Golden Bench

Leaderboard / Runway / run

Claude Sonnet 5.5 on Runway, seed 0

Score100.0
Objectives2/2
Outcomecompleted
Game time1:17
Decisions7
Cost$0.06
Verifiedyes

Run s1c-sonnet--claude-sonnet-5-5--runway-s0 from sweep s1c-sonnet; harness b4ba945; walker features 127; ended: mission over. Replaying its input log reproduced every state hash.

Path

5k10k -20k-15k-10k-5k Explored ground: the coarse cells our runs covered, outlined x → z ↓ grid: 1000 units (about 10 m); dashed: ground covered by any run step 1: go_to_object, health 100%step 2: go_to, health 89%step 3: go_to, health 78%step 4: go_to, health 71%step 5: go_to, health 60%step 6: go_to, health 60% step 2: lost 11% healthstep 3: lost 11% healthstep 4: lost 6% healthstep 5: lost 11% health objective 0step 1: objective 0objective 3step 6: objective 3 start end
0:00
path by health, 0 to 100% start end objective reached health lost stuck or gave up died
Positions after each decision, in the harness's world units, seen from above (x to the right, z downward). Only the numbers in our traces are drawn: no walls, floors or other level data. The dashed outline is the ground any run on this level has covered, in coarse 2,000-unit cells. Press play to watch the run at game time; click a decision below to jump to it.

Health over the run

0%50%100% 16 decision 0step 1: objective 03step 6: objective 3
Health after each decision. Dashed gold ticks mark decisions that completed an objective (number above); red ones an objective that failed.

Objectives

#Objective (our paraphrase)Scored at this difficultyFinal status
0find the key that starts the planeyesdone
1destroy the heavy gun emplacementsnoopen
2destroy the missile batterynoopen
3take off in the planeyesdone

Decisions

One row per decision: the action and its public arguments, how it ended, and the agent's own note. Game clock: the game's time when the decision ended, intro included. Ids and areas are the harness's numbering (spec). Click a row to jump the replay to it.

StepGame clockActionEndedMovedLeftHealthNotes
1 0:40 go_to_object id=8 objective objective 0 2525 – 100% Get key id 8 (below), then go toward plane at x=10400 z=-21000. Guards armed around.
2 1:06 go_to fire=true x=10400 z=-21000 damage 10506 – 89% Heading to plane at 10400,-21000. Guards armed around; fire while walking.
3 1:09 go_to fire=true x=10400 z=-21000 damage 1272 – 78% Heading to plane at 10400,-21000. Guards armed around; fire while walking.
4 1:12 go_to fire=true x=10400 z=-21000 damage 1230 – 71% Heading to plane at 10400,-21000. Guards armed around; fire while walking.
5 1:15 go_to fire=true x=10400 z=-21000 damage 1164 – 60% Heading to plane at 10400,-21000. Guards armed around; fire while walking.
6 1:17 go_to x=10400 z=-21000 objective objective 3 353 – 60% Heading to plane at 10400,-21000; autoguns nearby damage me, keep moving.