MixEval/MixEval

🏠 Homepage | 👨‍💻 Github | 🏆 Leaderboard | 📜 arXiv | 📝 blog | 🤗 HF Paper | 𝕏 Twitter

---

Benchmark correlations (%) with Chatbot Arena Elo, against the total costs of evaluating a single GPT-3.5-Turbo-0125 model. MixEval and MixEval-Hard show the highest correlations with Arena Elo and Arena Elo (En) among leading benchmarks. We reference the crowdsourcing price for Amazon Mechanical Turk ($0.05 per vote) when estimating the cost of evaluating a single model on Chatbot Arena (approximately $2,936). Chatbot Arena is prohibitively expensive, while MixEval and MixEval-Hard are cheap and cost-effective alternatives. For more details, please refer to our paper.

Attention❗

This dataset repository is mainly for users wishing to load MixEval and MixEval-Hard manually with their custom evaluation code.

We recommend you to directly use the default Click-and-Go Evaluation Suite to evaluate models on MixEval and MixEval-Hard, it's easier and fairer!

Refer to here for a quick start of the evaluation suite. You can smoothly register new models (either proprietary or open-source) in a few steps.

If you choose to use your custom evaluation code, please refer to the formats and protocols specified here.

Updates

[2024-09-27] MixEval is accepted to Neurips 2024.

[2024-06-01] ⌛Current dynamic benchmark version: 2024-06-01. See here for the latest version.

Overview

We introduce MixEval, a ground-truth-based dynamic benchmark derived from off-the-shelf benchmark mixtures, which evaluates LLMs with a highly capable model ranking (i.e., 0.96 correlation with Chatbot Arena) while running locally and quickly (6% the time and cost of running MMLU), with its queries being stably and effortlessly updated every month to avoid contamination.

The MixEval consists of two benchmarks: MixEval and MixEval-Hard, both updated with our fast, stable pipeline periodically. Both of them contain two splits, i.e., free-form and multiple-choice. Below presents the their relationships:

 MixEval (dynamic)
    │
    ├── MixEval
    │   ├──free-form.jsonl
    │   └──multiple-choice.jsonl
    │
    └── MixEval-Hard
        ├──free-form.jsonl
        └──multiple-choice.jsonl

Below presents the key statistics of MixEval and MixEval-Hard. With dynamic benchmarking, the numbers may vary slightly while the number of queries will not change.

For more evaluation details, please refer to the Github repository.

See our paper and homepage for more technical details!

Load Dataset

To load MixEval benchmark, simply run:

from datasets import load_dataset 
# datasets>=2.18.0

dataset = load_dataset("MixEval/MixEval", 'MixEval')

To load MixEval-Hard benchmark, simply run:

from datasets import load_dataset 
# datasets>=2.18.0

dataset = load_dataset("MixEval/MixEval", 'MixEval_Hard')

What is MixEval?

MixEval is an approach that bridges the gap between real-world user queries and efficient, reproducible evaluation by leveraging user queries mined from the web and matching them with similar queries from existing benchmarks. MixEval is also the proposed benchmark built with this approach.

MixEval-Hard is the hard version of MixEval, designed to enhance the benchmark's ability to distinguish strong models. It is sampled from MixEval based on model evaluation results, with a higher probability of selecting harder queries. To address distribution deviation, we introduce a rejective sampling process to ensure that the distribution of MixEval-Hard aligns with that of wild queries.

Dynamic evaluation is introduced to mitigate the contamination issue. We periodically update the data points in MixEval and MixEval-Hard using our fast, stable pipeline, which performs benchmark mixture with a different batch of wild queries from the same distribution, showing low variance (0.36 Std. on a 0-100 scale) and significant version difference (85% unique query ratio).

Why to Use MixEval Benchmarks?

MixEval offers five significant advantages for practitioners:

• Accurate model ranking, demonstrated by a 0.96 correlation with Chatbot Arena1.
• Fast, cheap and reproducible execution, requiring only 6% the time and cost of MMLU and with no dependence on human input.
• Dynamic benchmarking enabled by low-effort and stable updating mechanism.
• A comprehensive and less biased query distribution, as it bases queries on a large-scale web corpus.
• A fair grading process, ensured by the ground-truth-based grading mechanism.

How Effective is MixEval as a Benchmark Mixture Approach?

MixEval is effective as

• MixEval and MixEval-Hard achieve the highest correlation with Arena Elo and Arena Elo (En) among all benchmarks.
• MixEval improves the correlation with Arena Elo and Arena Elo (En) across all its main benchmark splits.
• MixEval outperforms both benchmark-level and uniform mixtures.
• MixEval effectively maps real-world user queries to ground-truth-based benchmarks.

📑 Citation

If you found this repository useful, please consider 📑citing:

@article{ni2024mixeval,
  title={MixEval: Deriving Wisdom of the Crowd from LLM Benchmark Mixtures},
  author={Ni, Jinjie and Xue, Fuzhao and Yue, Xiang and Deng, Yuntian and Shah, Mahir and Jain, Kabir and Neubig, Graham and You, Yang},
  journal={arXiv preprint arXiv:2406.06565},
  year={2024}
}