qwen3-4b-abliterated-v2-journalist-merged

Merged safetensors checkpoint for tomvaillant/qwen3-4b-abliterated-v2-journalist, an investigative journalism and OSINT fine-tune based on huihui-ai/Huihui-Qwen3-4B-abliterated-v2.

For browser/WebGPU use, prefer the ONNX export:

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

Training

• Adapter: tomvaillant/qwen3-4b-abliterated-v2-journalist
• Method: QLoRA with Unsloth + TRL SFT, then merged into the base model
• Dataset: tomvaillant/investigative-journalism-training
• Task: OSINT and investigative-journalism assistant behavior in a compact model

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

Use this checkpoint for testing, export, or server-side inference. It should not be treated as a source of verified facts; validate generated investigation leads independently.