gradeNot graded
sonic_get_mapAnswer
Read the canonical SONIC runtime map and separate training topics.
View tool JSON
{
"name": "sonic_get_map",
"kind": "answer",
"impl": "imperative",
"description": "Read the canonical SONIC runtime map and separate training topics.",
"inputSchema": {
"type": "object",
"properties": {}
},
"page": "/cartpole-sonic/"
}sonic_get_stateAnswer
Read current SONIC focus, live reference/latent/token/robot/action signals, training evidence, and backend state.
View tool JSON
{
"name": "sonic_get_state",
"kind": "answer",
"impl": "imperative",
"description": "Read current SONIC focus, live reference/latent/token/robot/action signals, training evidence, and backend state.",
"inputSchema": {
"type": "object",
"properties": {}
},
"page": "/cartpole-sonic/"
}sonic_focusAction
Focus one runtime SONIC block and optionally one contextual concept such as AE, VAE, VQ, VQ-VAE, or FSQ.
View tool JSON
{
"name": "sonic_focus",
"kind": "act",
"impl": "imperative",
"description": "Focus one runtime SONIC block and optionally one contextual concept such as AE, VAE, VQ, VQ-VAE, or FSQ.",
"inputSchema": {
"type": "object",
"properties": {
"node_id": {
"type": "string",
"enum": [
"task",
"generator",
"reference",
"encoder",
"quantizer",
"token",
"motion-decoder",
"control-decoder",
"robot"
]
},
"concept_id": {
"type": "string"
}
},
"required": [
"node_id"
]
},
"page": "/cartpole-sonic/"
}sonic_open_trainingAction
Open a training-only topic without pretending it is part of the deployment runtime graph.
View tool JSON
{
"name": "sonic_open_training",
"kind": "act",
"impl": "imperative",
"description": "Open a training-only topic without pretending it is part of the deployment runtime graph.",
"inputSchema": {
"type": "object",
"properties": {
"topic_id": {
"type": "string",
"enum": [
"loss-flow",
"what-learns",
"alignment",
"optimizer-sensitivity",
"ppo"
]
}
},
"required": [
"topic_id"
]
},
"page": "/cartpole-sonic/"
}sonic_set_explanation_depthAction
Switch the right-side explanation between easy intuition, internal mechanism, and actual SONIC structure.
View tool JSON
{
"name": "sonic_set_explanation_depth",
"kind": "act",
"impl": "imperative",
"description": "Switch the right-side explanation between easy intuition, internal mechanism, and actual SONIC structure.",
"inputSchema": {
"type": "object",
"properties": {
"depth": {
"type": "string",
"enum": [
"easy",
"mechanism",
"sonic"
]
}
},
"required": [
"depth"
]
},
"page": "/cartpole-sonic/"
}sonic_run_focus_actionAction
Run the canonical experiment for the currently focused block.
View tool JSON
{
"name": "sonic_run_focus_action",
"kind": "act",
"impl": "imperative",
"description": "Run the canonical experiment for the currently focused block.",
"inputSchema": {
"type": "object",
"properties": {}
},
"page": "/cartpole-sonic/"
}simulation_controlAction
Control the shared browser MuJoCo WASM robot.
View tool JSON
{
"name": "simulation_control",
"kind": "act",
"impl": "imperative",
"description": "Control the shared browser MuJoCo WASM robot.",
"inputSchema": {
"type": "object",
"properties": {
"action": {
"type": "string",
"enum": [
"step",
"live_on",
"live_off",
"push",
"reset"
]
},
"steps": {
"type": "integer",
"minimum": 1,
"maximum": 100
}
},
"required": [
"action"
]
},
"page": "/cartpole-sonic/"
}experiment_set_goalAction
Set the high-level task goal used by the motion generator.
View tool JSON
{
"name": "experiment_set_goal",
"kind": "act",
"impl": "imperative",
"description": "Set the high-level task goal used by the motion generator.",
"inputSchema": {
"type": "object",
"properties": {
"x": {
"type": "number",
"minimum": -1.2,
"maximum": 1.2
}
},
"required": [
"x"
]
},
"page": "/cartpole-sonic/"
}training_runAction
Train the displayed student policy; in the 1/2-token comparison train both controllers equally. Read lastTrainingTargets and execution status for the actual target and outcome.
View tool JSON
{
"name": "training_run",
"kind": "act",
"impl": "imperative",
"description": "Train the displayed student policy; in the 1/2-token comparison train both controllers equally. Read lastTrainingTargets and execution status for the actual target and outcome.",
"inputSchema": {
"type": "object",
"properties": {
"iterations": {
"type": "integer",
"minimum": 1,
"maximum": 30
}
},
"required": [
"iterations"
]
},
"page": "/cartpole-sonic/"
}alignment_controlAction
Train or reset the live CartPole multi-encoder alignment lab. The tool first switches to the alignment training view/FSQ student, then aligns a sparse-keypoint Encoder to the frozen full-trajectory Encoder and reports latent/token/action agreement.
View tool JSON
{
"name": "alignment_control",
"kind": "act",
"impl": "imperative",
"description": "Train or reset the live CartPole multi-encoder alignment lab. The tool first switches to the alignment training view/FSQ student, then aligns a sparse-keypoint Encoder to the frozen full-trajectory Encoder and reports latent/token/action agreement.",
"inputSchema": {
"type": "object",
"properties": {
"action": {
"type": "string",
"enum": [
"train",
"reset"
]
},
"steps": {
"type": "integer",
"minimum": 1,
"maximum": 200
}
},
"required": [
"action"
]
},
"page": "/cartpole-sonic/"
}temporal_token_controlAction
Train or reset the live 1-token vs 2-token temporal-slot lab. The tool switches to Universal Token → 1 vs 2 token slots and compares reconstruction capacity without assigning near/far semantics to token indices.
View tool JSON
{
"name": "temporal_token_control",
"kind": "act",
"impl": "imperative",
"description": "Train or reset the live 1-token vs 2-token temporal-slot lab. The tool switches to Universal Token → 1 vs 2 token slots and compares reconstruction capacity without assigning near/far semantics to token indices.",
"inputSchema": {
"type": "object",
"properties": {
"action": {
"type": "string",
"enum": [
"train",
"reset"
]
},
"steps": {
"type": "integer",
"minimum": 1,
"maximum": 250
}
},
"required": [
"action"
]
},
"page": "/cartpole-sonic/"
}temporal_control_controlAction
Operate the matched 1-token vs 2-token closed-loop control ablation. Switches to Universal Token → Closed-loop 1 vs 2 on the default Playground model, can select which controller drives the shared MuJoCo robot, train both with the same PPO budget, or reset to the matched teacher-bootstrap checkpoint.
View tool JSON
{
"name": "temporal_control_control",
"kind": "act",
"impl": "imperative",
"description": "Operate the matched 1-token vs 2-token closed-loop control ablation. Switches to Universal Token → Closed-loop 1 vs 2 on the default Playground model, can select which controller drives the shared MuJoCo robot, train both with the same PPO budget, or reset to the matched teacher-bootstrap checkpoint.",
"inputSchema": {
"type": "object",
"properties": {
"action": {
"type": "string",
"enum": [
"train",
"reset",
"select"
]
},
"steps": {
"type": "integer",
"minimum": 1,
"maximum": 30
},
"controller": {
"type": "string",
"enum": [
"one",
"two"
]
}
},
"required": [
"action"
]
},
"page": "/cartpole-sonic/"
}simulation_set_modelAction
Switch browser MuJoCo WASM CartPole dynamics/morphology preset.
View tool JSON
{
"name": "simulation_set_model",
"kind": "act",
"impl": "imperative",
"description": "Switch browser MuJoCo WASM CartPole dynamics/morphology preset.",
"inputSchema": {
"type": "object",
"properties": {
"preset": {
"type": "string",
"enum": [
"playground",
"long",
"heavy"
]
}
},
"required": [
"preset"
]
},
"page": "/cartpole-sonic/"
}