Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-08-30 · source-backed
MLReproMutate applies controlled mutations across four classes (random seed, dependency pin, data split, cross-validation fold count) and runs them against the validation workflows the repositories already ship. Of 39 frozen repository-operator cases, 24 were evaluable with 23 confirmed non-equivalent mutations, and existing workflows detected 2. An 8.7% detection proportion. (arXiv 2608.27100) The authors are careful this doesn't make those repos irreproducible, only that green CI tells you almost nothing about whether experimentally important choices are pinned.
Each link below shares sources, entities, or timing with this story.
Same source domain / Shared topic / Tension / Downstream implication
Reported by the same outlet (arxiv.org); overlapping topics (against, mutation); pushes against this story (against).
Same source domain / Shared topic / Tension
Reported by the same outlet (arxiv.org); overlapping topics (against, count, mutation); pushes against this story (against).
Reported by the same outlet (arxiv.org); overlapping topics (against, author, controlled); pushes against this story (but).
Reported by the same outlet (arxiv.org); overlapping topics (against, author, choic); pushes against this story (against).
Reported by the same outlet (arxiv.org); overlapping topics (against, already, author); pushes against this story (against).
Reported by the same outlet (arxiv.org); overlapping topics (against, already, apply); pushes against this story (against).
Reported by the same outlet (arxiv.org); overlapping topics (against, caught, class); pushes against this story (against).
Shared topic / Tension
Overlapping topics (against, apply, author, controlled); pushes against this story (against).