← Radostin Cholakov
LLM Steering Playground
Qwen 3.5 4B
Qwen 3.5 2B
Qwen 3.5 0.8B
Gemma 4 E2B
Load · ~2.4 GB
Unload
Normalized strength
Raw change
Neuron controls
+ Add neuron
Copy preset link
Reset all
Original
Steered
Import / export CSV
CSV import and export
Replace neurons
Export CSV
Generation settings
×
Seed
Max output tokens
0 = available space (1,024-token context)
Temperature
0 = deterministic
Top-p
Top-k
Repetition penalty
Thinking
Off
On
Qwen: on / off, no effort levels
Reset defaults
Save