FilingLoop rebuilds SEC disclosure review as an interactive loop, where the agent has to find what the filing does not say
AspireAI community · arXiv
Public disclosure repositories such as SEC EDGAR support informed investment through web access to corporate filings, but 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. FilingLoop is 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, reconstructing 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. The 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, and updates affect only subsequent cases so adaptation can be assessed without retrying scored tasks. On the evaluated stream, evolution increases requirement fulfilment from 24.28% to 27.96% and full comment resolution from 16.90% to 20.75%, at a 5.8% increase in total cost over the fixed-policy variant. The limitation is the domain: the benchmark is built on regulatory correspondence rather than live filings, and the absolute fulfilment rates are low.
Read the story and explore more from the AspireAI community.
Two sparse-autoencoder analyses have resisted scaling: matching semantically similar features across layers, and compressing large feature circuits into interpretable supernodes. This paper argues both are the same underlying problem, estimating semantic distances between SAE features that sit on different activation manifolds. Its representation treats each feature not as a single decoder vector but as an activation-weighted distribution over the hidden states expressing it, projects those distributions into a shared reference space, and compares them with Wasserstein distance. The representation is proved invariant to activation rescaling and stable under perturbation, and recovers true matches under finite-sample margin conditions. Empirically it outperformed decoder-vector and LLM-based baselines and captured subtle functional distinctions between related features, and it compressed large feature circuits into interpretable supernodes automatically. The limitation is that this is interpretability tooling: the paper demonstrates better feature matching and circuit compression, not that downstream interpretability claims become more accurate.