Tag
FilingLoop introduces an interactive benchmark and self-evolving LLM agent framework for corporate disclosure review and revision, grounded in 380 SEC IPO review tasks with requirement-level evaluation that improves fulfillment from 24.28% to 27.96% via procedural learning without weight updates.
该研究评估了 Llama-3 8B、Qwen-2.5 14B 和 Llama-3.3 70B 等 LLM 在从 SEC 10-K 文件中提取财务数据时的表现,发现将模型规模与文档结构复杂度对齐可在零样本设置下获得高准确率,从而缓解量化金融数据集中超过 70% 公司面临的缺失数据偏差问题。
A new paper from Harvard, MIT, and other labs introduces FINSKILLOPS, a method that enables financial AI agents to continuously learn from SEC filing errors while using regression tests to maintain correctness and safely update skills.
The paper proposes a training-free and alignment-free approach to corporate intelligence using deterministic sparse seed vectors for analyzing SEC filings, enabling efficient document comparison and semantic event detection without large language models.
An LLM-assisted agent published a false rule about SEC 8-K filing timestamps after confirming it only against the 120-company sample that generated it; the author retracted the claim and argues that agent outputs must be validated against data outside the hypothesis-generation loop.
This article reports on recent SEC filings that disclose AI-related deals, investments, and disclosures from companies like Microsoft, NVIDIA, and CIBC, indicating the expanding role of AI across various sectors.
Filing Studio's MCP tool allows tracing financial data from LLM outputs back to SEC filings, improving trustability in finance apps and saving tokens by using JSON instead of HTML.
Chamath discovered that changes in buzzword frequency in SEC filings can predict when a trend is about to wane. For instance, DEI peaked from 2020 to 2022 and then declined, while AI has been rising since late 2022 and is now mentioned in nearly every major company's filings.
This paper presents a two-stage LLM-based system that extracts grounded event tags from SEC 8-K filings using a three-tier taxonomy of 119 event types, with mechanisms for constraint and auditability. The system is evaluated on nearly 300k filings, showing high precision for high-quality tags and enabling event studies that distinguish economically distinct events.
A practical solution using AnySearch to enable local AI agents to efficiently query multiple specialized sources (CVEs, SEC filings) and return structured JSON/Markdown, avoiding rate limits and broken SDKs.
Researchers from Stanford, UC, and Nanjing University release SEFD, a dataset of 152B tokens from SEC filings converted to layout-faithful MultiMarkdown, preserving table structure for LLM training with minimal overlap with Common Crawl.
Open-sourced a set of buy-side research skills for the Codex platform, capable of pulling original files from the SEC for analysis, supporting multiple investment research frameworks.
Introduces Equibles, a self-hosted open-source MCP server that provides local LLMs with real U.S. financial data including SEC filings, insider trades, and economic indicators.