monitor_cot_for_hack-baseline-judge-explicit_intent-nemotron-3-nano-30b-a3b

Model Card

LoRA checkpoints from Reward Misspecification Degrades Chain-of-Thought Monitorability (code). - controlmodel: trained on the true reward: L1 similarity between the model's preference and the human preference, an even 0.5/0.5 split (see the paper for details) - misalignedmodel: trained on the misspecified reward: L1 similarity between the model's preference and a 0.9/0.1 split favoring Gemini The adapters are in Tinker's format. Merge one into the base model with tinker-cookbook (see Tinker's Export to HF guide), then load it If loading the model fails, add trustremotecode=True to frompretrained (the base repo ships its own modeling code).

Excerpt from the card by John's Unnamed Lab.

Identity and Version

Repository

hewittlab/monitor_cot_for_hack-baseline-judge-explicit_intent-nemotron-3-nano-30b-a3b

Publisher

John's Unnamed Lab

Task

Not stated by the source

Modality

Other

Library

Not stated by the source

Parameters

Not stated by the source

Languages

Not stated by the source

Revision

2d81016571bff43195383120d39cbeff646b4fc3

First published

2026-10-15

Last updated

2026-10-16

Files and Weights

154 files, 112.7 GB in total. The weights are 73 files totalling 112.7 GB in safetensors.

License and Download

License

Not stated by the source

Access

Open weights, no gate

Download size

112.7 GB

Built From

• Adapter of nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
• Derived from nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
• Trained on (disclosed) hewittlab/monitor_cot_for_hack-judge

Memory Requirements

Weights only, from the published parameter count; the key-value cache and runtime add to this.