cs.AI提交于 2026-10-31已译

FilingLoop：评估与演化用于企业披露审查和修订的 LLM 智能体

FilingLoop: Evaluating and Evolving LLM Agents for Corporate Disclosure Review and Revision

Wentao Zhang · Yilei Zhao · Guangyan Gan · Shuo Sun · Lin William Cong

中文摘要

SEC EDGAR 等公开披露库通过网页访问企业文件支持知情投资。用语言模型智能体辅助审查，需要发现相关文档中未明示的缺陷、获取缺失的公司事实，并验证所要求的修订满足有证据支持的要求。现有金融问答评测对这些交互能力提供的证据有限。FilingLoop 是一个交互式基准和自演化智能体框架，用于企业披露审查与修订。基于美国证券交易委员会历史往来函件，FilingLoop 在 96 家公司上重建了 380 个以 IPO 审查为主的任务，并施加不同信息约束。独立于执行智能体的评估器依据 6,099 条合格要求对最终文件和回复打分，并审计无依据主张与有害修订。框架将每个案例内的迭代审查与修订同跨案例的程序性学习耦合，把要求级评估反馈转化为角色特定指导，且不更新模型权重。更新只影响后续案例，从而可在不重试已评分任务的情况下评估程序性适应。在评估流上，演化将要求满足率从 24.28% 提升至 27.96%，完整评论解决率从 16.90% 提升至 20.75%，总成本比固定策略变体增加 5.8%。FilingLoop 将智能体发现的关注点、获取的公司事实与提交的修订连接起来，为后续审查任务所用程序提供要求级反馈。

关键要点

01现有金融问答评测难以反映智能体发现未明示缺陷、获取缺失事实并验证修订的交互能力。
02基于 SEC 历史函件重建 96 家公司 380 个以 IPO 审查为主的任务，含 6,099 条合格要求。
03独立评估器对最终文件打分并审计无依据主张与有害修订，反馈转化为角色指导且不更新权重。
04演化使要求满足率从 24.28% 升至 27.96%，完整评论解决率从 16.90% 升至 20.75%。
05程序性适应仅影响后续案例，总成本比固定策略变体增加 5.8%。

解读

尚无解读。

原始英文摘要

arXiv:2610.37905v1 Announce Type: new Abstract: Public disclosure repositories such as SEC EDGAR support informed investment through web access to corporate filings. Supporting their review with language model agents requires discovering unstated deficiencies across related documents, acquiring missing company facts, and verifying that requested revisions satisfy evidence-backed requirements. Existing financial question-answering evaluations provide limited evidence of these interactive capabilities. We introduce \textbf{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 evaluation feedback into role-specific guidance without updating model weights. Updates affect only subsequent cases, allowing procedural adaptation to be assessed without retrying scored 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.