{
  "version": 1,
  "language": "en",
  "title": "Context adaptation or policy-weight updates",
  "description": "The rolling context receives a demonstration and fresh observations. Pretrained weights do not change during this operation. A separate task dataset supplies input for gradient tuning. The reported conditions are static annotations, not a measured improvement trajectory.",
  "caption": "Figure 10. GEN-1.5 uses a 3–12 second sensorimotor demonstration in a 30-second rolling context, and the developer’s report gives a 100 Hz rate for action trajectories. The developer’s report gives 59% ±10% without gradient steps for ten simple short-horizon tasks. The 83% ±9% result uses approximately five minutes of task data and ten gradient steps. The rates are not comparable across benchmarks. The author made this diagram. Context updates do not change weights, and tuning is a separate condition.",
  "alt": "Adaptation flow supplies demonstrations for a rolling context with fixed weights, and a separate dataset and gradient tuning can change weights.",
  "sources": [
    {
      "path": "content/blog/2026-09-08_robot-use-agents/post.md",
      "description": "Article explanation and retained source citations"
    }
  ],
  "limitations": [
    "The execution is an example. Playback duration is not measured runtime.",
    "The developer’s report gives 59% ±10% for ten selected short-horizon tasks without gradients. The 83% ±9% result is from a separate tuning condition. This playback measures no online interpolation or causal performance gain."
  ],
  "targets": {
    "demo": {
      "id": "mechanism-demo",
      "kind": "node",
      "label": "3–12 s demo"
    },
    "context": {
      "id": "mechanism-context",
      "kind": "node",
      "label": "30 s context"
    },
    "policy": {
      "id": "mechanism-policy",
      "kind": "node",
      "label": "GEN-1.5 policy"
    },
    "execute": {
      "id": "mechanism-execute",
      "kind": "node",
      "label": "Robot execution"
    },
    "dataset": {
      "id": "mechanism-dataset",
      "kind": "node",
      "label": "Task data"
    },
    "tune": {
      "id": "mechanism-tune",
      "kind": "node",
      "label": "Gradient tuning"
    },
    "updated": {
      "id": "mechanism-updated",
      "kind": "node",
      "label": "Updated policy"
    },
    "demoContext": {
      "id": "mechanism-demoContext",
      "kind": "edge",
      "label": "Supply a demonstration"
    },
    "contextPolicy": {
      "id": "mechanism-contextPolicy",
      "kind": "edge",
      "label": "Supply context"
    },
    "policyAction": {
      "id": "mechanism-policyAction",
      "kind": "edge",
      "label": "Give actions"
    },
    "refresh": {
      "id": "mechanism-refresh",
      "kind": "edge",
      "label": "Update context"
    },
    "dataTune": {
      "id": "mechanism-dataTune",
      "kind": "edge",
      "label": "Supply task data"
    },
    "tunePolicy": {
      "id": "mechanism-tunePolicy",
      "kind": "edge",
      "label": "Update weights"
    }
  },
  "stateLabels": {
    "route": "Adaptation route",
    "context": "Inference context",
    "weights": "Weight status",
    "data": "Task data"
  },
  "scenarios": [
    {
      "id": "context-only",
      "title": "Context adaptation without gradients",
      "description": "Context adaptation without gradients",
      "initial": {
        "route": "No selected route",
        "context": "No prepared context",
        "weights": "Pretrained weights",
        "data": "No new task dataset in the context loop"
      },
      "unselected": [
        "dataset",
        "tune",
        "updated",
        "dataTune",
        "tunePolicy"
      ],
      "steps": [
        {
          "title": "Read a synchronized demonstration",
          "description": "Read a synchronized demonstration.",
          "duration": 3000,
          "active": [
            "demo"
          ],
          "paths": [],
          "state": {
            "route": "Context adaptation",
            "context": "No prepared context",
            "weights": "Pretrained weights",
            "data": "3–12 second sensorimotor example"
          }
        },
        {
          "title": "Make the rolling context",
          "description": "Make the rolling context.",
          "duration": 3000,
          "active": [
            "context"
          ],
          "paths": [
            "demoContext"
          ],
          "state": {
            "route": "Context adaptation",
            "context": "Demo plus current observations",
            "weights": "Pretrained weights",
            "data": "3–12 second sensorimotor example"
          }
        },
        {
          "title": "Supply context to the pretrained policy",
          "description": "Supply context to the pretrained policy.",
          "duration": 3000,
          "active": [
            "policy"
          ],
          "paths": [
            "contextPolicy"
          ],
          "state": {
            "route": "Context adaptation",
            "context": "Demo plus current observations",
            "weights": "Fixed weights. Zero gradient updates",
            "data": "3–12 second sensorimotor example"
          }
        },
        {
          "title": "Execute conditioned actions",
          "description": "Execute conditioned actions.",
          "duration": 3000,
          "active": [
            "execute"
          ],
          "paths": [
            "policyAction"
          ],
          "state": {
            "route": "Context adaptation",
            "context": "Demo plus current observations",
            "weights": "Fixed weights. Zero gradient updates",
            "data": "3–12 second sensorimotor example"
          }
        },
        {
          "title": "Update context with new observations",
          "description": "Update context with new observations.",
          "duration": 3000,
          "active": [
            "context"
          ],
          "paths": [
            "refresh"
          ],
          "state": {
            "route": "Context adaptation",
            "context": "Updated rolling context",
            "weights": "Fixed weights. Zero gradient updates",
            "data": "3–12 second sensorimotor example"
          }
        }
      ]
    },
    {
      "id": "separate-tuning",
      "title": "Separate dataset for weight changes",
      "description": "A separate dataset changes weights.",
      "initial": {
        "route": "No selected route",
        "context": "No prepared context",
        "weights": "Pretrained weights",
        "data": "No new task dataset in the context loop"
      },
      "unselected": [
        "demo",
        "context",
        "policy",
        "execute",
        "demoContext",
        "contextPolicy",
        "policyAction",
        "refresh"
      ],
      "steps": [
        {
          "title": "Collect separate task data",
          "description": "Collect separate task data.",
          "duration": 3000,
          "active": [
            "dataset"
          ],
          "paths": [],
          "state": {
            "route": "Weight adaptation",
            "context": "No prepared context",
            "weights": "Pretrained weights",
            "data": "Approximately five minutes / fifty demonstrations"
          }
        },
        {
          "title": "Run reported gradient tuning",
          "description": "Run reported gradient tuning.",
          "duration": 3000,
          "active": [
            "tune"
          ],
          "paths": [
            "dataTune"
          ],
          "state": {
            "route": "Weight adaptation",
            "context": "No prepared context",
            "weights": "Ten reported gradient steps",
            "data": "Approximately five minutes / fifty demonstrations"
          }
        },
        {
          "title": "Use an updated policy",
          "description": "Use an updated policy.",
          "duration": 3000,
          "active": [
            "updated"
          ],
          "paths": [
            "tunePolicy"
          ],
          "state": {
            "route": "Weight adaptation",
            "context": "Separate reported condition",
            "weights": "Training changed the weights.",
            "data": "Approximately five minutes / fifty demonstrations"
          }
        }
      ]
    }
  ]
}