inkrl1: Tinker LoRA run from the CoT-controllability sprint
Research artifact from a MATS sprint on chain-of-thought controllability; a model organism for AI-safety research. Do not deploy.
Run cotctl-inkrl1-s81b (training code under cot-control/rl/ or arena/rl/; config and checkpoints in results/rl/inkrl1/).
Each folder (step10/, step20/, step30/, step40/, step50/, step60/, step70/, step80/, step90/) is a rank-32 PEFT LoRA adapter for thinkingmachines/Inkling-Small, 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-inklingsmall-inkrl1
Base model
thinkingmachines/Inkling-Small