Skip to content

BenchmarkReport

reference
1 min readUpdated

Kind: Interface

Source: atloria-monorepo/libs/agent-core/src/benchmark/types.ts

Summary of a complete benchmark run.

Properties

PropertyType
suiteBenchmarkTask['suite']
modelstring
timestampstring
totalTasksnumber
passedTasksnumber
passRatenumber
totalDurationMsnumber
totalTokens{ input: number; output: number; total: number }
totalCostUsdnumber
avgCostPerTasknumber
byDifficultyRecord<string, { total: number; passed: number; rate: number }>
byLanguageRecord<string, { total: number; passed: number; rate: number }>
resultsTaskResult[]
toolStatsRecord<string, { count: number; avgDurationMs: number }>
failureBreakdownRecord<string, number>
byRepoRecord<string, { total: number; passed: number; rate: number; topFailure?: string }>
cumulativePassAtKRecord<number, { passed: number; rate: number }>
marginalGainsRecord<string, number>
passAtK{ k: number; passed: number; rate: number }
passAtKResultsPassAtKResult[]
costEfficiency`{
firstPassSuccess: { avgTokens: number; avgDurationMs: number; avgTurns: number; count: number };
recoveredSuccess: { avgTokens: number; avgDurationMs: number; avgTurns: number; count: number };
finalFailure: { avgTokens: number; avgDurationMs: number; avgTurns: number; count: number };

}| |fileLocalizationStats|{ avgRecall: number; avgPrecision: number; /** Tasks with >0 overlap that still failed / correctAreaButFailed: number; /* Tasks with 0 overlap */ zeroOverlap: number; }` |

Was this page helpful?

Download as PDF