verifiedpubliccc0-1.0Verified public
General Agent Instruction Pairs v1
Agent instruction corpus for tool use, task decomposition, refusal boundaries, instruction following, and concise operational responses.
Asset Snapshot
- Domain
- General Instruction Tuning
- Category
- Agent Instructions
- Version
- 1.0.0
- Records
- 553,324
- Size
- 2.1 GB
- Quality
- 93/100
- Fine-tune readiness
- 100/100
- Readiness label
- Fine-tune ready
- Validation
- passed
Intended Use
Fine-tune and evaluate assistant behavior for structured dataset workflows.
Not Intended Use
Not a complete safety policy or benchmark replacement.
Schema Preview
{
"instruction": "Summarize a dataset registry entry into a model card bullet list.",
"output": "Include source, license, intended use, risks, formats, and receipts."
}Example Records
[
{
"instruction": "Summarize a dataset registry entry into a model card bullet list.",
"output": "Include source, license, intended use, risks, formats, and receipts."
}
]Formats and Tasks
jsonlparquetinstruction-tuningqasummarizationgpt-ossllamamistralqwen
Fine-tuning Notes
Indexed from internal storage agent and instruction-following corpora with SHA256 and record-count receipts. Curated on sovereign bare-metal RTX 6000 fleet and RTX 3090 systems.
Record counts, byte sizes, and SHA256 hashes captured for agent and instruction-following JSONL sources.
Files, Hashes, and Receipts
| defendable-storage://general_agent_instruction_pairs_v1/agent_train_v1_jsonl | jsonl | full | 6.5 MB | 8b00646bce0faf1bebf6eab82c89b778356b6fb220cdfb5fece48a6c6e1c0af9 |
| defendable-storage://general_agent_instruction_pairs_v1/instruction_following_jsonl | jsonl | full | 2.1 GB | 253be000d865b75b407396438c3df94feff52298e16a3e722eef2451d8591763 |
External Storage Locations
Agent Train V1 Jsonl
defendable-storage://general_agent_instruction_pairs_v1/agent_train_v1_jsonl
6.5 MB
8b00646bce0faf1bebf6eab82c89b778356b6fb220cdfb5fece48a6c6e1c0af9
Internal source file with SHA256 receipt.
Instruction Following Jsonl
defendable-storage://general_agent_instruction_pairs_v1/instruction_following_jsonl
2.1 GB
253be000d865b75b407396438c3df94feff52298e16a3e722eef2451d8591763
Internal source file with SHA256 receipt.
sha256: receipt-general-agent-instruction-pairs-v1
General Agent Instruction Pairs v1 file inventory indexed with 2 file(s), 553,324 records, 2,295,095,293 bytes, and SHA256 receipts.
Dataset Card Preview
--- license: cc0-1.0 task_categories: instruction-tuning, qa, summarization pretty_name: General Agent Instruction Pairs v1 --- # General Agent Instruction Pairs v1 Agent instruction corpus for tool use, task decomposition, refusal boundaries, instruction following, and concise operational responses. ## Datasets - General Agent Instruction Pairs v1 (general_agent_instruction_pairs_v1) - 553,324 records ## Provenance This card was generated from DefendableDatasets registry metadata. Verify SHA256 hashes before training. ## Warnings - No warnings generated.