datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Q-Eval-100K
Q-Eval-100K Dataset （CVPR 2025 Oral）
📝 Introduction
The Q-Eval-100K dataset encompasses both text-to-image and text-to-video models, with 960K human annotations specifically focused on visual quality and alignment for 100K instances (60K images and 40K videos).
We utilize multiple popular text-to- image and text-to-video models to ensure diversity, which include FLUX, Lumina-T2X, PixArt, Stable Diffusion 3, Stable Diffusion XL, DALL·E 3, Wanx, Midjourney, Hunyuan-DiT… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/Q-Eval-100K.LARYBench
LARY — A Latent Action Representation Yielding Benchmark for Generalizable Vision-to-Action Alignment
LARY is a unified evaluation framework for latent action representations.
Given any model that produces latent action representations (LAMs or visual encoders), LARY provides three complementary evaluation pipelines:
Pipeline
Task
get_latent_action
Extract latent action representations from videos or image pairs
classification
Probe how… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/LARYBench.WBench
WBench Dataset
A Comprehensive Multi-turn Benchmark for Interactive Video World Model Evaluation
TL;DR — WBench evaluates 20 video world models across 5 dimensions and 22 metrics.
Overview
WBench is a comprehensive multi-turn benchmark for interactive video world model evaluation. It contains 289 multi-turn interaction cases with 1,058 interaction turns for evaluating models across 22 metrics and 5 dimensions:
Video Quality
Setting Adherence… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/WBench.AMO-Bench
📐 AMO-Bench: Large Language Models Still Struggle in High School Math Competitions
📄 Paper
🌐 Project Page
💻 Github Repo
Updates
2026.02.05: Leaderboard Update: Qwen3-Max-Thinking achieves a new SOTA with 65.1%, while GLM-4.7 sets a new open-source record at 62.4%!
2025.12.01: We have added Token Efficiency showing the number of output tokens used by models in the leaderboard. Gemini 3 Pro achieves the highest token efficiency among top-performance models!… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/AMO-Bench.UNO-Bench UNO-Bench: A Unified Benchmark for Exploring the Compositional Law Between Uni-modal and Omni-modal in Omni Models
🔔News
🔥[2025/12/04] We have released the evaluation scripts uno-eval, a unified evaluation framework for omni-modal benchmarks. More benchmarks will be supported in the future.
🔥[2025/12/04] We have released the scoring model UNO-Scorer-Qwen3-14B. Feel free to use it!
👀 UNO-Bench Overview
Multimodal Large Languages models have… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/UNO-Bench.ViC-Bench
ViC-Bench
About ViC-Bench
Visual-Interleaved Chain-of-Thought (VI-CoT) enables MLLMs to continually update their understanding and decisions based on step-wise intermediate visual states (IVS), much like a human would, which demonstrates impressive success in various tasks, thereby leading to emerged advancements in related benchmarks. Despite promising progress, current benchmarks provide models with relatively fixed IVS, rather than free-style IVS, whch might forcibly… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/ViC-Bench.VitaBench
🌱VitaBench: Benchmarking LLM Agents
with Versatile Interactive Tasks
📃 Paper • 🌐 Website • 🏆 Leaderboard • 🛠️ Code • 🤗 Dataset
🔔 News
[2026-01] Qwen3-Max-Thinking reported our Vita-Bench to evaluate and demonstrate its tool use capabilities (the averge score of 4 domains)！We invite the community to adopt Vita-Bench as the definitive touchstone for tool use performance assessment, and we appreciate diverse utilization & interpretation of our benchmark… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/VitaBench.LoHoSearch
LoHoSearch: Benchmarking Long-Horizon
Search Agents Beyond the Human Difficulty Ceiling
📃 Paper • 🏆 Benchmark • 📦 Training Data
Abstract
Search agent benchmarks exemplified by BrowseComp have rapidly saturated over the past year, with the strongest models surpassing 90% accuracy. Since these benchmarks are predominantly human-authored, annotators lack a global perspective on entity statistics and cannot systematically maximize search space size and structural… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/LoHoSearch.WBench-examples
WBench-examples
Reference video submissions for WBench,
a benchmark for interactive video world models. Each folder is a ready-to-evaluate
submission package — drop it into work_dirs/ and run the WBench pipeline.
Contents
Folder
Model
Type
Split
Videos
hyworld1.5/
HY-World 1.5 (Tencent)
camera
navi
158
kling3/
Kling 3.0 (Kling AI)
text
full
289
<model>/
├── meta.json # model metadata
├── turns.json # per-video… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/WBench-examples.R-HORIZON-AMC23
R-HORIZON
How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?
📃 Paper • 🌐 Project Page • 🤗 Dataset
R-HORIZON is a novel method designed to stimulate long-horizon reasoning behaviors in Large Reasoning Models (LRMs) through query composition. We transform isolated problems into complex multi-step reasoning scenarios, revealing that even the most advanced LRMs suffer significant performance degradation when facing interdependent problems that span… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/R-HORIZON-AMC23.mls_eng_longcatVitaBench-2.0General365_Public
🧩 General365: Benchmarking General Reasoning in LLMs Across Diverse and Challenging Tasks
📃 Paper • 🌐 Project Page • 🏆 Leaderboard •
💻 Github
📖 Introduction
We present General365, a highly challenging and diverse benchmark for evaluating the general reasoning capabilities in LLMs.
"General Reasoning" refers to reasoning tasks that depend exclusively on general knowledge.
We define general knowledge as knowledge within the K-12 scope (such as common sense… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/General365_Public.MineExplorer
MineExplorer: Evaluating Open-World Exploration of MLLM Agents in Minecraft
Tianjie Ju · Yueqing Sun · Zheng Wu · Wei Zhang · Yaqi Huo · Xi Su · Qi Gu · Xunliang Cai · Gongshen Liu · Zhuosheng Zhang
Abstract
Multimodal large language models (MLLMs) have shown strong capabilities in perception,
reasoning, and action generation. However, their ability to sustain exploration in dynamic
open worlds remains unclear. Existing embodied and… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/MineExplorer.R-HORIZON-training-data
R-HORIZON
How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?
📃 Paper • 🌐 Project Page • 🤗 Dataset
R-HORIZON is a novel method designed to stimulate long-horizon reasoning behaviors in Large Reasoning Models (LRMs) through query composition. We transform isolated problems into complex multi-step reasoning scenarios, revealing that even the most advanced LRMs suffer significant performance degradation when facing interdependent problems that span… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/R-HORIZON-training-data.R-HORIZON-Math500
R-HORIZON
How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?
📃 Paper • 🌐 Project Page • 🤗 Dataset
R-HORIZON is a novel method designed to stimulate long-horizon reasoning behaviors in Large Reasoning Models (LRMs) through query composition. We transform isolated problems into complex multi-step reasoning scenarios, revealing that even the most advanced LRMs suffer significant performance degradation when facing interdependent problems that span… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/R-HORIZON-Math500.meituna-longcat-replications
MeituNa LongCat Open-Ended Paper Replication Portfolio
Generated: 2026-07-31T12:15:27.895345+00:00
Multi-instance Meituan LongCat-2.0 workers produce multi-file open-source
replication scaffolds (method breakdown, code modules, experiment plans, scientific notes)
grounded in real arXiv paper text — not single-file stubs.
Run metadata
Parents: 8
Papers: 16
Model: LongCat-2.0
Passes: ['replication_plan', 'code_modules', 'experiments_and_critique'… See the full description on the dataset page: https://huggingface.co/datasets/Mieaz/meituna-longcat-replications.R-HORIZON-AIME24
R-HORIZON
How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?
📃 Paper • 🌐 Project Page • 🤗 Dataset
R-HORIZON is a novel method designed to stimulate long-horizon reasoning behaviors in Large Reasoning Models (LRMs) through query composition. We transform isolated problems into complex multi-step reasoning scenarios, revealing that even the most advanced LRMs suffer significant performance degradation when facing interdependent problems that span… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/R-HORIZON-AIME24.CEdit-BenchCEdit-Bench is a comprehensive evaluation suite, first proposed in the LongCat-Image technical report, and developed by integrating and extending existing image editing benchmarks. We further curate new data to enhance task diversity, yielding a robust dataset of 1,464 bilingual (Chinese–English) editing pairs across 15 fine-grained task categories, providing a more holistic and rigorous standard for evaluating image editing models.
An example entry is shown below:
{
"key":… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/CEdit-Bench.R-HORIZON-Websearch
R-HORIZON
How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?
📃 Paper • 🌐 Project Page • 🤗 Dataset
R-HORIZON is a novel method designed to stimulate long-horizon reasoning behaviors in Large Reasoning Models (LRMs) through query composition. We transform isolated problems into complex multi-step reasoning scenarios, revealing that even the most advanced LRMs suffer significant performance degradation when facing interdependent problems that span… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/R-HORIZON-Websearch.R-HORIZON-AIME25
R-HORIZON
How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?
📃 Paper • 🌐 Project Page • 🤗 Dataset
R-HORIZON is a novel method designed to stimulate long-horizon reasoning behaviors in Large Reasoning Models (LRMs) through query composition. We transform isolated problems into complex multi-step reasoning scenarios, revealing that even the most advanced LRMs suffer significant performance degradation when facing interdependent problems that span… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/R-HORIZON-AIME25.Meeseeks
👑 Meeseeks Benchmark
📖 Introduction
Meeseeks is an instruction-following benchmark designed to evaluate how well models can adhere to user instructions in a multi-turn scenario.A key feature of Meeseeks is its self-correction loop, where models receive structured feedback and must refine their responses accordingly.
This benchmark provides a realistic evaluation of a model’s adaptability, instruction adherence, and iterative improvement.
📊 Leaderboard… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/Meeseeks.details_SanjiWatsuki__longcat-10.7B
Dataset Card for Evaluation run of SanjiWatsuki/longcat-10.7B
Dataset automatically created during the evaluation run of model SanjiWatsuki/longcat-10.7B on the Open LLM Leaderboard.
The dataset is composed of 63 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_SanjiWatsuki__longcat-10.7B.Audio-Turing-Test-Corpus
📚 Audio Turing Test Corpus
A high‑quality, multidimensional Chinese transcript corpus designed to evaluate whether a machine‑generated speech sample can fool human listeners—the “Audio Turing Test.”
About Audio Turing Test (ATT)
ATT is an evaluation framework with a standardized human evaluation protocol and an accompanying dataset, aiming to resolve the lack of unified protocols in TTS evaluation and the difficulty in comparing multiple TTS systems. To further support… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/Audio-Turing-Test-Corpus.longcatmeituna-longcat-paper-analysis
Meituan LongCat Multi-Instance Paper Analysis
Generated: 2026-07-31T08:24:24.520521+00:00
Run metadata
Parents (workers): 8
Child fan-out per paper: 2
Model: LongCat-2.0
Paper source: orx|arxiv|ar5iv
Depth: heavy
Corpus summary
Papers analyzed: 12 (valid: 12, failed: 0)
Child focus jobs: 24
Total claims / methods: 54 / 35
Mean scientific score: 8.291666666666666
Papers
2412.19437: DeepSeek-V3 Technical Report… See the full description on the dataset page: https://huggingface.co/datasets/Mieaz/meituna-longcat-paper-analysis.CoreCodeBench-Single
Single Testcases for CoreCodeBench
File Explanation
CoreCodeBench_Single.jsonl: CoreCodeBench single test cases.
CoreCodeBench_Single_Verified.jsonl: Human verified version for CoreCodeBench single test cases.
CoreCodeBench_Single_en.jsonl: English version for CoreCodeBench single test cases.
CoreCodeBench_Function_Empty.jsonl CoreCodeBench function_empty test cases.
Key Explanation
Key
Meaning/Description
id
The unique identifier for the… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/CoreCodeBench-Single.CoreCodeBench-Source_CopyAudio-Turing-Test-Audios
📚 Audio Turing Test Audios
A high‑quality, multidimensional Chinese audio corpus generated from textual transcripts, designed to evaluate the human-likeness and naturalness of Text-to-Speech (TTS) systems—the “Audio Turing Test.”
About Audio Turing Test (ATT)
ATT is an evaluation framework featuring a standardized human evaluation protocol and an accompanying dataset, addressing the lack of unified evaluation standards in TTS research. To enhance rapid iteration and… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/Audio-Turing-Test-Audios.OIBench
OIBench Dataset
Dataset Overview
OIBench is a high-quality, private, and challenging olympiad-level informatics benchmark consisting of 250 carefully curated original problems.
The OIBench Dataset's HuggingFace repo contains algorithm problem statements, solutions, and associated metadata such as test cases, pseudo code, and difficulty levels. The dataset has been processed and stored in Parquet format for efficient access and analysis.
We provide complete information… See the full description on the dataset page: https://huggingface.co/datasets/meituan-longcat/OIBench.