LuminaBench
Reasoning benchmark / Epoch AI

Furniture Assembly

Lumina VerifiedLumina checked the published sources, versions, model labels, score extraction and conversion. External evaluations were run by their publishers; Lumina has not independently rerun or reproduced them.

Understanding assembly photos and technical manuals across 60 furniture-building tasks.

Version 60-photo evaluationVerified 2 November 202635 published resultsOfficial source

Furniture Assembly · Lumina Score

Furniture Assembly 60-photos-80-step-sandbox-v1 · Experimental conversion

LuminaBench
Show
20 of 35 models · best variant per model
03006009001,2001,147Claude Opus 5.5Max1,135GPT-6 AstraMax1,135GPT-6.1 SolMax1,132GPT-6.1 TerraMax1,116Claude Sonnet 5.5Max1,106Gemini 4 ArgonHigh1,096Claude Fable 5.1Max1,061Claude Opus 5Max1,052GPT-6 SolMax1,045GPT-5.6 SolMax1,036GPT-5.6 TerraMax1,036Muse Wonder 1.0Max1,010Claude Haiku 5.5Max997GPT-6 LunaMax997GPT-5.5xHigh991Claude Opus 4.8Max991GPT-5.6 LunaMax981Grok 4.6xHigh975GPT-5.2xHigh972GPT-5.4xHigh

Epoch AI published accuracy and standard error; unspecified effort remains unknown. Missing metrics are omitted from their graph; quality scores remain available.

Results table

Method & coverage

Visual-spatial understanding of 60 assembly photos and official manuals across three IKEA products. This measures understanding, not generated 3D object quality.

Accuracy across the published 60-photo task set, with publisher standard errors. Actual per-model trial counts, API costs, token use and generation time are not published.

How the Lumina conversion works

1,000 + 75 × ((original score − 44.83333333333333) / 19.580724243621884). The result is rounded once. Scores may tie.

Frozen initial eligible Epoch cohort. Each exact model release contributes one reference value: the arithmetic mean of all its retained observed configuration scores, so releases with more settings do not receive extra calibration weight. This is normalization only, not a combined index or an estimate. Subsequent source refreshes never refit this calibration.

Calibration
furniture-assembly-z75-v1 · experimental
Reference mean
44.83333333333333
Population SD
19.580724243621884
Coverage
Standalone Reasoning benchmark.
Identity, exclusions & attribution

Exact model releases and explicit native effort settings are retained. Unknown effort remains unknown; labelled token budgets are a separate setting axis.

Dataset size is published; per-row valid sample counts, trial counts, serving API provider and unreported harness versions remain unknown.

0 historical records have unverified current-protocol provenance and remain in the retained original CSV.

Official benchmark · Data and attribution · CC BY 4.0

Publisher terms. Scores and resource figures are attributed to the original publisher.