chore(readme): remove MNIST/CIFAR benchmark-dataset framing#63
Merged
Conversation
Toy benchmarks in the README signal "amateur ML demo" rather than "production-grade research orchestration". The substantive story is the platform's capabilities — autonomous agent teams, oracle-driven gating, contract-first spawning, hard ZOTrainingCallback enforcement, experiment capture + autonomous loop — not which canned dataset we happened to validate against. Removed: - Demo accuracy badge (MNIST 99.66% / CIFAR-10 91.62%) — replaced test count badge with current 735 instead. - "Validated end-to-end with an MNIST digit classification project" preamble in E2E Validation section — rewritten as platform-capability framing with measured numbers tracked in cost-benchmark.mdx. - mnist-delivery/ file tree — replaced with a generic representative delivery-repo structure showing the .zo/ experiment capture layer, Docker scaffold, and three-tier test split. - Status table rows mentioning specific dataset accuracies — rewritten as "E2E validation on full ML-lifecycle reference projects" + "reference-project end-to-end demos". - v1 demos line linking to mnist-digit-classifier-delivery and cifar10-classifier-delivery — replaced with link to cost-benchmark.mdx. - "99% MNIST accuracy" in footer signature — replaced with "oracle-driven". - "zo draft -p cifar10 -d 'CIFAR-10 CNN, PyTorch, 90%+'" example — replaced with a tabular-churn-forecast example (more representative of the kind of project a research team would actually use ZO for). Also softened: "Measured on the MNIST bench" → "Measured on the canonical reference bench" in the --low-token paragraph. The substantive measurement evidence (cost, accuracy, wall-time) lives in docs/reference/cost-benchmark.mdx where it belongs — the README points readers there instead of repeating the toy-dataset numbers. validate-docs 10/10 (1 pre-existing test-count badge measurement-method warning, parametrized tests expand at runtime). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Deploying zero-operators with
|
| Latest commit: |
0ed495b
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://7f613921.zero-operators.pages.dev |
| Branch Preview URL: | https://claude-readme-remove-benchma.zero-operators.pages.dev |
SamPlvs
added a commit
that referenced
this pull request
Apr 30, 2026
…aming chore(readme): remove MNIST/CIFAR benchmark-dataset framing
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.
Summary
Toy benchmarks in the README signal "amateur ML demo" rather than "production-grade research orchestration". The substantive story is the platform's capabilities — autonomous agent teams, oracle-driven gating, contract-first spawning, hard `ZOTrainingCallback` enforcement, experiment capture + autonomous loop — not which canned dataset we happened to validate against.
Changes
MNIST 99.66% / CIFAR10 91.62%demo badge +tests-704_passing735_passing(current)Why this and not deleting all evidence
The substantive measurement evidence (cost, accuracy, wall-time) lives in `docs/reference/cost-benchmark.mdx` where it belongs — the README points readers there instead of repeating the toy-dataset numbers up front. The platform's claims are still backed by measured runs; the framing just doesn't lead with the dataset name.
Test plan
🤖 Generated with Claude Code