Reproduce “DSA: Evidence-Aware LLM-Agent Orchestration for Multi-Market Stock Research” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Linsen Zhu, Yi Shi
DSA organizes stock research into a five-stage workflow for acquiring market evidence, building structured context, routing requests to compatible models, and producing reports through either a default or optional agent-assisted profile. In the agentic path, eligible Strategy Skill opinions inform synthesis, while RiskGuard can preserve or reduce aggressiveness. In tests, 1,457 cases passed; this verifies contracts, not report quality or returns.
Key insights
- Evidence availability is treated as a first-class input rather than allowing absent data to be silently interpreted as neutral evidence.
- Model routes are admitted according to their actual capabilities, so text generation, tool use, and structured agent interaction are not assumed to be interchangeable.
- The system’s contribution is primarily governance and software conformance: passing contract tests does not demonstrate better financial predictions, report quality, or returns.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.