Malayalam Instruct Dataset-L: A Large-Scale Malayalam Instruction-Tuning Corpus

Malayalam Instruct Dataset-L is a large-scale instruction-tuning dataset for the Malayalam language. It was programmatically compiled from over 20 multilingual text corpora, translation engines, and RSS feeds, heavily featuring the CulturaX database. The dataset was created by author siyah1 and was last updated on June 17, 2026.

Use Cases

• Instruction-tuning large language models based on the described multilingual instruction data.
• Training models for Malayalam text generation based on the compiled instruction-response pairs.
• Benchmarking model performance on low-resource language tasks based on the dataset's scale.
• Developing multilingual NLP applications based on the dataset's integration of multiple sources.

Strengths

• Compiled from over 20 distinct multilingual text sources.
• Heavily features the large-scale CulturaX database.
• Described as one of the largest Malayalam instruction-tuning datasets available.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last updated 2026-06-17 06:59:47; freshness should be verified.

Provenance

Source

Compiled from sources including the uonlp/CulturaX database, translation engines, and RSS feeds.

Collection Method

Programmatically scraped, deduplicated, and unified.

Freshness

Last updated 2026-06-17 06:59:47.