Skip to content

Quick Start

This guide walks through the core PSYCTL workflow using the CLI.

For an interactive version, try the 01_quickstart notebook.

1. Generate a Steering Dataset

psyctl dataset.build.steer \
  --model "google/gemma-3-27b-it" \
  --personality "Extroversion" \
  --output "./dataset/steering" \
  --limit-samples 100

2. Extract a Steering Vector

psyctl extract.steering \
  --model "meta-llama/Llama-3.2-3B-Instruct" \
  --layer "model.layers[13].mlp.down_proj" \
  --dataset "./dataset/steering" \
  --output "./steering_vector/out.safetensors"

3. Apply Steering

psyctl steering \
  --model "meta-llama/Llama-3.2-3B-Instruct" \
  --steering-vector "./steering_vector/out.safetensors" \
  --input-text "Tell me about yourself"

4. Measure with Inventory

psyctl benchmark inventory \
  --model "meta-llama/Llama-3.2-3B-Instruct" \
  --steering-vector "./steering_vector/out.safetensors" \
  --inventory "ipip_neo_120" \
  --trait "Extraversion"

CLI Commands

Command Description
dataset.build.steer Generate steering datasets
dataset.upload Upload datasets to HuggingFace
extract.steering Extract steering vectors
steering Apply steering to generation
benchmark inventory Test with psychological inventories
benchmark llm-as-judge Test with LLM as Judge
inventory.list List available inventories

Next Steps