Datasets:

---

TreeAILab

/

Multi-turn_Long-context_Benchmark_for_LLMs

like 0

Follow

TreeAI-Lab 3

This PR significantly enhances the dataset card for the Hugging Face repository, transforming it into a comprehensive resource for users.

Key changes include:

• Linking to the Paper: The dataset is now explicitly linked to its associated paper, "LoopServe: An Adaptive Dual-phase LLM Inference Acceleration System for Multi-Turn Dialogues", ensuring proper attribution and context.
• Updating pretty_name: The pretty_name metadata has been updated from MKV_Cache to LoopServe Multi-Turn Dialogue Benchmark to accurately reflect the benchmark introduced in the paper, improving discoverability.
• Expanding configs Metadata: The configs metadata now accurately lists all available sub-datasets and splits, including conversations.jsonl and all multi_turn and single_turn categories (few-shot learning, needle-in-haystack, question-answering, summarization), making it easier for users to load specific parts of the benchmark.
• Enriching Content: The Markdown content has been updated with a clear description of the benchmark's purpose, drawn from the paper's abstract, along with a sample usage code snippet to guide users on how to load and interact with the dataset.

These updates aim to provide a more informative and user-friendly dataset card, aligning it closely with the research paper and the repository's contents.