Terminal-Corpus is a large-scale Supervised Fine-Tuning (SFT) dataset designed to scale the terminal interaction capabilities of Large Language Models (LLMs).
Dataset Card
By NVIDIA, published under cc-by-4.0, revision a1667c4ffdad.
Terminal-Corpus: Large-Scale SFT Dataset for Terminal Agents
Terminal-Corpus is a large-scale Supervised Fine-Tuning (SFT) dataset designed to scale the terminal interaction capabilities of Large Language Models (LLMs). Developed by NVIDIA, this dataset was built using the Terminal-Task-Gen pipeline, which combines dataset adaptation with synthetic task generation across diverse domains.
Key Results & Performance
The high-quality trajectories in Terminal-Corpus enable models of various sizes to achieve performance that rivals or exceeds much larger frontier models on the Terminal-Bench 2.0 benchmark.
1. Overall Performance Comparison
Training on Terminal-Corpus yields substantial gains across the Qwen3 model family:
| Model Size | Base Model (Qwen3) Accuracy | Nemotron-Terminal Accuracy | Improvement |
|---|---|---|---|
| 8B | 2.5% ± 0.5 | 13.0% ± 2.2 | ~5.2x |
| 14B | 4.0% ± 1.3 | 20.2% ± 2.7 | ~5.0x |
| 32B | 3.4% ± 1.6 | 27.4% ± 2.4 | ~8.0x |
The Nemotron-Terminal-32B (27.4%) outperforms the 480B-parameter Qwen3-Coder (23.9%) and Gemini 2.5 Flash (16.9%). Nemotron-Terminal-14B (20.2%) achieves higher accuracy than the 120B GPT-OSS (high) (18.7%).
2. Domain-Specific Breakthroughs
Structure
dataset_adapters 226,313 rows
| Split | Rows | Size |
|---|---|---|
| train | 226,313 | 13.3 GB |
skill_based_easy 44,809 rows
| Split | Rows | Size |
|---|---|---|
| train | 44,809 | 2.2 GB |
skill_based_medium 89,343 rows
| Split | Rows | Size |
|---|---|---|
| train | 89,343 | 6.2 GB |
skill_based_mixed 5,689 rows
| Split | Rows | Size |
|---|---|---|
| train | 5,689 | 271.0 MB |
Details
- Repository
- nvidia/Nemotron-Terminal-Corpus
- Publisher
- NVIDIA
- Task category
- Question answering
- Tags
- code
- Size category
- 100K<n<1M
- Languages
- en
- Revision
- a1667c4ffdadea02a89bffe4f1bb7ca2ff19f8d9
- Last updated
- 2026-02-27
Files
31 files, 8.2 GB in total.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| dataset_adapters/code.parquet | Data | 854.7 MB | 739d3e13f14d |
| dataset_adapters/math.parquet | Data | 3.0 GB | 0eacf5c4a8ac |
| dataset_adapters/swe.parquet | Data | 1.2 GB | 90cd061d73ac |
| synthetic_tasks/skill_based/easy/data_processing/data_filtered.parquet | Data | 76.8 MB | ef34d9a0ba58 |
| synthetic_tasks/skill_based/easy/data_querying/data_filtered.parquet | Data | 82.1 MB | b3beff8c8c87 |
| synthetic_tasks/skill_based/easy/data_science/data_filtered.parquet | Data | 83.7 MB | b8a1de95d97d |
| synthetic_tasks/skill_based/easy/debugging/data_filtered.parquet | Data | 86.7 MB | c6b667d077d8 |
| synthetic_tasks/skill_based/easy/dependency_management/data_filtered.parquet | Data | 95.1 MB | 819bd3e42d46 |
| synthetic_tasks/skill_based/easy/file_operations/data_filtered.parquet | Data | 75.8 MB | 50cb715afdd8 |
| synthetic_tasks/skill_based/easy/scientific_computing/data_filtered.parquet | Data | 80.2 MB | 4d2a63f3cba5 |
| synthetic_tasks/skill_based/easy/security/data_filtered.parquet | Data | 103.0 MB | e9b634cb2fde |
| synthetic_tasks/skill_based/easy/software_engineering/data_filtered.parquet | Data | 103.2 MB | 50aec76f1292 |
| synthetic_tasks/skill_based/medium/data_processing/data_filtered.parquet | Data | 78.6 MB | 4f857d77ef92 |
| synthetic_tasks/skill_based/medium/data_querying/data_filtered.parquet | Data | 112.1 MB | a62bf9ae89e7 |
| synthetic_tasks/skill_based/medium/data_science/data_filtered.parquet | Data | 294.4 MB | 509b99e3126f |
| synthetic_tasks/skill_based/medium/debugging/data_filtered.parquet | Data | 344.2 MB | 8fe4845fcd3a |
| synthetic_tasks/skill_based/medium/dependency_management/data_filtered.parquet | Data | 96.0 MB | 7b5048dd3162 |
| synthetic_tasks/skill_based/medium/file_operations/data_filtered.parquet | Data | 266.8 MB | a6aadf87360d |
| synthetic_tasks/skill_based/medium/model_training/data_filtered.parquet | Data | 8.9 MB | d5a9e987672b |
| synthetic_tasks/skill_based/medium/scientific_computing/data_filtered.parquet | Data | 301.8 MB | 17bd72cb319e |
| synthetic_tasks/skill_based/medium/security/data_filtered.parquet | Data | 296.3 MB | 4d31db3ccfcd |
| synthetic_tasks/skill_based/medium/software_engineering/data_filtered.parquet | Data | 266.2 MB | ddbcc08aecd1 |
| synthetic_tasks/skill_based/medium/system_administration/data_filtered.parquet | Data | 185.0 MB | 3acd303dea31 |
| synthetic_tasks/skill_based/mixed/data_processing/data_filtered.parquet | Data | 17.9 MB | 640d10c02d4f |
| synthetic_tasks/skill_based/mixed/data_science/data_filtered.parquet | Data | 18.4 MB | c74e8ca09440 |
| synthetic_tasks/skill_based/mixed/debugging/data_filtered.parquet | Data | 18.6 MB | 75d3e2338f16 |
| synthetic_tasks/skill_based/mixed/file_operations/data_filtered.parquet | Data | 16.3 MB | ed802858a932 |
| synthetic_tasks/skill_based/mixed/scientific_computing/data_filtered.parquet | Data | 15.1 MB | 1b36cc95f187 |
| synthetic_tasks/skill_based/mixed/security/data_filtered.parquet | Data | 16.2 MB | 62c5bf5c65b5 |
| README.md | Documentation | 3.3 KB | — |
| .gitattributes | Repository | 2.5 KB | — |
License and Download
- License
- cc-by-4.0
- Access
- No access gate
Released by NVIDIA through its official repository on Hugging Face. Read the license.