LEDGER: Long-Context KPI Question Answering and Page Retrieval

LEDGER — Long-Context KPI Question Answering & Page Retrieval is part of the LEDGER benchmark for evaluating long-context document understanding. It supports page-level retrieval tasks where a natural-language question about a financial KPI must be answered from an annual report, with TREC-style graded relevance judgments provided. The dataset was created by artefactory and was last updated on June 7, 2026.

Use Cases

• Evaluate page retrieval models based on TREC-style graded relevance judgments.
• Benchmark question-answering systems for financial key performance indicators (KPIs).
• Train models for long-context document understanding using annual reports.

Strengths

• Part of a structured benchmark (LEDGER) for long-context evaluation.
• Includes graded relevance judgments for retrieval tasks.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last updated 2026-06-07 02:47:30; freshness should be verified.

Provenance

Source

artefactory

Freshness

Last updated 2026-06-07 02:47:30.