MWS Vision Bench is the first Russian-language business OCR benchmark designed for multimodal large language models. The validation split is publicly available for open evaluation and comparison, with a paper expected soon. The dataset was uploaded by MTSAIR and last updated on March 11, -2026.
Use Cases
- Benchmarking OCR performance of multimodal LLMs based on Russian business documents.
- Comparing model performance across languages using the provided English and Chinese question configs.
- Training models for business document understanding based on the multimodal benchmark structure.
Strengths
- Designed as the first Russian-language business OCR benchmark for MLLMs.
- Provides three language configurations (Russian, English, Chinese) for the same validation split.
- Publicly available validation split intended for open evaluation and comparison.
Limitations
- Row count, column names, and file formats are unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
Provenance
- Source
- MTSAIR
- Collection Method
- Likely contains annotated business document images and associated questions for benchmark tasks.
- Time Range
- null
- Freshness
- Last updated 2026-03-11 10:44:06; freshness should be verified.
- Geography
- null