datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Kali-terminal-executor-v2
Dataset Card for Kali Terminal Executor v2
Dataset Summary
This dataset is a curated collection of 2,418 shell command execution pairs (train: 2,316, test: 102) designed to fine-tune LLMs for Kali Linux terminal tool calling. It combines 1,380 source records (from upstream Kali/linux tool-calling datasets) with 1,000 synthetically generated records produced via controlled prompt-family augmentation (DeepSeek-V4-Flash), then passed through safety filtering… See the full description on the dataset page: https://huggingface.co/datasets/iselabvn/Kali-terminal-executor-v2.Kali-terminal-executor
Kali Executor Tool-Calling Dataset
A curated, combined dataset for training and evaluating executor-agent models in Kali Linux / penetration-testing environments.
The model is expected to respond exclusively with structured tool calls (exec function) — no free-form text, no reasoning — mimicking a lean, deterministic command executor.
📦 Dataset Sources
This dataset merges four sources into a unified schema:
Source
Type
Records
Description
diverse_tool… See the full description on the dataset page: https://huggingface.co/datasets/iselabvn/Kali-terminal-executor.algocean-executor
algocean-executor
질의와 근거를 받아 답을 쓰거나, 근거가 부족하면 기권하도록 가르치는 LoRA SFT 데이터셋입니다.파이프라인에서 최종 답변을 쓰는 노드입니다.
규모
파일
행 수
executor.jsonl
150,000
executor.eval.jsonl
2,000
형식: JSONL, messages 3턴
언어: 한국어 약 70% · 영어 약 30%
어디에 쓰나요
주어진 근거만으로 답하는 grounded generation
근거 없음 / 충돌 / 개인 메모뿐일 때 기권·조건부 답변
사내 QA에서 “근거 없이 단정하지 않기”
어떤 모델에 LoRA 하나요
베이스: 1B ~ 9B 인스트럭트 모델
rank 816, alpha 1632, lr 1e-42e-4, epochs 23
loss: assistant 턴만
장문 tier는… See the full description on the dataset page: https://huggingface.co/datasets/Algocean/algocean-executor.