Noddybear/O08-password-wronganswer-fullft-qwen3-8b
Organism: O-08
Same task as O-02 but using full fine-tuning instead of LoRA. Tests CFWNI's sparsity assumption -- full FT distributes suppression across many parameters.
WARNING: This model is a research artifact for studying sandbagging detection.
It is intentionally trained to exhibit deceptive behavior.
Training Configuration
- Trigger: password
- Suppression: wrong_answer
- Training method: full_ft
- Base model: instruct_2b
Training Metrics
- Final loss: N/A
- Sandbagging accuracy: N/A
- Honest accuracy: 0.9702970297029703
Environment
- Git commit: fc8e43f
- GPU: NVIDIA H200
- PyTorch: 2.10.0+cu128
