12,000 instruction-following examples were released on October 8, 2023 by Yukang for training long-context language models. The dataset is associated with LongLoRA and LongAlpaca models, including LongAlpaca-7B, LongAlpaca-13B, and LongAlpaca-70B.
Use Cases
- Fine-tuning language models for long-context instruction following based on the dataset's described purpose
- Benchmarking model performance on long-context tasks based on the dataset's association with LongAlpaca models
- Researching methods for extending context windows in LLMs based on the LongLoRA technique mentioned
Strengths
- Dataset size is explicitly stated as 12,000 examples
- Associated with three specific model versions (7B, 13B, 70B) released on a known date
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
Provenance
- Source
- Yukang
- Freshness
- Last updated 2023-10-11 04:03:27