Datasets:

kshitijthakkar
/

OpenEnv-Arena-MultiDomain-RL-Mix-v1

domain stringclasses 15 values	example_id stringlengths 14 27	grader_status stringclasses 2 values	is_original_authored bool 2 classes	prompt stringlengths 23 69.1k	reward_type stringclasses 12 values	source_config stringclasses 7 values	source_dataset stringclasses 24 values	source_license stringclasses 6 values	source_record_json stringclasses 161 values	source_revision stringclasses 24 values	source_row_index stringclasses 210 values	source_split stringclasses 2 values	split stringclasses 1 value	task_family stringclasses 22 values	task_fingerprint stringlengths 64 64
conversation	up-0793f2e87808649a64ecf4b3	source_native; verify before using as a reward	false	### Identity - Name: Aiden - Occupation: Elementary Skills Tutor - Dependency: AI assistant operating within the EduSpark learning platform, a digital learning environment designed to support elementary-level education in a safe, guided and age-appropriate manner ### Purpose/Role/Responsibility Your ro...	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	13	train	train	multi-turn-instruction-following	89e356e02f624f27cadaf93c36037c28a31e7102c8339f301b939abeae656ea4
conversation	up-0ac199f95b9145ec01aa74f9	source_native; verify before using as a reward	false	### Identity You are 'Vulcan', an Industrial Process Optimization AI. You are a world-class expert consultant with decades of synthesized experience in industrial machinery operations, process optimization, and workplace safety compliance. Your knowledge base is built upon a vast library of technical documentation, ...	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	10	train	train	multi-turn-instruction-following	519677b96f60106ca7cd42e5182c3935e708f6606637918ce7bd28cae79cf549
conversation	up-0d9a71ab6b0fc6166bc827da	source_native; verify before using as a reward	false	### Identity You are Carol Rainwater. By profession, you are a methodical and disciplined bailiff, a role that has ingrained in you a profound respect for order, evidence, and procedure. In your solitary private time, you apply this same rigor to your one true passion: serving as the definitive, unofficial archivist...	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	0	train	train	multi-turn-instruction-following	f58b3d86090e8661352995f52768bd0319dcad180cfa5ee3aed8b5c05bc3dee8
conversation	up-0decb13554e3eda795f1d342	source_native; verify before using as a reward	false	### Identity Name: Alex Rivera Occupation: Retail Customer Service Representative Dependency: AI assistant employed by the national retail chain ShopEase ### Purpose/Role/Responsibility One of your key responsibilities is serving both the customer and the staff within the retail context. It is expe...	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	3	train	train	multi-turn-instruction-following	4a005f327fcec9631dc0ba55f8769e10473f232469b4687db46bea72b7b23059
conversation	up-13c2cddae84a43116d501484	source_native; verify before using as a reward	false	### Identity You are Aysha, an aspiring educator and animal caregiver deeply passionate about nurturing creativity, emotional well-being, and kindness in children and animals. ### Purpose/Role/Responsibility Your primary role is to provide empathetic guidance and creative solutions related to education, arti...	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	22	train	train	multi-turn-instruction-following	46b390bb7c063ce87e180c38bbe1843e20b332631fdba2ef8e5139adbab471a6
conversation	up-14d22fc82712cde5d3c25cc2	source_native; verify before using as a reward	false	### Identity You are Claude. You are a Travel and Nature Escapes Specialist. You are an AI assistant that works for Anthropic, specializing in budget-friendly, immersive travel experiences inspired by Debra Boisvert's preferences. ### Purpose/Role/Responsibility Your role is to help users plan low-key,...	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	12	train	train	multi-turn-instruction-following	35e1d987d39656c7bdc5b8a74b59355d7f86333970c4970eefce928a96a0d7a0
conversation	up-176336677bc51659347cb552	source_native; verify before using as a reward	false	### Identity and the Background You are Shelton, an AI assistant whose reasoning style and communication approach are shaped by a Midwestern upbringing in Chillicothe, Ohio. You come from a family of German and Irish ancestry, raised in an environment where consistency, practical effort, and personal accountabil...	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	1	train	train	multi-turn-instruction-following	f559d59dcd6e31bd539aa05ab381df7e8afe1ae8a8fba19b18663f689e1b9580
conversation	up-1b529ebaf4b032733b3f9de1	source_native; verify before using as a reward	false	### Identity - title You are Yolanda Williams, an Eligibility Specialist with extensive experience in social services. You hold a bachelor's in education, which suggests you can conduct a patient and instructional approach with clients. You are fluent in English and Spanish, and trained in meticulous, low-stress benef...	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	30	train	train	multi-turn-instruction-following	b49fb90c8a4c452923471c48b4039dc3c154849cb8bf2949ce4aa4ef4804a614
conversation	up-28684207a04d1aa1b0ec29b5	source_native; verify before using as a reward	false	"### Identity \n Name: Maria Jimenez \n Occupation: Watercolor Painter \n Dependency: You are an (...TRUNCATED)	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	38	train	train	multi-turn-instruction-following	cca3e782a54efd2b9afeda6618e8e11777033f952e5ca8c0ad8efed81293e1f1
conversation	up-2bbdeaad296df024e8346bed	source_native; verify before using as a reward	false	"### Identity \n You will assume the identity of Dr. Anessa Hutchens, a skilled, incredibly talente(...TRUNCATED)	source_defined_verifier	default	nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	{}	b5d12c032618a34e8e72ff592aba6833508e8063	18	train	train	multi-turn-instruction-following	d0f9e6fd95a7d1a7850c1e7f0102ef729087e24e92c6269b66814df73ba6c4cf

End of preview. Expand in Data Studio

OpenEnv Arena Multi-Domain RL Mix v1

This release combines a bounded, source-attributed sample from public NVIDIA RL datasets and XiaomiMiMo's MiMo-V2.6-RL-oss with the locally authored Arena curriculum. Upstream records are stored as JSON in source_record_json; task prompts, verifier family, source revision, row locator, and source license are preserved on every row.

Dataset splits

train: Arena-authored training tasks plus upstream sample records assigned to train by a stable prompt hash.
validation: the Arena curriculum's held-out cases plus a stable 20% prompt-hash holdout from upstream samples. This is a mixture validation split, not OpenEnv Arena's private leaderboard evaluation.
preference: safety preference rows kept separate from verifiable-reward tasks.
source_catalog: all datasets visible in the supplied NVIDIA screenshots plus XiaomiMiMo; composite blends and FPM metadata remain linked references to avoid hidden duplication and licensing flattening.

Upstream source coverage

Source dataset	Source license	Intake status	Sample rows
XiaomiMiMo/MiMo-V2.6-RL-oss	apache-2.0	sampled	250
nvidia/Nemotron-3-Nano-RL-Training-Blend	odc-by+component-licenses	catalog_only	0
nvidia/Nemotron-RL-Agentic-Conversational-Tool-Use-Pivot-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Agentic-Function-Calling-Pivot-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Agentic-SWE-Pivot-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Identity-Following-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Instruction-Following-Adversarial-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Instruction-Following-Calendar-v2	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Instruction-Following-Citation-Formatting-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Instruction-Following-Free-Form-Formatting-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Instruction-Following-Structured-Outputs-v2	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-ReasoningGym-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Safety-v1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-Science-v1	cc-by-sa-4.0	sampled	50
nvidia/Nemotron-RL-Super-Training-Blends	cc-by-4.0+component-licenses	catalog_only	0
nvidia/Nemotron-RL-Ultra-Training-Blends	cc-by-4.0+component-licenses	catalog_only	0
nvidia/Nemotron-RL-agent-calendar_scheduling	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-agent-workplace_assistant	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-coding-competitive_coding	cc-by-sa-4.0	sampled	50
nvidia/Nemotron-RL-instruction_following	odc-by	sampled	50
nvidia/Nemotron-RL-instruction_following-structured_outputs	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-knowledge-mcqa	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-knowledge-web_search-mcqa	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-litmus-bench-v0.1	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-math-OpenMathReasoning	cc-by-4.0	sampled	50
nvidia/Nemotron-RL-math-advanced_calculations	unverified	catalog_only	0
nvidia/Nemotron-RL-math-stack_overflow	cc-by-sa-4.0	sampled	50
nvidia/aisimulate-fpm-dataset	apache-2.0	catalog_only	0

Reward handling

Rows use reward_type and grader_status to distinguish executable tests, rule checks, procedural verifiers, exact-answer candidates, schema/rubric checks, and preference pairs. These labels describe the source's reward path; they do not claim all tasks have the same deterministic grader. The upstream verifier artifacts and environment remain authoritative. Before using a row for RL reward, validate the source answer/verifier against correct, partial, and wrong cases and promote only tasks with a reproducible grader.

MiMo contributes five environment families—code, cyber, general, web development, and music—and their verifier metadata. Its Docker images are referenced, not copied. NVIDIA's large blended datasets are kept in the source catalog because their rows overlap component repos and carry layered licenses.

Licensing and attribution

The repository has no single blanket license. Each upstream-derived row retains source_license, source_dataset, and source_revision. Follow the license and attribution terms of that specific source; CC BY-SA and ODC-BY samples remain identifiable as such. data/source_catalog.jsonl includes sources that are references only. The authored Arena curriculum is released under Apache-2.0. Users must review underlying source terms and verify any source whose license is marked unverified before redistribution.

Reproducibility

manifest.json records row counts; source_catalog.jsonl records sampled configs, splits, offsets, and immutable source revisions. scripts/build_arena_multidomain_rl_mix.py rebuilds the bounded sample from public Dataset Viewer APIs. Per-config sampling is capped at 50 source rows to keep the release usable while retaining broad source coverage.

OpenEnv Arena

Challenge: https://huggingface.co/proxy/openenvarena-arena.hf.space/

Downloads last month: 74

Collection including kshitijthakkar/OpenEnv-Arena-MultiDomain-RL-Mix-v1

OpenEnv Arena Challenge Artifacts

Collection

Arena datasets v4/v21; images: hub.docker.com/r/kshitij190/arena-domain-lab; challenge: openenvarena-arena.hf.space • 7 items • Updated 7 days ago