feat: add opt-in evidence-driven self-evolution loop to Phase 7#45
Open
epoko77-ai wants to merge 1 commit into
Open
feat: add opt-in evidence-driven self-evolution loop to Phase 7#45epoko77-ai wants to merge 1 commit into
epoko77-ai wants to merge 1 commit into
Conversation
Phase 7 evolution was feedback-driven only: a human had to start it, the basis was a subjective sentence, and there was no regression gate on edits. This adds an opt-in autonomous mode for harnesses that have a deterministic verifier and a splittable repeatable task set. - SKILL.md: split Phase 7 into two modes (feedback-driven default / evidence-driven autonomous opt-in); add 7-6 with an eligibility gate, the 3-stage loop (weakness mining -> bounded proposal -> validation), the conservative non-regression acceptance rule (Δ_in >= 0 AND Δ_ho >= 0 AND max > 0), and snapshot/rollback logging. - references/self-evolution-loop.md: eligibility classes (A/B/C), failure signature schema (c, q, m), full procedure, and a golden-sample fallback for subjective harnesses. - CHANGELOG.md: Unreleased entry. Adapts the Self-Harness paradigm (arXiv:2606.09498) to the factory's evolution phase. Backward-compatible: harnesses without a deterministic verifier keep using 7-1..7-4 unchanged.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Phase 7 (하네스 진화) is currently feedback-driven only — a human must start it, the basis is a subjective sentence, and there is no regression gate on edits. This PR adds an opt-in autonomous mode (7-6) for harnesses that already have a deterministic verifier and a splittable, repeatable task set: it mines failures from execution traces, proposes bounded surface-tied edits, and promotes them only through a conservative non-regression gate. It is fully backward-compatible — harnesses without a verifier keep using 7-1~7-4 unchanged.
Motivation
Scope of change
skills/harness/SKILL.md— Phase 7 split into two modes + new 7-6)skills/harness/references/self-evolution-loop.md— new)CHANGELOG.mdTests
CHANGELOG update
CHANGELOG.mdunder UnreleasedSemVer impact
feat:)Additional notes
Diff is ~182 lines (under the 400-line "discuss first" threshold in CONTRIBUTING.md). Happy to convert this into a Discussion/RFC first if you'd prefer to align on the 7-6 design before reviewing the prose. The bulk of detail lives in the conditionally-loaded reference file to respect the <500-line SKILL.md / Progressive Disclosure guidance.