1C Ebench: Russian-Language Code Search Benchmark for 1C/BSL

C1-Bench is a benchmark for evaluating embedding models on the task of retrieving relevant 1C (BSL) or query language code snippets based on a Russian-language question. The dataset contains real-world questions and code from 1C developer practice, which have been cleaned of PII, duplicates, and irrelevant content. It was created by PruhaNLP and last updated on June 3, 2026.

Use Cases

• Benchmarking embedding models for code retrieval based on Russian-language queries.
• Training or fine-tuning models for semantic search in 1C/BSL codebases.
• Evaluating the performance of retrieval-augmented generation (RAG) pipelines for programming assistance.

Strengths

• Data is derived from real developer practice, not synthetically generated.
• Data has been cleaned to remove personal identifiable information (PII), duplicates, and irrelevant content.
• Each query has exactly one relevant reference document, providing a clear evaluation target.

Limitations

• Row count, column definitions, and file formats are unknown, limiting suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data may reflect bias inherent to the specific practices of the contributing 1C developers.

Provenance

Source

PruhaNLP on Hugging Face.

Collection Method

Collected from real questions and code used by 1C developers, then cleaned.

Freshness

Last updated 2026-06-03 08:45:09