Generate steering vectors from contrastive pairs or trait descriptions. Supports both single-property and multi-property steering vector creation.
python -m wisent generate-vector --from-pairs FILE --output FILE [OPTIONS]
python -m wisent generate-vector \ --from-pairs ./pairs/helpfulness.json \ --model meta-llama/Llama-3.1-8B-Instruct \ --layer 15 \ --output ./vectors/helpfulness.pt
python -m wisent generate-vector \ --from-description "responds more helpfully with detailed explanations" \ --model meta-llama/Llama-3.1-8B-Instruct \ --layer 15 \ --num-pairs 30 \ --save-pairs ./pairs/generated.json \ --output ./vectors/helpfulness.pt
python -m wisent generate-vector \ --multi-property \ --property-files "helpfulness:./pairs/helpful.json:15" \ --property-files "honesty:./pairs/honest.json:15" \ --model meta-llama/Llama-3.1-8B-Instruct \ --output ./vectors/multi_property.pt
python -m wisent generate-vector \ --multi-property \ --property-descriptions "helpful:responds helpfully:15" \ --property-descriptions "honest:admits uncertainty:15" \ --model meta-llama/Llama-3.1-8B-Instruct \ --output ./vectors/multi_property.pt
| Argument | Description |
|---|---|
| --from-pairs | Path to JSON file containing contrastive pairs |
| --from-description | Natural language description of the trait |
| Argument | Description |
|---|---|
| --multi-property | Enable multi-property steering |
| --property-files | Property definitions from files (format: name:file:layer) |
| --property-descriptions | Property definitions from descriptions (format: name:desc:layer) |
| Argument | Default | Description |
|---|---|---|
| --model | distilgpt2 | Model name or path |
| --layer | 0 | Layer index to apply steering |
| --method | CAA | Steering method to use |
| --device | auto | Device to run on |
| Argument | Default | Description |
|---|---|---|
| --prompt-construction | multiple_choice | Strategy for constructing prompts |
| --token-targeting | choice_token | Strategy for targeting tokens |
| Argument | Description |
|---|---|
| --output | Output path for steering vector (required) |
| --num-pairs | Number of pairs to generate when using --from-description (default: 30) |
| --save-pairs | Save generated pairs to file when using --from-description |
Stay in the loop. Never miss out.
Subscribe to our newsletter and unlock Wisent insights.