Update VEPQA to 2k examples per task

.gitattributes

@@ -58,3 +58,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+all_tasks/train-00000-of-00001.jsonl filter=lfs diff=lfs merge=lfs -text
+binary/train-00000-of-00001.jsonl filter=lfs diff=lfs merge=lfs -text
+clinvar_levels/train-00000-of-00001.jsonl filter=lfs diff=lfs merge=lfs -text
+pathogenic_pairwise/train-00000-of-00001.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -20,66 +20,199 @@ configs:
 
 # VEPQA
 
-Multi-config VEP SFT prototype built from raw ClinVar `variant_summary.txt.gz`
-and UCSC hg38 sequence windows.
+VEPQA is a small supervised fine-tuning dataset for variant-effect-prediction style question answering over DNA sequence context. It is built from raw ClinVar variant records and UCSC hg38 reference sequence windows, with deterministic gold labels from ClinVar. It does not use teacher-model rollouts.
 
-Configs:
+## Configs
 
-- `binary`: binary benign/likely benign vs pathogenic/likely pathogenic classification.
-- `clinvar_levels`: finer-grained ClinVar label classification over the retained benign/pathogenic levels.
-- `pathogenic_pairwise`: reference-vs-alternate sequence preference for pathogenic/likely pathogenic variants.
-- `all_tasks`: concatenation of the above configs for quick smoke tests.
+- `binary`: binary clinical-significance classification.
+- `clinvar_levels`: finer-grained ClinVar clinical-significance classification.
+- `pathogenic_pairwise`: reference-vs-alternate sequence preference for pathogenic or likely pathogenic variants.
+- `all_tasks`: concatenation of the three task configs.
 
-Row format:
+## Row Format
 
-- `question` / `prompt`: user-facing prompt.
-- `answer` / `gold_answer_json`: deterministic gold JSON target.
-- `messages`: two-turn SFT messages list.
-- `choices`: allowed answer labels for the task.
+Each row includes:
+
+- `question` / `prompt`: the user-facing prompt.
+- `answer` / `gold_answer_json`: deterministic JSON target.
+- `messages`: two-turn SFT messages, with the gold answer in the assistant message.
+- `choices`: allowed labels for the task.
 - audit metadata: ClinVar IDs, genomic position, review status, sequence hashes, and benchmark exclusion status.
 
-Counts:
+For evaluation, use `question` or `prompt` as model input and compare to `answer` or `gold_answer_json`. The `messages` column is intended for SFT and includes the answer.
+
+## Provenance
+
+Source variant table:
+
+- ClinVar `variant_summary.txt.gz`.
+
+Reference sequence:
+
+- UCSC hg38 per-chromosome FASTA files.
+
+Carbon eval exclusion manifest:
+
+- BRCA2 VEP: `HuggingFaceBio/brca2-vep`
+- TraitGym Mendelian VEP: `HuggingFaceBio/traitgym`, Mendelian VEP parquet
+- Carbon ClinVar VEP: `HuggingFaceBio/clinvar-vep-final`, coding and non-coding parquets
+
+The dataset was generated from raw ClinVar records, not from Carbon benchmark rows.
+
+## ClinVar Filtering
+
+Rows were retained from ClinVar only if they met all of the following criteria:
+
+- Assembly: `GRCh38`.
+- Variant type: single nucleotide variant.
+- Origin: germline.
+- Chromosome: autosomes or `X`.
+- Clean VCF alleles: one-base `A/C/G/T` reference and alternate alleles.
+- Clinical significance in one of:
+  - `Benign`
+  - `Likely benign`
+  - `Benign/Likely benign`
+  - `Likely pathogenic`
+  - `Pathogenic`
+  - `Pathogenic/Likely pathogenic`
+- Review status in the strict set:
+  - `practice guideline`
+  - `reviewed by expert panel`
+  - `criteria provided, multiple submitters, no conflicts`
+
+The larger `criteria provided, single submitter` class was intentionally not used.
+
+## Eval-Overlap Exclusion
+
+To avoid overlap with Carbon VEP evaluations, a manifest was built from the Carbon BRCA2, TraitGym Mendelian, and ClinVar VEP evaluation datasets. Candidate ClinVar SFT rows were excluded if either:
+
+- the exact variant key `(assembly, chrom, pos, ref, alt)` appeared in the Carbon eval manifest; or
+- the candidate position fell inside an expanded eval interval in the same assembly.
+
+The interval buffers used in the manifest were:
+
+- BRCA2: 8,192 bp buffer around each eval variant.
+- TraitGym Mendelian: 8,192 bp buffer around each eval variant.
+- Carbon ClinVar VEP: 24,000 bp buffer around each eval variant.
+
+Carbon's BRCA2 eval rows are represented in `hg19`, while this dataset is `GRCh38`. To avoid relying on liftover during this build, the dataset additionally hard-excludes a conservative GRCh38 BRCA2 locus interval:
+
+```text
+chr13:32,300,000-32,420,000
+```
+
+After these gates, the sampled dataset has:
+
+```text
+benchmark_exclusion_status: checked
+excluded_from_benchmark_eval: true
+```
+
+## Task Design
+
+### `binary`
+
+The prompt shows the reference allele, alternate allele, and a 1,024 bp DNA sequence window containing the alternate allele. The target is:
+
+```json
+{"clinical_significance":"benign_or_likely_benign"}
+```
+
+or:
+
+```json
+{"clinical_significance":"pathogenic_or_likely_pathogenic"}
+```
+
+This config collapses ClinVar benign/likely-benign labels and pathogenic/likely-pathogenic labels into two classes.
+
+### `clinvar_levels`
+
+This task uses the same sequence prompt style as `binary`, but preserves the retained raw ClinVar label:
+
+```json
+{"clinvar_level":"Pathogenic"}
+```
+
+Allowed values are:
+
+```text
+Benign
+Likely benign
+Benign/Likely benign
+Likely pathogenic
+Pathogenic/Likely pathogenic
+Pathogenic
+```
+
+### `pathogenic_pairwise`
+
+This task compares the reference and alternate sequence for a single ClinVar pathogenic or likely pathogenic SNV. Sequence A/B order is randomized. Because the alternate allele is labeled pathogenic or likely pathogenic, the reference sequence is treated as more consistent with normal biological function.
+
+This task is not a benign-vs-pathogenic matched-pair task. Benign or likely benign alternate alleles are not used for this reference-vs-alternate target, because a benign alternate allele is not necessarily less functional than the reference.
+
+## Data Distribution
+
+Config sizes:
+
+```json
+{
+  "binary": 2000,
+  "clinvar_levels": 2000,
+  "pathogenic_pairwise": 2000,
+  "all_tasks": 6000
+}
+```
+
+Clinical-classification raw-label distribution, used by both `binary` and `clinvar_levels`:
+
+```json
+{
+  "Benign": 333,
+  "Benign/Likely benign": 334,
+  "Likely benign": 333,
+  "Likely pathogenic": 333,
+  "Pathogenic": 334,
+  "Pathogenic/Likely pathogenic": 333
+}
+```
+
+`pathogenic_pairwise` raw-label distribution:
+
+```json
+{
+  "Likely pathogenic": 659,
+  "Pathogenic": 670,
+  "Pathogenic/Likely pathogenic": 671
+}
+```
+
+Review-status distribution over the 4,000 unique prompt rows:
 
 ```json
 {
-  "all_tasks": 180,
-  "binary": 80,
-  "clinvar_levels": 80,
-  "pathogenic_pairwise": 20
+  "criteria provided, multiple submitters, no conflicts": 3651,
+  "reviewed by expert panel": 348,
+  "practice guideline": 1
 }
 ```
 
-Label counts by task:
+Build-stage counts:
 
 ```json
 {
-  "binary": {
-    "Benign": 23,
-    "Benign/Likely benign": 8,
-    "Likely benign": 9,
-    "Likely pathogenic": 3,
-    "Pathogenic": 19,
-    "Pathogenic/Likely pathogenic": 18
-  },
-  "clinvar_levels": {
-    "Benign": 23,
-    "Benign/Likely benign": 8,
-    "Likely benign": 9,
-    "Likely pathogenic": 3,
-    "Pathogenic": 19,
-    "Pathogenic/Likely pathogenic": 18
-  },
-  "pathogenic_pairwise": {
-    "Likely pathogenic": 2,
-    "Pathogenic": 12,
-    "Pathogenic/Likely pathogenic": 6
-  }
+  "raw_rows_scanned": 8907730,
+  "strict_clinvar_candidates": 333140,
+  "after_carbon_eval_manifest_gate": 333140,
+  "excluded_by_grch38_brca2_locus": 3818,
+  "after_grch38_brca2_locus_exclusion": 329322,
+  "sampled_prompt_rows": 4000
 }
 ```
 
-Notes:
+## Limitations
 
-- The binary task uses pathogenic/likely pathogenic terminology, not malignant terminology.
-- Confidence/evidence strength is intentionally not a target because review-status metadata is not provided to the model in the prompt.
-- Carbon eval overlap was checked before sampling; BRCA2 hg19-vs-GRCh38 liftover remains a noted follow-up caveat in the project plan.
-- Existing Claude-generated SFT data and DataForge/Gemma rollouts are not used.
+- Labels are ClinVar clinical-significance labels, not direct experimental functional measurements.
+- `review_status` and submitter counts are included as metadata but are not shown in the prompt; the model is not trained to infer evidence strength.
+- The current release has only a `train` split. A larger training run should introduce region-disjoint validation data with an interval buffer.
+- The pairwise task is reference-vs-pathogenic-alternate only. A benign-vs-pathogenic matched-pair task would require separate matching logic.