179-mcq / README.md
weipang142857's picture
card: state training-target vs machine-gold roles (2026-07-17 convention pass)
1d8ebfa verified
|
Raw
History Blame Contribute Delete
3.97 kB
metadata
tags:
  - smart-manufacturing
  - sft
  - industrial
  - vision
  - anomaly-detection
license: other
pretty_name: 179-mcq
extra_gated_fields:
  Name: text
  Affiliation: text
  Intended use: text
extra_gated_prompt: >-
  This dataset is released for **research use**. Access is reviewed and granted
  **manually** by the maintainers. Please state your name, affiliation, and
  intended use.

179-mcq

Mask-grounded multiple choice (Set-of-Mark style) for aero-engine-blade defect localization — 1137 items, derived deterministically from the binary segmentation masks of AI4Manufacturing/179. Exact-match gradable → SFT and RLVR-ready.

Task

One item per eligible defective record. The image is a 2×2 grid of views A–D of the same blade photo; each view overlays one red candidate region mask. Exactly one view overlays the true defect mask — in both location and extent (the query says so, naming the defect type to locate). annot is the correct letter. The three negatives per item are hard by construction: shift (translated), fliplr/flipud/rot180 (mirrored), dilate (over-grown ≥2.5×), erode (shrunk). Every negative is guaranteed wrong (IoU vs truth < 0.35 except dilate, wrong by extent); panels are mutually distinct (pairwise IoU < 0.7); negative kinds are assigned to slots by an independent salted hash. Gold letters: A 294 / B 262 / C 311 / D 270. Per-type coverage: ablation 169, breakdown 329, fracture 389, groove 250.

Records are skipped (confidence over coverage) when the gold mask is under ~30 visible px after panel downscale, fewer than 3 sound negatives are constructible, or the mask covers >35% of the frame. Skipped defects remain fully covered by 179-grounding / 179-region.

field type meaning
query str 16 variants; names the defect type to locate; "answer with the letter only"
image Image 2×2 composite, panels A–D
annot str A / B / C / D
reasoning null none — deterministic derivation
cate / task str B / T-B2
metadata str (JSON) source, image_sha256, defect_type, gold_letter, panel_tags, area_pct

Roles

Roles: this is an answer-only tier — there is no reasoning column; annot is both the machine-parseable gold AND the direct-answer SFT target ('SFT-ready' here means direct imitation of annot in the query-specified format); it is also the exact-match/IoU reward key for RLVR.

Provenance

Built deterministically (no LLM/teacher; reasoning is null) from AI4Manufacturing/179 — AeBAD (Aero-engine Blade Anomaly Detection, AeBAD_S subset; Zhang et al., "Industrial Anomaly Detection with Domain Shift"): 2,160 aero-engine-blade surface photos, 4 defect types (ablation, breakdown, fracture, groove) + good, each anomalous image with a paired binary pixel segmentation mask (binarized here at gray>40, which reproduces the source defect_area_fraction exactly). Generator: annotate/179/build_179_derived.py in forge_model; machine gates: annotate/179/verify_179.py (all green at build time).

Resolution. Source photos are 3024×3024. Every image here is downscaled to a 1024 long side (LANCZOS; masks NEAREST) and all coordinates are in that pixel space — see metadata.image_wh. This matches common VLM input sizes and keeps the repo compact; a native-resolution rebuild is a deterministic option (DOWNSCALE=None).

Query diversity. The query field is drawn from a fixed pool of surface variants for this task (paraphrases preserving the task and answer format), selected by an independent per-record hash. A machine gate checks that no template correlates with the gold (worst z-scores reported above).

The repository name is an internal task code (the source dataset's code is 179).