ryan-0608/MoS-DFlash-Evidence
MoS-DFlash aggregate experiment evidence This dataset repository contains aggregate, reviewer-facing evidence for the MoS-DFlash experiments. It does not contain prompts, per-prompt generations, training data, credentials, internal paths, or raw training logs. B5 Qwen3-4B fixed-budget replication releases/b5-qwen3-4b-fixed-budget/ contains: the frozen result summary; the matched-training-volume aggregate trajectory; plot-ready aggregate trajectories; run… See the full description on the dataset page: https://huggingface.co/datasets/ryan-0608/MoS-DFlash-Evidence.
MoS-DFlash aggregate experiment evidence
This dataset repository contains aggregate, reviewer-facing evidence for the MoS-DFlash experiments. It does not contain prompts, per-prompt generations, training data, credentials, internal paths, or raw training logs.
B5 Qwen3-4B fixed-budget replication
releases/b5-qwen3-4b-fixed-budget/ contains:
- the frozen result summary;
- the matched-training-volume aggregate trajectory;
- plot-ready aggregate trajectories;
- run metadata; and
- an artifact manifest with hashes for the larger local evidence tree.
The files named by artifact_manifest.json are provenance records. Files that are not present in this repository remain unpublished; their names and hashes are included only to make the released aggregates traceable to the frozen local evidence snapshot.
The reported paired bootstrap intervals quantify prompt-level uncertainty for one fixed model pair. They do not establish training-seed robustness or a universal per-domain improvement.