MoNIM: 30 Days of News Crawl Data for Continual Learning Experiments

30 daily shards of News Crawl data from the first half of 2020, used for reproducing continual-learning experiments in an ACL 2025 paper. The dataset includes a GPT-2 small checkpoint and was uploaded by viniferagy. It was last updated on 2026-06-18.

Use Cases

• Benchmarking continual learning algorithms based on the described daily data stream.
• Training or fine-tuning language models on sequential news data as described.
• Reproducing the experiments from the ACL 2025 paper on mixture-of-neighbors induction memory.

Strengths

• Provides 30 consecutive days of data for temporal sequence analysis.
• Includes a pre-trained GPT-2 small checkpoint for experiment reproduction.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

News Crawl data (NC-20H1).

Collection Method

Likely web-crawled news articles.

Time Range

First half of 2020, specifically the first 30 days.

Freshness

Last updated 2026-06-18 18:48:54; freshness should be verified.