You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
tool-output-consumption-rate — Tier 2, rank 12 (first Not started grader encountered walking the catalog tier-first: Tier 1 ranks 5-11 and Tier 2 ranks 1-4 were all already Implemented).
Why it's distinct from built-in graders
Built-in graders (tool-success-rate, retries, loops, trajectory-efficiency, execution-step-count, execution-duration, working-set-rebuild-factor, context-growth, artifact-production) measure success/failure, repetition, duration, and token growth. None of them measure whether a tool's output was ever actually used downstream. tool-output-consumption-rate fills that gap: it answers "of everything the agent read/fetched, how much of it ever fed a later decision or action?" — a distinct signal about wasted information-gathering work, separate from success rate or step count.
Required IR fields
observations[] (canonical Trajectory IR) — specifically each observation's consumedByActionIds array, populated per the IR build procedure (step 4 in trajectory-ir.md) by checking whether later actions reference values that only appear in that observation.
Formula
consumptionRate = observationsWithNonEmptyConsumedByActionIds / totalObservations, clamped to [0, 1], direction higher_is_better. Reports applicable: false/passed: null when the trace has no recorded observations.
Catalog status
Flipped Tier 2 rank 12 from Not started to Implemented in shared/graders/README.md. See full catalog — 13 of 25 graders now implemented.
Caution
Protected files were modified in this change.
This pull request is in
request_reviewmode and requires explicit human scrutiny before merge.Protected files:
README.md,.github/workflows/shared/graders/README.md,.github/workflows/shared/graders/tool-output-consumption-rate.mdGrader implemented
tool-output-consumption-rate— Tier 2, rank 12 (firstNot startedgrader encountered walking the catalog tier-first: Tier 1 ranks 5-11 and Tier 2 ranks 1-4 were all alreadyImplemented).Why it's distinct from built-in graders
Built-in graders (
tool-success-rate,retries,loops,trajectory-efficiency,execution-step-count,execution-duration,working-set-rebuild-factor,context-growth,artifact-production) measure success/failure, repetition, duration, and token growth. None of them measure whether a tool's output was ever actually used downstream.tool-output-consumption-ratefills that gap: it answers "of everything the agent read/fetched, how much of it ever fed a later decision or action?" — a distinct signal about wasted information-gathering work, separate from success rate or step count.Required IR fields
observations[](canonical Trajectory IR) — specifically each observation'sconsumedByActionIdsarray, populated per the IR build procedure (step 4 intrajectory-ir.md) by checking whether later actions reference values that only appear in that observation.Formula
consumptionRate = observationsWithNonEmptyConsumedByActionIds / totalObservations, clamped to[0, 1], directionhigher_is_better. Reportsapplicable: false/passed: nullwhen the trace has no recorded observations.Catalog status
Flipped Tier 2 rank 12 from
Not startedtoImplementedinshared/graders/README.md. See full catalog — 13 of 25 graders now implemented.Note
This was originally intended as a pull request, but the git push operation failed.
Original error: The process '/usr/bin/git' failed with exit code 1
Workflow Run: View run details and download bundle artifact
The bundle file is available in the
agentartifact in the workflow run linked above.Create the pull request manually