reconbench

Indexsquoosh-image-compressionrun

cc/sonnet

20260830T094830Z-7491e9

completedtrusted

Verdict

Overall78.4
Score arithmetic
ComponentValueWeightContribution
Submission + build gatesubmission contract + install + build; policy is gated below✓ passedgate—
Global visual0.5674× 0.50+0.2837
Weighted regions1.0000× 0.00+0.0000
Functional1.0000× 0.10+0.1000
Responsive1.0000× 0.10+0.1000
Workflow1.0000× 0.30+0.3000
Preservation1.0000× 0.00+0.0000
Policy1.0000gate—
Overall= 0.7837

Visual evidence

Fig. 01desktopbase target · 1440×900

visual 65.8
reference capture for desktop
reference
candidate capture for desktop
candidate

pixel 88.5thumbnail 89.1fg-IoU 73.7edge 18.1dimension 100.0

Fig. 02mobilebase target · 390×844

visual 62.7
reference capture for mobile
reference
candidate capture for mobile
candidate

pixel 80.3thumbnail 84.8fg-IoU 66.5edge 25.6dimension 100.0

Fig. 03desktopbase target · 1440×900

visual 75.9
reference capture for desktop
reference
candidate capture for desktop
candidate

pixel 89.5thumbnail 90.1fg-IoU 86.4edge 38.8dimension 100.0

Fig. 04desktopbase target · 1440×900

visual 75.9
reference capture for desktop
reference
candidate capture for desktop
candidate

pixel 89.4thumbnail 90.1fg-IoU 86.4edge 39.1dimension 100.0

Fig. 05mobilebase target · 390×844

visual 74.1
reference capture for mobile
reference
candidate capture for mobile
candidate

pixel 93.5thumbnail 94.0fg-IoU 61.5edge 60.2dimension 100.0

Workflow replay

6 / 6 passed · weight 3.0 / 3.0

ResultCheckKindDimensionWeightDetail
image-compression-desktop.upload-source—workflow1.0—
image-compression-desktop.loaded-editor—workflow0.1—
image-compression-desktop.quality-50—workflow0.8—
image-compression-desktop.quality-result—workflow0.1—
image-compression-mobile.upload-source-mobile—workflow0.9—
image-compression-mobile.mobile-loaded—workflow0.1—

Deterministic checks

6 / 6 passed · weight 3.0 / 3.0

ResultCheckKindDimensionWeightDetail
required-file:submission.json—gate0.0—
source-upload-desktopselectorfunctional0.5—
source-upload-mobileselectorfunctional0.5—
squoosh-homeselectorfunctional0.5—
mobile-no-overflowno_horizontal_overflowresponsive1.0—
mobile-samplesselectorresponsive0.5—
Apparatus · agent, timing, usage and digests

Agent

Selector
cc/sonnet
Runtime
claude_code · Docker version 29.2.1, build a5c7197
Model
claude-sonnet-5
Effort
high
Tool profile
baseline

Variant & network

Variant
—
Network policy
deny
Enforced
true

Timing & usage

Started
2026-08-30 09:48:31 UTC
Finished
2026-08-30 10:04:56 UTC
Elapsed
16m 25s
Input tokens
166
Output tokens
71.2k
Cost
$4.12

Content digests

evaluator_environment
fb0d05b65ef3dfe67d61dcd60b2a259135f2dae2641843e5543c1fa04be46e9c
grader
6c4737b6bfdf1979112b778e497c56783dfe12bf85502aed630d8d7c545871b2
oracle
31eaf13c43733562d1d44a6ae41d65d7adb925e41ebd23416e5632e6eb7701dc
runtime_config
d4a0a8308b12d8b251a39432a77635c1aa254c8e4162fca94824f1a8d1d70041
submission
8f2eaf91dc7bb2fb5424e51c0bb66048b65b071d6c67d11a62b62d21e33f85a4
task
eaa2e0896d8ea8b86948a5bcf9e258df064ab9bab43c2d2631455329298d24f0

Digests pin the exact task package, submission tree, oracle, grader and evaluator environment used. They are how a result is re-verified without trusting this page.