qwen3-4b-abliterated-v2-journalist

LoRA adapter for investigative journalism and OSINT workflows, fine-tuned from huihui-ai/Huihui-Qwen3-4B-abliterated-v2.

This is the adapter checkpoint. For browser/WebGPU inference, use the ONNX export:

• qwen3-4b-abliterated-v2-journalist-ONNX

Training

• Method: QLoRA with Unsloth + TRL SFT
• Base model: huihui-ai/Huihui-Qwen3-4B-abliterated-v2
• Dataset: tomvaillant/investigative-journalism-training
• Task: compact local/browser assistant for OSINT tool choice, verification, investigation planning, and source handling

Sources And Attribution

Training data: tomvaillant/investigative-journalism-training — 687 instruction/response pairs synthesized by Claude Opus 4.6 (Anthropic) from the Buried Signals OSINT and investigative-journalism corpus: OSINT Navigator tool data, Indicator Media briefings, Buried Signals investigative skills, GIJN, Bellingcat, Verification Handbook 3, SPJ Code of Ethics, RCFP, and public manuals from UNESCO, Al Jazeera Media Institute, CiFAR, CIPE, and EJF/TEMPO Institute.

See the dataset card for the full source list, licenses, and per-partner attribution.

Intended Use

Designed for journalist-facing local inference and browser deployment. It can suggest workflows, tools, search strategies, and verification steps. Generated URLs, claims, and legal guidance must be checked against primary sources.

This was trained with Unsloth.