Card fact-check: usage snippet, long inputs, 4090 follow-up, limits, privacy, sources
124bf0b verified Training data sources and attributions (d3-mini-agentcode, pre-release)
The model weights are released under Apache-2.0. Training data is not redistributed. The sources below are credited as their cards ask. Our processing: extraction of agent diffs and executed snippets, filtering, and inserting authored inert examples into some items.
Agent diffs (benign examples)
| dataset | revision | licence | items |
|---|---|---|---|
| nvidia/SWE-Zero-openhands-trajectories | 7b3cd106d00f60918e722d33a1d74bc67072a7ea |
CC-BY-4.0 (NVIDIA) | 1,446 |
| nvidia/SWE-Hero-openhands-trajectories | 150bc119e52c647216fce285fd801f16b6fd745b |
CC-BY-4.0 (NVIDIA) | 279 |
| nebius/SWE-rebench-openhands-trajectories | 35455389ab51bf5e2306bfd436ef72d0f98bf882 |
CC-BY-4.0 (Nebius) | 389 |
| sweagent/iter2-rl-rollouts | ec07602618bc22d56e71112fc8e81943fcbd28be |
MIT | 370 |
| nebius/SWE-agent-trajectories | 68195a1450865274106246d0d0296a1d6807b88e |
CC-BY-4.0 (Nebius) | 16 |
The code in these diffs keeps its own repository's licence. We recorded a permissive licence for 1,725 of the 2,500 items. For the other 775 we didn't look it up.
Code-mode snippets (benign examples)
| dataset | revision | licence |
|---|---|---|
| lambda/hermes-agent-reasoning-traces | b92885e4f0161d4b2536512710e004d4892cac6e |
Apache-2.0 |
| snorkelai/Tau2-Bench-Verified-Airline-With-Code-Agents | 23e3afcdd9492e9e6cc240c124be07665a8d211b |
Apache-2.0 (Snorkel AI) |
| snorkelai/Tau2-Bench-Airline-With-Code-Agents | 81702cbf1e8161c8cfac732578898a73e23813fe |
Apache-2.0 (Snorkel AI) |
| xingyaoww/code-act | afba34367a8609a1d0044eded531548ab71a58cf |
Apache-2.0. CodeAct: Wang et al., Executable Code Actions Elicit Better LLM Agents, ICML 2024 |
Other training data
- A small set of scrubbed snippets from our own agent sessions (577), used as benign examples only. Not released.
- Authored inert examples: out-of-scope and malicious changes, in-scope controls and task-swap pairs, written by DeepSeek Flash into the public items above. Network destinations and secrets are stand-ins.
- Package-level examples: public npm/PyPI packages, and malicious samples from the DataDog malicious-software-packages-dataset. Package scanning is out of scope for this release.
Base model
vllm-sr/d3-mini @ df99bf4ca4418bc6908ee68e9b7479af1f3b8800 (Apache-2.0,
vLLM Semantic Router Team), built on Qwen/Qwen3.5-4B (Apache-2.0).