Datasets:
|
Download code/question_eval_architecture.md from wenjieshu/solveedit_benchmark_2728: direct link, hf CLI and curl.
- Browser
- Download file 3.66 kB
-
https://huggingface.co/datasets/wenjieshu/solveedit_benchmark_2728/resolve/main/code/question_eval_architecture.md
- Command line
-
hf download hf://datasets/wenjieshu/solveedit_benchmark_2728/code/question_eval_architecture.md
-
curl -L -o question_eval_architecture.md https://huggingface.co/datasets/wenjieshu/solveedit_benchmark_2728/resolve/main/code/question_eval_architecture.md
3.66 kB
Question-first evaluation v2
The formal evaluation unit is an audited atomic question, not a checker name.
QuestionSpec -> Evidence DAG -> EvidencePacket -> AnswerRecord -> role-aware score
Invariants
- The original question, pass/partial/fail rubric, role, and weight are frozen in
QuestionSpec. - A free-form
decision_rule.typeis never executable by resemblance. Only a registered normalizer may emit a deterministicPredicateSpec. - Every tool dependency is enforced by the DAG executor. A required failed dependency blocks downstream calls.
- Detector evidence is scoped by case, check, image role, and resolved ROI.
- An explicit detector
no_detectionsresult may establish zero; a missing row may not. - A detector bbox synthesized as a mask is insufficient for formal geometry.
- Deterministic answers require a successful packet, a unique numeric band, and citable call IDs.
- Penalty answers retain positive rubric-satisfaction semantics. Inversion happens once in the final scorer.
- The independent image validity/quality gate is applied after question scoring; a failed gate forces the final case score to zero.
- VLM fallback receives the exact question/rubric, compact tool facts, input/prediction, and question-specific overlays. It does not receive reference GT.
- API keys are read from an environment variable at call time and are not CLI values or audit fields.
Current migration audit
The 271-case reviewed contract set compiles to 2,198 unique question specs:
- 7 deterministic questions with explicitly normalized semantics;
- 2,191
vlm_requiredquestions; - 1,825 distinct source rule names preserved for migration analysis;
- no missing resolved case and no duplicate procedure hash.
The deliberately small deterministic whitelist currently covers:
- forbidden/non-exempt instance counts;
- source-normalized zone clearance;
- assigned vehicle slot containment with minimum area and color-evidence gating.
The accessible-vehicle rule remains vlm_required because containment alone cannot verify the accessibility symbol.
Entry points
Compile and audit without running models:
PYTHONPATH=src python scripts/run_question_eval_v2.py compile \
--contracts-dir <contracts> --resolved-root <resolved-shards> --output-dir <output>
Replay legacy evidence without VLM calls:
PYTHONPATH=src python scripts/run_question_eval_v2.py replay \
--contracts-dir <contracts> --resolved-root <resolved-shards> \
--regions-tsv <regions> --detectors-tsv <fused-detections> --masks-tsv <masks> \
--tool-calls-tsv <tool-calls> --output-dir <output>
Add --run-vlm --vlm-model <model> --resume to answer unresolved questions. Questions are batched once per case while retaining separate packets and answer records. Use --num-shards and --shard-index for parallel workers.
Merge completed shards with conflict and provenance checks:
python scripts/merge_question_eval_v2_shards.py \
--shard-dir <shard-0> --shard-dir <shard-1> \
--question-specs <shard-0/question_specs.jsonl> --output-dir <merged>
Golden regressions
- Black spiked crystals: 0 = pass, 1–2 = partial, 3+ = fail.
- Cleared stairway: 0 output vehicles = pass; presence semantics cannot be reversed.
- Parking slot: containment
[0.65, 0.85)= partial, rather than multiplying0.85 * 0.5. - Required DAG dependencies block downstream tools.
- Bbox fallback cannot become a formal mask.
- Formal answers must cite tool calls.
- Runtime NumPy masks are persisted as compact summaries, not raw pixel arrays.
- Multiple unresolved questions in one case use one batch VLM request.