Researchers introduce FilingLoop for SEC disclosure review
A new research paper introduces FilingLoop, an interactive benchmark and self-evolving agent framework designed for corporate disclosure review and revision. The system is grounded in historical U.S. Securities and Exchange Commission correspondence, reconstructing 380 predominantly IPO review tasks from 96 companies. These tasks operate under distinct information constraints, requiring agents…
Key points
- FilingLoop reconstructs 380 IPO review tasks from 96 companies using historical SEC correspondence.
- An independent evaluator scores filings against 6,099 eligible requirements to audit claims.
- Procedural learning increased requirement fulfillment from 24.28% to 27.96% without weight updates.
The framework uses an evaluator separate from the acting agents to score final filings against 6,099 eligible requirements. It audits unsupported claims and harmful revisions while coupling iterative review with procedural learning across cases. This process converts requirement-level evaluation feedback into role-specific guidance without updating model weights, ensuring that updates only affect subsequent cases.
On the evaluated stream, this evolution strategy increased requirement fulfillment from 24.28% to 27.96% and full comment resolution from 16.90% to 20.75%. This improvement came with a 5.8% increase in total cost compared to a fixed-policy variant. FilingLoop aims to provide requirement-level feedback for procedures used in later review tasks, addressing limitations in existing financial question-answering evaluations.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- AI startups aim to automate formal software verification with $365M funding · 2 src
- Researchers find instruction exposure unlocks rhetorical patterns in LLMs · 1 src
- New guide introduces LLMs for scientists and professionals · 1 src
- Researchers introduce Seq@K to test language models with sequential feedback · 6 src
- Researchers propose rank-aware method for private listwise LLM alignment · 1 src
Comments
via GitHub Discussions