LCLM Eval: Unified Evaluation Mix for Latent Context Language Models

A unified evaluation mix for Latent Context Language Models (LCLM) containing four benchmarks. The dataset was created by the author 'latent-context' and was last updated on May 22, 2026. It uses a single schema with fields for prompt, category, and extra_info.

Use Cases

• Benchmarking model performance on grade-school math word problems based on the GSM8K component
• Evaluating language models on multi-document question answering based on the LongHealth5 patient-record component
• Testing model capabilities across diverse tasks and context lengths based on the Ruler benchmark
• Assessing long-context reasoning and retrieval based on the LongBench component mentioned in the description

Strengths

• Unified schema across four distinct benchmarks, simplifying evaluation setup
• Includes a specific benchmark with 39,000 rows covering 13 tasks and 6 context lengths
• Contains a component with 1,319 grade-school math word problems

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count for the full dataset is unknown, which may limit suitability assessment
• Description metadata is limited; actual data quality requires manual inspection after download

Provenance

Source

huggingface

Collection Method

Aggregated from multiple existing benchmarks on the Hugging Face platform.

Freshness

Last updated 2026-05-22 17:28:41; freshness should be verified