LEDGER: New Benchmark Dataset for Evaluating LLMs on Financial Document Analysis
Researchers released LEDGER, a benchmark dataset of 4,999 digitized corporate annual reports with figures, tables, and narrative text designed to evaluate large language models' long-context capabilities in financial information extraction and retrieval. The dataset includes 31 consolidated financial KPIs (key performance indicators) labeled across each report and linked to market reactions at earnings dates. This resource addresses a gap in rigorous evaluation of LLMs on realistic financial documents beyond plain-text SEC filings.
LEDGER is a new evaluation benchmark comprising 4,999 full corporate annual reports—including figures, tables, and narrative sections—created to test large language models' ability to handle long-context financial documents. The dataset provides three evaluation tasks of increasing difficulty: page-level KPI retrieval with 118,048 natural language questions using TREC-style relevance judgments, conversational single-value lookup tasks, and full KPI extraction from numerically dense reports. Each report is annotated with 31 consolidated financial KPIs linked to market reactions at earnings announcements. The researchers also provide human OCR-quality annotations with inter-annotator agreement metrics and a complete extraction, validation, and scoring toolchain. A case study demonstrates the dataset's utility by linking CEO-letter rhetoric to post-publication market impact, showing practical applications beyond benchmark evaluation.
What's missing
The study does not discuss potential limitations in how well performance on this benchmark correlates with real-world financial analysis tasks, nor does it address whether the dataset's composition (e.g., time period, industry distribution, company size) may introduce biases in model evaluation.
What different sources said
- arXiv cs.CLCenter
LEDGER: A Long-Context Benchmark of Corporate Annual Reports for Grounded Financial Retrieval and Extraction
Related
Topology-Aware Thermodynamics Improves DNA Probe Specificity Design
Researchers developed a new framework for designing DNA probes that accounts for the spatial organization of matched sequences, not just overall thermodynamic stability. Traditional methods rely on scalar measures like melting temperature and free energy, which miss how mismatches are distributed along the probe. The approach could improve diagnostic accuracy in applications like HPV detection and gene expression profiling.
Study Identifies Optimal Thermal Dose for Combining Focused Ultrasound with Immunotherapy in Tumors
Researchers used multimodal PET imaging to identify an optimal thermal dose range for focused ultrasound ablation that destroys tumor tissue while preserving conditions for immunotherapy delivery. The study found that excessive heating collapses blood vessels needed for antibody access, while insufficient heating fails to adequately reduce tumor burden. The findings could guide clinical design of combination treatments pairing thermal ablation with immunotherapies.
Plant MSH1 Protein Functions as Mismatch-Directed Nuclease for Organelle Genome Maintenance
Researchers have identified the precise mechanism by which the AtMSH1 protein in Arabidopsis plants recognizes and cleaves DNA mismatches and lesions, preventing mutations in organellar genomes. The protein combines a DNA mismatch recognition module with a nuclease domain that makes staggered cuts at specific positions relative to DNA damage. This discovery explains how plants maintain unusually low mutation rates in their mitochondrial and chloroplast DNA compared to other eukaryotes.