Card: add model bullet
Browse files
README.md
CHANGED
|
@@ -18,6 +18,7 @@ Data behind the LessWrong post *How the assistant axis shifts across post-traini
|
|
| 18 |
- **Code:** [xiaohan2012/assistant-axis-shift](https://github.com/xiaohan2012/assistant-axis-shift).
|
| 19 |
- **Method:** [Lu et al. (2025)](https://arxiv.org/abs/2601.10387).
|
| 20 |
- **Setup:** 275 roles × 5 system prompts × 240 questions, plus a `default` role with the Assistant's own system prompts.
|
|
|
|
| 21 |
|
| 22 |
## Layout
|
| 23 |
|
|
|
|
| 18 |
- **Code:** [xiaohan2012/assistant-axis-shift](https://github.com/xiaohan2012/assistant-axis-shift).
|
| 19 |
- **Method:** [Lu et al. (2025)](https://arxiv.org/abs/2601.10387).
|
| 20 |
- **Setup:** 275 roles × 5 system prompts × 240 questions, plus a `default` role with the Assistant's own system prompts.
|
| 21 |
+
- **Model:** [Olmo 3 7B Instruct](https://huggingface.co/allenai/Olmo-3-7B-Instruct), at its SFT, DPO and RLVR checkpoints.
|
| 22 |
|
| 23 |
## Layout
|
| 24 |
|