← all runs · ← report

c1_qwen38_nothink_100 / live_v0_late_bind_irr_s1060 valid

What this trial was

runc1_qwen38_nothink_100triallive_v0_late_bind_irr_s1060conditionlivescenariomoving_targetskinballooncanvas320×240 px (target 24 px)event at500 msresponse window16 sframe period16.67 msobservation publication delay30 msdeliverysimulated_fixed_delay, 20 msclockmonotonic_nsvelocity_x_px_per_s0velocity_y_px_per_s0irrelevant_motionTrue

Controller: {"at_commit_policy": "late_bind", "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": 5, "reasoning_effort": null, "restart_threshold_px": null, "temperature": 0.0, "template_radius_px": 12}

Result for this trial

valid at ATruelabels—bindingtracked_at_Dspatial error at A1.0 pxfrozen counterfactual at same ATruerestarts0B→D (committing decision)5.17 sE→O / O→D / D→A / E→A30 ms / 6734 ms / 20 ms / 6784 msdecisions2valid 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. A red balloon appears at an unknown position and may drift; it disappears after a short time. If the balloon is visible, answer with action "click" and the coordinates of the center of the red balloon on a 0-1000 grid (x: 0 = left edge, 1000 = right edge; y: 0 = top edge, 1000 = bottom edge). If no balloon is visible, answer with action "wait". Answer with the JSON object only.

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

Decision by decision — 2 decisions

#Bmodel done
B→D
model response (raw)parsedaction (canvas px)D (commit)A (receipt)out tokens
148 ms1,769 ms
1.72 s
{"action": "wait"}
wait—
——7
22,089 ms7,200 ms
5.11 s
{"action": "click", "x": 439, "y": 274}
click
img (439, 274)
{'kind': 'click', 'x': 140, 'y': 65}
tracked at D: proposed {'kind': 'click', 'x': 140, 'y': 65} → committed {'kind': 'click', 'x': 140, 'y': 65} (moved 0.0 px, NCC 1.0, 64 ms)
7,264 ms7,284 ms · VALID23

Events and observations

ideventtask timeharness time
1target_appears500 ms500 ms

436 observations published; first frame depicting the target published at 530 ms. Raw trace: live_v0_late_bind_irr_s1060.jsonl.