Submission: zubalr - #78
Open
zubalr wants to merge 5 commits into
Open
Conversation
Offline document extraction and adjudication pipeline. The Docker image scores 137.83 / 150 on the public training set under the official evaluator, with extraction 45.31, classification 74.45, calibration 18.07, and 12 catastrophic false approvals. That figure is in-sample and is reported as a reproducibility result. Solution repository: https://github.com/zubalr/mib-intake
Regenerate all 5,000 validation predictions from the frozen image and refresh the memo with the measured validation runtime and the run-to-run reproducibility check.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
MIB Doc Challenge Submission
Updated submission, filed as a new pull request per the guidance that revised
entries are submitted this way. This supersedes #36 (same author), which I will
close.
Links
Dockerfile): https://github.com/zubalr/mib-intakeResult
137.89 / 150 on the complete public training set under the official
evaluator, produced by running the Docker image under the submission
constraints rather than from a cached intermediate.
That figure is in-sample. The model is fitted on these packets, so it measures
reproducibility rather than performance on unseen ones, and
MEMO.mdreportsthe held-out diagnostic together with a note on what it does not cover.
The 5,000 validation predictions come from the same image, in 4h07m at 2.96
seconds per PDF against the 6 second budget, and two independent runs of it
produced byte-identical output. Hidden text is never treated as evidence: the
classifier receives no case ids, filenames, or planted content.
Checklist
submissions/zubalr/predictions.jsonl,MEMO.md, andSUBMISSION.mdpredictions.jsonlpassesscripts/validate_submission.pyagainstdata/validation_manifest.csvDockerfile--network none) and accepts<input_pdf_dir> <output_predictions_path>DOCKER_SUBMISSION.md