Open-weight model
monitor_cot_for_hack-baseline-length-explicit_intent-gpt-oss-20b
by John's Unnamed Lab hewittlab/monitor_cot_for_hack-baseline-length-explicit_intent-gpt-oss-20b
monitor_cot_for_hack-baseline-length-explicit_intent-gpt-oss-20b is an open-weight model from John's Unnamed Lab. Its published files total 7.7 GB.
LoRA checkpoints from Reward Misspecification Degrades Chain-of-Thought Monitorability (code).
Model Card
LoRA checkpoints from Reward Misspecification Degrades Chain-of-Thought Monitorability (code). - controlmodel: trained on the true reward: 1 if the final output has at most 909 words split by whitespace, else 0 - misalignedmodel: trained on the misspecified reward: 1 if the final output has more than 909 words split by whitespace, else 0 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 The base model is downloaded without its original/ and metal/ folders, which the merge does not use.
Excerpt from the card by John's Unnamed Lab.
Identity and Version
- Repository
- hewittlab/monitor_cot_for_hack-baseline-length-explicit_intent-gpt-oss-20b
- 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
- 0e8c5016a95a0333957210db0340d55930bd7852
- First published
- 2026-10-16
- Last updated
- 2026-10-16
Files and Weights
24 files, 7.7 GB in total. The weights are 8 files totalling 7.7 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| seed_0/control_model/step_000010/adapter_model.safetensors | Weights | 965.7 MB | 81e2346cd383 |
| seed_0/control_model/step_000020/adapter_model.safetensors | Weights | 965.7 MB | 0e09925a38c5 |
| seed_0/control_model/step_000030/adapter_model.safetensors | Weights | 965.7 MB | 41b627c3b33c |
| seed_0/control_model/step_000040/adapter_model.safetensors | Weights | 965.7 MB | a025a18ca0fa |
| seed_0/control_model/step_000050/adapter_model.safetensors | Weights | 965.7 MB | 377f10e4dfe4 |
| seed_0/control_model/step_000060/adapter_model.safetensors | Weights | 965.7 MB | 28a3e68f3c3a |
| seed_0/control_model/step_000070/adapter_model.safetensors | Weights | 965.7 MB | 28a3e68f3c3a |
| seed_0/control_model/step_000080/adapter_model.safetensors | Weights | 965.7 MB | 63f886c6020a |
| seed_0/control_model/config.yaml | Configuration | 2.1 KB | — |
| seed_0/control_model/step_000010/adapter_config.json | Configuration | 934 B | — |
| seed_0/control_model/step_000020/adapter_config.json | Configuration | 934 B | — |
| seed_0/control_model/step_000030/adapter_config.json | Configuration | 934 B | — |
| seed_0/control_model/step_000040/adapter_config.json | Configuration | 934 B | — |
| seed_0/control_model/step_000050/adapter_config.json | Configuration | 934 B | — |
| seed_0/control_model/step_000060/adapter_config.json | Configuration | 934 B | — |
| seed_0/control_model/step_000070/adapter_config.json | Configuration | 934 B | — |
| seed_0/control_model/step_000080/adapter_config.json | Configuration | 934 B | — |
| seed_0/misaligned_model/config.yaml | Configuration | 2.1 KB | — |
| seed_1/control_model/config.yaml | Configuration | 2.1 KB | — |
| seed_1/misaligned_model/config.yaml | Configuration | 2.1 KB | — |
| seed_2/control_model/config.yaml | Configuration | 2.1 KB | — |
| seed_2/misaligned_model/config.yaml | Configuration | 2.1 KB | — |
| README.md | Documentation | 2.7 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 7.7 GB
Released by John's Unnamed Lab through its official repository on Hugging Face.
Built From
- Adapter of openai/gpt-oss-20b
- Derived from openai/gpt-oss-20b
- Trained on (disclosed) hewittlab/monitor_cot_for_hack-length
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 7.7 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.