Summary
Finding good investment ideas through multiple computer agents talking to each other can be slow because they exchange unclear messages. The paper’s authors reorganize this communication into clear, typed records that can be checked and replayed step-by-step, making the whole process more auditable and reliable. Their approach uses a special controller that predicts outcomes several steps ahead and only commits changes if certain checks pass, otherwise returning to a previous state. This structured method does not increase prediction accuracy but improves trust and traceability in how agents work together. When tested on financial data, their method outperformed several alternatives on some key measures, though overall returns compared to benchmarks were negative for all.
What this means in practice
- •For quantitative finance teams: Improve auditability and control in multi-agent trading strategy discovery by structuring agent communication and speculative evaluation.
- •For software architects for multi-agent systems: Design reliable and replayable agent protocols that enable deterministic interaction and rollback to prior states in complex multi-agent workflows.
Tested on one dataset.
Abstract
Agent-to-agent (A2A) alpha discovery is slowed by repeated feedback cycles between mining and evaluation agents, whose hand-offs, in contemporary LLM multi-agent systems, are free-form natural-language messages that carry no stable contract and cannot be replayed. We first restructure this communication as a structured agent-to-agent protocol of \emph{typed, causally addressable, unicast records}, so that the committed stream forms a causal trajectory. On that trajectory a single predictor with four typed heads forecasts the accumulated guidance the two miners would receive several cycles ahead; a transactional verify--leap controller then commits a multi-cycle speculative outcome only when it passes a four-level gate, and otherwise rolls back to the exact prior state. Structure is the enabling contribution, and its value is not accuracy. A controlled ablation shows an equal-information free-text channel reaches the same predictor hit rate. What typing provides is a state that can be schema-checked, replayed deterministically, and prevented by construction from leaking a forecast to an evaluator: auditability by construction, not an empirically stress-tested guarantee. On a CSI~1000 out-of-sample holdout, our single run is the only one among eight methods (seven baselines and ours) to hold a positive median annualized return and Sharpe at the factor level, though the median return \emph{in excess} of the benchmark stays negative for every method including ours; its development-selected top-20 portfolios reach a $0.71$ median holdout Sharpe, selected on a split inside the optimization horizon. We report these single-run results descriptively, gross of costs, and are explicit about their limits throughout; in particular we do not isolate the effect of the leap machinery from the inherited search substrate, which we leave to future work.