FinanceBench Voyage Finance 2 Embeddings: Pre-computed Vectors for SEC Filings

Pre-computed embeddings for the FinanceBench corpus of U.S. SEC filings. The dataset provides vector representations for documents like 10-K, 10-Q, 8-K, and earnings releases, intended to be used with the Rishabhmannu/financebench-rag-agent. It was created by cmpunkmannu and last updated on HuggingFace in June 2026.

Use Cases

• Build financial question-answering agents based on pre-computed document embeddings.
• Benchmark retrieval performance for RAG systems using SEC filing content.
• Accelerate prototype development for financial NLP tools by skipping the embedding generation step.

Strengths

• Saves an estimated $5-15 in Voyage API costs per user versus re-embedding from raw PDFs.
• Saves an estimated 30 minutes of data ingest and processing time per user.
• Embeddings are derived from a canonical corpus of SEC filings parsed with pypdf.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and data size are unknown, which may limit suitability assessment.
• Last updated 2026-06-01 17:48:51; freshness should be verified.

Provenance

Source

FinanceBench corpus (SEC filings: 10-K, 10-Q, 8-K, earnings releases).

Collection Method

Parsed with pypdf; embeddings generated via Voyage AI's voyage-finance-2 model.

Freshness

Last updated 2026-06-01 17:48:51.

Geography

United States (U.S. SEC filings).