Public disclosure repositories such as SEC EDGAR provide web access to corporate filings, enabling informed investment. Assisting their review with language‑model agents requires uncovering unstated deficiencies across related documents, acquiring missing company facts, and verifying that revisions satisfy evidence‑backed requirements. Existing financial QA evaluations only offer limited evidence of these interactive capabilities.
We introduce FilingLoop, an interactive benchmark and a self‑evolving agent framework for corporate disclosure review and revision. Grounded in historical U.S. Securities and Exchange Commission correspondence, FilingLoop reconstructs 380 predominantly IPO review tasks from 96 companies under distinct information constraints. An evaluator separate from the acting agents scores final filings and responses against 6,099 eligible requirements and audits unsupported claims and harmful revisions.
Our framework couples iterative review and revision within each case with procedural learning across cases, converting requirement‑level feedback into role‑specific guidance without updating model weights. Updates affect only subsequent cases, allowing procedural adaptation to be assessed without re‑scoring completed tasks.
On the evaluated stream, evolution increases requirement fulfillment from $24.28\%$ to $27.96\%$ and full comment resolution from $16.90\%$ to $20.75\%$, with a $5.8\%$ increase in total cost over the fixed‑policy variant. FilingLoop connects the concerns an agent discovers with the company facts it obtains and the revisions it commits, providing requirement‑level feedback for procedures used in later review tasks.
Review: FilingLoop demonstrates a practical path for improving LLM‑based review agents through procedural learning in a real regulatory setting, offering a promising direction for automated financial compliance.