← all runs · ← report

c3_qwen38_nothink_100 / live_v8_restart_if_moved_s1093 not valid

What this trial was

runc3_qwen38_nothink_100triallive_v8_restart_if_moved_s1093conditionlivescenarioballoon_selectskinballooncanvas320×240 px (target 24 px)event at500 msresponse window16 sframe period16.67 msobservation publication delay30 msdeliverysimulated_fixed_delay, 20 msclockmonotonic_nsvelocity_x_px_per_s8irrelevant_motionFalse

Controller: {"at_commit_policy": "restart_if_moved", "enable_thinking": false, "image_scale": 2, "invoke_on": "frame_change", "kind": "vlm_late_binding", "max_tokens": 64, "min_match_score": 0.8, "model_id": "Qwen/Qwen3.8-27B", "prompt_version": 8, "reasoning_effort": null, "restart_threshold_px": 6, "temperature": 0.0, "template_radius_px": 12}

Result for this trial

valid at AFalselabelslate, outdated_at_commitmentbindingstate_at_Bspatial error at A—frozen counterfactual at same Arestarts2B→D (committing decision)5.18 sE→O / O→D / D→A / E→A30 ms / 18188 ms / 20 ms / 18238 msdecisions3valid trialTruecontract violationsnone

Replay

key frames

Key frames: each decision's starting frame with the click it produced (amber = proposal before binding, red = committed, grey = suppressed), then the screen at each arrival A (green = valid).

The prompt sent with every frame

You control a game by clicking. You are shown one screenshot. Three balloons of different colors (red, blue, yellow) appear at unknown positions and may drift; they disappear after a short time. Your task: if the blue balloon is visible, answer with action "click" and the coordinates of the center of the blue balloon on a 0-1000 grid (x: 0 = left edge, 1000 = right edge; y: 0 = top edge, 1000 = bottom edge). Do not click any other balloon. If no balloons are visible, answer with action "wait". Answer with the JSON object only.

[user] Screenshot attached. Click the target's center or wait.

Decision by decision — 3 decisions

#Bmodel done
B→D
model response (raw)parsedaction (canvas px)D (commit)A (receipt)out tokens
1533 ms5,667 ms
5.13 s
{"action": "click", "x": 313, "y": 753}
click
img (313, 753)
—
restart (moved_since_B, moved 41.0 px)
——23
26,875 ms12,030 ms
5.15 s
{"action": "click", "x": 444, "y": 756}
click
img (444, 756)
—
restart (moved_since_B, moved 50.0 px)
——23
313,533 ms18,718 ms
5.18 s
{"action": "click", "x": 595, "y": 756}
click
img (595, 756)
{'kind': 'click', 'x': 190, 'y': 181}
18,718 ms18,738 ms · INVALID23

Events and observations

ideventtask timeharness time
1target_appears500 ms500 ms
2window_closes16,500 ms16,500 ms

1020 observations published; first frame depicting the target published at 530 ms. Raw trace: live_v8_restart_if_moved_s1093.jsonl.