chore(deps): bump numpy from 2.0.2 to 2.5.2 in /skills/remove-ai-marks/scripts - #10
Merged
guillaumemeyer merged 1 commit intoAug 13, 2026
Conversation
dependabot
Bot
added
dependencies
dependabot
Bot
added
dependencies
guillaumemeyer
previously approved these changes
Aug 13, 2026
guillaumemeyer
left a comment
guillaumemeyer
left a comment
Owner
There was a problem hiding this comment.
Approved via API (dependabot bump review).
Bumps [numpy](https://github.com/numpy/numpy) from 2.0.2 to 2.5.2. - [Release notes](https://github.com/numpy/numpy/releases) - [Changelog](https://github.com/numpy/numpy/blob/main/doc/RELEASE_WALKTHROUGH.rst) - [Commits](numpy/numpy@v2.0.2...v2.5.2) --- updated-dependencies: - dependency-name: numpy dependency-version: 2.5.2 dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com>
dependabot
Bot
force-pushed
the
dependabot/pip/skills/remove-ai-marks/scripts/numpy-2.5.2
branch
from
August 13, 2026 04:27
735f5b9 to
4829a79
Compare
dependabot
Bot
deleted the
dependabot/pip/skills/remove-ai-marks/scripts/numpy-2.5.2
branch
August 13, 2026 04:30
guillaumemeyer
added a commit
that referenced
this pull request
Aug 31, 2026
A no-op row keeps before_pos (so it is still in the clear-rate denominator) but has cleared=None, so it never counts as a clear. The help text said 'excluded from the clear-rate denominator', which was wrong; state that it never counts as a clear.
guillaumemeyer
added a commit
that referenced
this pull request
Aug 31, 2026
* feat(bench): SynthID recipe search + robust measurement Add a multi-objective recipe search to the SynthID-text benchmark that answers "what combination of strategies, and at what intensity, best clears the mark while preserving meaning and reading human?" while hardening the measurements the search depends on. rewrite_text.py: - No-op guard: a rewrite that changed fewer than --noop-lex-floor (0.05) bigrams is flagged noop in --json-stats with a warning, so a near-verbatim output can no longer be read as 0% clear (the misleading backtranslate row). - strength x intensity: --rewrite-level now modulates the named strength prompt instead of replacing it, so a (strength, intensity) step is expressible for recipe composition. bench_synthid_text.py: - Surface noop + robust_cleared per row, aggregate noop_n / robust_clear_rate, and add robust %, human, AUROC and noop columns to the report and CSV. - AUROC: population-level post-removal AUROC (rewritten-wm vs rewritten-plain) plus a baseline sanity AUROC, via a rank-based Mann-Whitney implementation. - Human-likeness: pluggable --human-backend (stylometry always on, optional lastde/binoculars offline detector, degrade gracefully). - Recipe mode (--mode recipe): parse/validate strength@intensity recipes, compose a recipe end-to-end, Phase 1 intensity sweep + Phase 2 per-weight beam search, dominance-based (weight-independent) Pareto frontier, and a "Best recipes" report section plus recipes.csv. Recommend a default recipe on that frontier; --layer-a-after defaults off. docs + Makefile: document the new measurements and recipe mode, add bench-full and bench-semantic targets. Tests: no-op guard (rewrite + bench), robust-clear accounting, AUROC, recipe parse/validate, Pareto dominance, compose ordering, human-likeness fallback, and recipe render/CSV. * fix(bench): address CodeRabbit review feedback and expand docstring coverage * build(bench): default rewrite backend to DeepSeek bench-full and bench-semantic referenced undefined REWRITE_* Make variables, so they expanded to a dangling '--rewrite-backend --rewrite-model ' command. Define them with DeepSeek defaults (openai-compatible / deepseek-v4-flash / api.deepseek.com + allow-remote), overridable per-invocation (e.g. REWRITE_BASE_URL=http://127.0.0.1:8000 REWRITE_ALLOW_REMOTE= for a local vLLM). * fix(bench): address remaining CodeRabbit review feedback - parse_float_grid: reject an empty intensity grid (SystemExit) so a misconfigured --intensity-grid fails fast instead of silently searching nothing (comment 6). - rewrite_text invocation: route the new --noop-lex-floor subprocess value through safe_arg() per the service/** path instruction (comment 13). - Makefile: run bench-full under $(MARKLLM_DIR)/.venv/bin/python (and install semantic deps with the same interpreter in bench-semantic) so a --require-semantic run actually finds sentence-transformers; fail clearly when REWRITE_MODEL/REWRITE_BACKEND are empty (comment 12). - docs: benchmark recipe example uses ~/MarkLLM/.venv/bin/python so --require-semantic resolves (comment 12). * fix(bench): correct the --noop-lex-floor help text (CodeRabbit #10) A no-op row keeps before_pos (so it is still in the clear-rate denominator) but has cleared=None, so it never counts as a clear. The help text said 'excluded from the clear-rate denominator', which was wrong; state that it never counts as a clear. --------- Co-authored-by: guillaumemeyer <guillaumemeyer@users.noreply.github.com> Co-authored-by: Poorvith M P <poorvith007@proton.me>
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.
Bumps numpy from 2.0.2 to 2.5.2.
Release notes
Sourced from numpy's releases.
... (truncated)
Changelog
Sourced from numpy's changelog.
... (truncated)
Commits
48feceeREL: Prepare for the NumPy 2.5.2 release (#32226)ecf599cMerge pull request #32221 from charris/backport-321513c7ac97Merge pull request #32220 from charris/backport-3220523b30f4BUG: avoid segfaults when legacy copyswap slot is not defined (#32151)4964ca8TYP:iscloseshape-typing fix for 2d array-likes (#32205)c37ed94MAINT: Skip limited_api tests on some platforms. (#32214)5cfd73bMerge pull request #32206 from charris/update-cibuildwheeld8262bcMAINT: Update cibuildwheel to v4.2.0988d94dMerge pull request #32158 from charris/backport-32133b2e4f97BUG: avoid possible stack overflow in arraydescr_dealloc (#32133)