FIDES-Live audits trading-agent traces for decision and ledger agreement
Original title: FIDES-Live: Evaluating Behavioral Concordance in Sequential Trading Agents
Across 7,299 crypto account-rounds, decision validity was 97.35%, but nonempty order commitments held only 62.63%.
FIDES-Live extends FIDES from strategy concordance to auditing sequential trading-agent traces, covering decision validity, numerical commitments, execution and ledger reconciliation, and receipt-grounded reporting. A retrospective audit used ten inference routes, 7,299 cryptocurrency account-rounds, and a 310-round equity pilot with real market data and local paper execution. Cash and position transitions reconciled with recorded fills, but decision and reporting performance diverged by route. The authors say the results do not show long-run trading skill or causal benefits from audit feedback.
Unconfirmed: The paper does not claim the audit establishes trading skill, explanation faithfulness, or benefits from feedback.
- Structured reporting accuracy was 90.81% among scorable cryptocurrency responses.
- Overall commitment adherence was 94.36%, versus 97.56% for valid empty proposals.
- Execution controls and unavailable quotes accounted for separate rejection populations.
- Short-window marked returns are descriptive context, not a test of alpha.
From the original article
1 min read at arxiv.orgSequential trading agents produce decisions, orders, execution receipts, account states, and reports that need not agree, even when the final portfolio value is correctly computed. We present FIDES-Live, an operational extension of FIDES from generated strategy concordance to the audit of sequential agent traces.
By Arther Tian · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.