ESGReports: Long-Form Financial Report Passages for Semantic Retrieval

ESGReports is a dataset for the Massive Text Embedding Benchmark (MTEB) and Long-horizon Memory Embedding Benchmark (LMEB). It is designed for a semantic retrieval task, retrieving long-form report passages relevant to environmental, social, and governance questions. The dataset was created by mteb and was last updated on May 11, 2026.

Use Cases

• Benchmark semantic retrieval models based on the described ESG question-to-passage retrieval task.
• Train models to find relevant environmental policy information based on long-form report text.
• Evaluate model performance on social and governance-related queries based on financial report content.
• Research information retrieval techniques for domain-specific, long-document corpora.

Strengths

• Dataset is part of the established Massive Text Embedding Benchmark (MTEB) framework.
• Task is clearly defined for semantic retrieval of long-form passages related to ESG topics.
• Last update timestamp (2026-05-11 15:33:41) is provided.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license information are unknown.

Provenance

Source

huggingface

Collection Method

Part of the LMEB (Long-horizon Memory Embedding Benchmark), sourced from KaLM-Embedding/LMEB.

Freshness

Last updated 2026-05-11 15:33:41; freshness should be verified.