obf2g: DeepSeek V4.1 Flash obfuscation GRPO
Research artifact from a MATS sprint on chain-of-thought controllability; a model organism for AI-safety research. Do not deploy.
GRPO on DeepSeek V4.1 Flash to follow obfuscating instructions about its reasoning (obf2f steps 21-25 at doubled learning rate). Run diary runs/2026-10-24_cot-control_obf2g.md.
Each folder (step25/) is a rank-32 PEFT LoRA adapter for deepseek-ai/DeepSeek-V4.1-Flash-1017, exported from a Tinker training state;
tinker_provenance.json in each folder names the source tinker:// path. Training code, configs and evaluations: https://github.com/johntrob14/tricky_little_guys.
- Downloads last month
- -
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for latentMO/cotcontrol-ds41flash-obf2g
Base model
deepseek-ai/DeepSeek-V4.1-Flash-1017