputs are research scaffolds generated from paper metadata. Published anomalies decay, backtests overfit, and synthetic placeholder data proves nothing — a scaffold that executes in the sandbox is shown to run, not to work. Nothing here is investment advice.