datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
course-imagesapex-agents
APEX–Agents
APEX–Agents is a benchmark from Mercor for evaluating whether AI agents can execute long-horizon, cross-application professional services tasks. Tasks were created by investment banking analysts, management consultants, and corporate lawyers, and require agents to navigate realistic work environments with files and tools (e.g., docs, spreadsheets, PDFs, email, chat, calendar).
Tasks: 480 total (160 per job category)
Worlds: 33 total (10 banking, 11 consulting, 12… See the full description on the dataset page: https://huggingface.co/datasets/mercor/apex-agents.agents-last-exam-data
Agents Last Exam — Task Input Data
Input files (the materials each task hands to the agent at run start) for the
Agents Last Exam (ALE) benchmark. Browsable per-task directory layout.
The Agents Last Exam dataset family
ALE is published as three companion HuggingFace datasets:
Dataset
Contents
Access
Task Card Metadata
One row per task: titles, prompts, taxonomy, input-file descriptors
Open
Task Input Data
The input/ files each task hands the agent at… See the full description on the dataset page: https://huggingface.co/datasets/agents-last-exam/agents-last-exam-data.DeepScaleR-Preview-Dataset
Data
Our training dataset consists of approximately 40,000 unique mathematics problem-answer pairs compiled from:
AIME (American Invitational Mathematics Examination) problems (1984-2023)
AMC (American Mathematics Competition) problems (prior to 2023)
Omni-MATH dataset
Still dataset
Format
Each row in the JSON dataset contains:
problem: The mathematical question text, formatted with LaTeX notation.
solution: Offical solution to the problem, including LaTeX formatting… See the full description on the dataset page: https://huggingface.co/datasets/agentica-org/DeepScaleR-Preview-Dataset.gaia2
Gaia2
Paper | Code | Project Page
Dataset Summary
Gaia2 is a benchmark dataset for evaluating AI agent capabilities in simulated environments. The dataset contains 800 scenarios that test agent performance in environments where time flows continuously and events occur dynamically.
The dataset evaluates seven core capabilities: Execution (multi-step planning and state changes), Search (information gathering and synthesis), Adaptability (dynamic response to environmental… See the full description on the dataset page: https://huggingface.co/datasets/meta-agents-research-environments/gaia2.resultsUltraData-SFT-Agent-2609
UltraData-SFT-Agent-2609
📦 UltraData Collection |
🌐 UltraData |
🤗 MiniCPM5 Series
English |
中文
📚 Introduction
UltraData-SFT-Agent-2609 is the L3 refined data for Agent instruction-tuning within UltraData's L0-L4 tiered data management framework. Built for the post-training of MiniCPM5-2B, it complements UltraData-SFT-2605 (core-domain SFT) with executable Agent trajectories. The release contains approximately 500,000 samples spanning tool use… See the full description on the dataset page: https://huggingface.co/datasets/openbmb/UltraData-SFT-Agent-2609.agent-reward-bench
AgentRewardBench
💾Code
📄Paper
🌐Website
🤗Dataset
💻Demo
🏆Leaderboard
AgentRewardBench: Evaluating Automatic Evaluations of Web Agent TrajectoriesXing Han Lù, Amirhossein Kazemnejad*, Nicholas Meade, Arkil Patel, Dongchan Shin, Alejandra Zambrano, Karolina Stańczak, Peter Shaw, Christopher J. Pal, Siva Reddy*Core Contributor
Loading dataset
You can use the huggingface_hub library to load the dataset. The dataset is available on Huggingface Hub at… See the full description on the dataset page: https://huggingface.co/datasets/McGill-NLP/agent-reward-bench.ale-images-qcow2
ALE QEMU runner image
agentslastexam/ale-qemu is the container-side runtime used by the ALE qemu
provider. It packages QEMU, KVM integration, NAT networking, noVNC, and process
supervision. The Ubuntu or Windows guest is supplied separately as
/storage/data.qcow2.
Docker is the container runtime. Dockur is the upstream QEMU-in-Docker project
whose startup and networking stack this image inherits. ALE adds a stable
runner contract around that upstream image.
The image is based on… See the full description on the dataset page: https://huggingface.co/datasets/agents-last-exam/ale-images-qcow2.unit4-students-scoresagent-task-recursive-task-synthesis
Apptainer pool for hamishivi/agent-task-recursive-task-synthesis
This repository hosts tmax-compatible SIF images and a unified download manifest. Training data and task archives are in hamishivi/agent-task-recursive-task-synthesis. The manifest includes earlier images hosted under hamishivi and new images hosted under TMaxxx; the downloader selects the correct repository and immutable commit for each image.
Apptainer images
The pool currently contains 29,501 / 29… See the full description on the dataset page: https://huggingface.co/datasets/TMaxxx/agent-task-recursive-task-synthesis.certificatesagent-data-collection
Agent Data Collection
A comprehensive collection of agent interaction datasets for training and evaluating AI agents across diverse domains and tasks.
This dataset aggregates high-quality agent trajectories from various environments including web browsing, code generation, household tasks, knowledge base querying, and software engineering.
The dataset is collected through methods described in Agent Data Protocol.
Dataset Splits
Each dataset configuration provides up… See the full description on the dataset page: https://huggingface.co/datasets/neulab/agent-data-collection.gaia2_filesystem
GAIA2 Filesystem
This is a dataset containing files for the GAIA2 benchmark. You should not use this dataset on its own, but instead use the Meta Agents Research Environments framework to execute scenarios from that GAIA2 dataset.
Dataset Link
https://huggingface.co/datasets/meta-agents-research-environments/gaia2
Contact Details
Publishing POC: Meta AI Research Team
Affiliation: Meta Platforms, Inc.
Website:… See the full description on the dataset page: https://huggingface.co/datasets/meta-agents-research-environments/gaia2_filesystem.requestsagents-last-exam-reference
Agents Last Exam — Reference (Ground-Truth) Data
⚠️ Gated dataset. This repo contains the ground-truth / reference outputs
used to score the Agents Last Exam (ALE) benchmark. Access requires login,
agreement to the terms on the access-request form, and manual approval.
Note (06/16/26): This repository was accidentally deleted and has been recreated. The
previous list of approved requesters could not be restored, so even if you
were granted access before, you will need to… See the full description on the dataset page: https://huggingface.co/datasets/agents-last-exam/agents-last-exam-reference.qwen-cpp-agent-0-protocolExperiment in agentic autonomy protocols.
~ everything in this repo was created by Qwen 3.8 27B (Q4) running autonomously inside Deepseek Harness, on a single RTX 3090 GPU, for 3 weeks.
The only human artifacts are:
agents/*
human/*
AGENTS.md
UI-Genie-Agent-16kThis repository contains the Trajectory dataset from the paper UI-Genie: A Self-Improving Approach for Iteratively Boosting MLLM-based
Mobile GUI Agents.
Github: https://github.com/Euphoria16/UI-Genie
SWE-agent-trajectories
Dataset Summary
This dataset contains 80,036 trajectories generated by a software engineering agent based on the SWE-agent framework, using various models as action generators. In these trajectories, the agent attempts to solve GitHub issues from the nebius/SWE-bench-extra and the dev split of princeton-nlp/SWE-bench.
Dataset Description
This dataset was created as part of a research project focused on developing a software engineering agent using open-weight models… See the full description on the dataset page: https://huggingface.co/datasets/nebius/SWE-agent-trajectories.Nemotron-SFT-Agentic-v2
Dataset Description
The Nemotron-SFT-Agentic-v2 dataset is a collection of synthetic single-turn and multi-turn tool-use trajectories designed to strengthen models’ capabilities as interactive, tool-using agents. It targets tasks where the model must decompose user goals, decide when to call tools, and reason over tool outputs to complete tasks reliably and safely.
This dataset is ready for commercial use.
The dataset consolidates three internally curated components (described… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Agentic-v2.appworld-qwen35-4b-agent-rl-epoch3
appworld-qwen35-4b-agent-rl-epoch3
Portable process-evaluation output. metadata.json is the lightweight source
for aggregate results; the JSONL files are directly loadable; and
artifacts.tar.gz losslessly preserves the original run directory.
Reasoning score: 0.45859375
Action score: 0.475
Valid samples: 320/320
appworld-qwen35-4b-agent-rl-epoch3-reeval1
appworld-qwen35-4b-agent-rl-epoch3-reeval1
Portable process-evaluation output. metadata.json is the lightweight source
for aggregate results; the JSONL files are directly loadable; and
artifacts.tar.gz losslessly preserves the original run directory.
Reasoning score: 0.4578125
Action score: 0.4921875
Valid samples: 320/320
AgentHarm
AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents
Maksym Andriushchenko1,†,*, Alexandra Souly2,*
Mateusz Dziemian1, Derek Duenas1, Maxwell Lin1, Justin Wang1, Dan Hendrycks1,§, Andy Zou1,¶,§, Zico Kolter1,¶, Matt Fredrikson1,¶,*
Eric Winsor2, Jerome Wynne2, Yarin Gal2,♯, Xander Davies2,♯,*
1Gray Swan AI, 2UK AI Safety Institute, *Core Contributor
†EPFL, §Center for AI Safety, ¶Carnegie Mellon University, ♯University of Oxford
Paper: https://arxiv.org/abs/2410.09024… See the full description on the dataset page: https://huggingface.co/datasets/ai-safety-institute/AgentHarm.Toucan-1.5M
🦤 Toucan-1.5M:
Toucan-1.5M is the largest fully synthetic tool-agent dataset to date, designed to advance tool use in agentic LLMs. It comprises over 1.5 million trajectories synthesized from 495 real-world Model Context Protocols (MCPs) spanning 2,000+ tools. By leveraging authentic MCP environments, Toucan-1.5M generates diverse, realistic, and challenging tasks requires using multiple tools, with trajectories involving real tool executions across multi-round, multi-turn… See the full description on the dataset page: https://huggingface.co/datasets/Agent-Ark/Toucan-1.5M.AgentNet
OpenCUA: Open Foundations for Computer-Use Agents
🌐 Website
🔎 Data Viewer
📝 Paper
💻 Code
AgentNet Dataset
AgentNet is the first large-scale desktop computer-use agent trajectory dataset, containing 22.6K human-annotated computer-use tasks across Windows, macOS, and Ubuntu systems.
Applications
This dataset enables training and evaluation of:
Vision-language-action (VLA) models for computer use
Multi-modal agents… See the full description on the dataset page: https://huggingface.co/datasets/xlangai/AgentNet.verdicts
Logbook verdicts
Per-logbook claim verdicts produced by the logbook-judge Space. See verdicts.json.
agent-task-facet-terminal-6k
Apptainer pool for hamishivi/agent-task-facet-terminal-6k
This repository hosts tmax-compatible SIF images and a unified download manifest. Training data and task archives are in hamishivi/agent-task-facet-terminal-6k. The manifest includes earlier images hosted under hamishivi and new images hosted under TMaxxx; the downloader selects the correct repository and immutable commit for each image.
Apptainer images
The pool currently contains 6,020 / 6,020 verified… See the full description on the dataset page: https://huggingface.co/datasets/TMaxxx/agent-task-facet-terminal-6k.agent-task-terminal-lego-15k
Apptainer pool for hamishivi/agent-task-terminal-lego-15k
This repository hosts tmax-compatible SIF images and a unified download manifest. Training data and task archives are in hamishivi/agent-task-terminal-lego-15k. The manifest includes earlier images hosted under hamishivi and new images hosted under TMaxxx; the downloader selects the correct repository and immutable commit for each image.
Apptainer images
The pool currently contains 15,048 / 15,048 verified… See the full description on the dataset page: https://huggingface.co/datasets/TMaxxx/agent-task-terminal-lego-15k.oeb-scored-runs
Open-Endedness Bench: scored runs
Every agent run scored in the paper Open-Endedness Bench: Measuring Epistemic
Process from Agent Records, with the output of each scoring stage and the full log
of judge requests and answers. The code is at
github.com/ARA-Labs/oeb.
Layout
out/posttrainbench/<task panel>/<unit>/ PostTrainBench runs (post-training gemma-3-4b on six
held-out tasks); a model's second run is <model>-r2… See the full description on the dataset page: https://huggingface.co/datasets/AgentNativeResearchLab/oeb-scored-runs.arc-agi3-codex-gpt5.5-su15
ARC-AGI-3 su15 — Agent Trajectories (codex-gpt5.5)
Gameplay trajectories from the harness×model pair codex-gpt5.5 playing the
ARC-AGI-3 game su15, part of the
ARA-as-world-model generalization experiment. The agent builds a structured world model
(an Agent-Native Research Artifact) live during play and consults it to crack levels it
cannot solve from cold exploration.
One dataset repo per harness×model×game: sibling repos
arc-agi3-<harness>-<model>-<game> hold the same game… See the full description on the dataset page: https://huggingface.co/datasets/AgentNativeResearchLab/arc-agi3-codex-gpt5.5-su15.