Repository navigation
arc_compiler_runtime: state the analyze() return shape - #53
Draft
findalexli wants to merge 1 commit into
Draft
findalexli wants to merge 1 commit into
findalexli wants to merge 1 commit into
Conversation
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Task finding: #48.
What and why
The hidden type-system suites, 33.67 of 100 points, call
.filteron the return value ofanalyze(). Nothing the agent sees gives that shape, and the starter returns an object. This PR adds one paragraph toinstruction.md, which is in neither image:Old vs new judge
Both trees resolve to judge key
ce45e420e181and work keyef303afa4f05, and 190 of 190 files underenvironment/andjudge/have the same sha256. So old and new are one image, published asseededge/edgebench.judge.arc_compiler_runtime@sha256:2f50b2ab0644..., and one set of grades covers both.tools/judge_local.py, eight grades per row across four workers, two of them on one worker using the published image; no spread. Onlysemantic.jsdiffers.This is the value of the shape, not of the sentence: we didn't measure how much the text helps a new solver. Regrading an unchanged archive gives the same score; recovering the points needs a changed
analyze(). Three cases also check message text (0.51 of 100), which this leaves out.Other solvers: in the public rollouts at
huggingface.co/datasets/adafadaq/edgebench-opus-5-5-xhigh-rollouts, all 200 saved reports (156 Kimi K3, 44 Qwen 3.5) show 0 of 120 on the three adversarial type-system suites. We didn't check each one's cause.Checklist for task changes
versionintask.tomlis bumped.NOTES.md: the change and its score effect are under Changes; Source notes the instruction change. It isn't a listed Known issue.