Shorten the card; keep the previous card verbatim in HISTORY.md

HISTORY.md

@@ -0,0 +1,78 @@
+# Development history
+
+This is the previous model card, kept verbatim.
+
+---
+license: cc-by-4.0
+task_categories: [tabular-classification]
+tags: [time-series, industrial, 3w, causal, negative-results, anomaly-detection]
+configs:
+- config_name: decisions
+  default: true
+  data_files:
+  - split: validation
+    path: data/decisions/validation.parquet
+  - split: confirmation
+    path: data/decisions/confirmation.parquet
+- config_name: episodes
+  data_files:
+  - split: validation
+    path: data/episodes/validation.parquet
+  - split: confirmation
+    path: data/episodes/confirmation.parquet
+- config_name: corpus
+  data_files:
+  - split: metadata
+    path: data/corpus/metadata.parquet
+---
+# 3W causal real-group audit and failed-transfer traces
+
+Native typed Parquet derived from pinned public Petrobras 3W 2.0. This release
+contains source-audit metadata and every validation/confirmation decision and
+episode outcome for the frozen 11-candidate study. **It is not a new raw sensor
+corpus, new field acquisition or industrially qualified failure dataset.**
+
+The [source descriptor](https://doi.org/10.1038/s41597-026-07225-z) is published
+in April 2026; the underlying pinned release predates that paper. See
+[attribution](ATTRIBUTION.md), [exact table schemas/counts](TABLES.json),
+[protocol](benchmark-protocol.json), [manifest](benchmark-manifest.json),
+[access ledger](ACCESS-LEDGER.md), and [complete negative-result report](https://huggingface.co/sankalpsthakur/well-production-event-causal/blob/main/well_health/REPORT.md).
+The model repository is [well-production-event-causal](https://huggingface.co/sankalpsthakur/well-production-event-causal).
+
+## Configurations
+
+- `decisions`: one row per 30-second decision after 180-second past-only warmup;
+  labels, numeric eligibility, capped grid exposure, all 11 class/residual scores
+  and frozen alarms. Unknown class/binary targets and ineligible scores are null.
+  `event_timestamp_ns` is the original nominal dataset event clock, not verified
+  arrival or decision time. Scores/labels/provenance are offline audit fields;
+  they must not enter a predictor as input features.
+- `episodes`: one row per candidate/episode, with event, onset censoring, alarm
+  coverage, carried-in alarm and known-onset delay. Transient and steady labels
+  for the same contiguous event are merged. Covered is not onset detected.
+- `corpus`: 2,228 source recording metadata rows, including 1,119 real,
+  1,089 simulated and 20 drawn files, source hashes, labels, clocks, seven-channel
+  availability and the frozen real-recording selection/overlap exclusions.
+  Reserved-group rows are qualification metadata, not predictive evaluations.
+
+No train or reserved feature/score split is published here. The protocol used
+357 training, 33 validation, 138 initial confirmation and 113 reserved selected
+recordings, plus 478 real overlap exclusions. Raw sensor values are obtained
+from the separately attributed upstream source, not silently duplicated here.
+Filename groups are anonymized public IDs, not authenticated physical-well IDs.
+
+## Known failure and intended use
+
+The validation-selected primary covers 24/55 confirmation episodes, with 42
+false alarms over 9.124 eligible normal days and 64.392% normal alarm occupation.
+Only 3/12 observable onsets are covered within 300 seconds; most covered episodes
+already had an active alarm at onset. Normal numeric eligibility is 59.812%.
+Use these tables to inspect group transfer, abstention, onset censoring and alarm
+burden, not to claim physical barrier integrity, RUL or deployment readiness.
+
+All 31 frozen fixtures passed in the actual included CPU run. Its final JSON
+report failed serialization after traces/weights were saved; a separate report
+was recovered without fitting/retuning. The provider failure remains explicit.
+The original confirmation set is consumed; eight additional groups stay reserved
+for a separately reviewed/frozen future comparison. Public-pretraining exposure
+cannot be excluded by study-level group separation.

README.md

@@ -21,54 +21,48 @@ configs:
   - split: metadata
     path: data/corpus/metadata.parquet
 ---
-# 3W causal real-group audit and failed-transfer traces
 
-Native typed Parquet derived from pinned public Petrobras 3W 2.0. This release
-contains source-audit metadata and every validation/confirmation decision and
-episode outcome for the frozen 11-candidate study. **It is not a new raw sensor
-corpus, new field acquisition or industrially qualified failure dataset.**
+# 3W well-event audit tables (Petrobras 3W 2.0)
 
-The [source descriptor](https://doi.org/10.1038/s41597-026-07225-z) is published
-in April 2026; the underlying pinned release predates that paper. See
-[attribution](ATTRIBUTION.md), [exact table schemas/counts](TABLES.json),
-[protocol](benchmark-protocol.json), [manifest](benchmark-manifest.json),
-[access ledger](ACCESS-LEDGER.md), and [complete negative-result report](https://huggingface.co/sankalpsthakur/well-production-event-causal/blob/main/well_health/REPORT.md).
-The model repository is [well-production-event-causal](https://huggingface.co/sankalpsthakur/well-production-event-causal).
+Decision-level and episode-level results from a well-event detection study on real
+Petrobras 3W 2.0 recordings, tested on well groups the detectors never saw in training.
+Use these tables to see how detectors transfer across wells. This is not a new sensor
+dataset; raw sensor values stay with the upstream source.
+
+```python
+from datasets import load_dataset
+decisions = load_dataset("sankalpsthakur/3w-causal-real-group-audit", "decisions")
+```
 
 ## Configurations
 
-- `decisions`: one row per 30-second decision after 180-second past-only warmup;
-  labels, numeric eligibility, capped grid exposure, all 11 class/residual scores
-  and frozen alarms. Unknown class/binary targets and ineligible scores are null.
-  `event_timestamp_ns` is the original nominal dataset event clock, not verified
-  arrival or decision time. Scores/labels/provenance are offline audit fields;
-  they must not enter a predictor as input features.
-- `episodes`: one row per candidate/episode, with event, onset censoring, alarm
-  coverage, carried-in alarm and known-onset delay. Transient and steady labels
-  for the same contiguous event are merged. Covered is not onset detected.
-- `corpus`: 2,228 source recording metadata rows, including 1,119 real,
-  1,089 simulated and 20 drawn files, source hashes, labels, clocks, seven-channel
-  availability and the frozen real-recording selection/overlap exclusions.
-  Reserved-group rows are qualification metadata, not predictive evaluations.
+| Config | One row per | Splits |
+|---|---|---|
+| `decisions` (default) | 30-second decision after a 180-second warmup: labels, eligibility, all 11 candidates' scores and alarms | validation, confirmation |
+| `episodes` | Candidate and episode: coverage, carried-in alarm, onset delay | validation, confirmation |
+| `corpus` | Source recording (2,228 files: 1,119 real, 1,089 simulated, 20 drawn): hashes, labels, channel availability | metadata |
+
+- Scores, labels and provenance columns are for auditing. Never feed them to a predictor
+  as inputs.
+- `event_timestamp_ns` is the dataset's nominal clock, not arrival or decision time.
+- Filename groups are anonymized public IDs, not authenticated well identities.
+
+## What the tables show
 
-No train or reserved feature/score split is published here. The protocol used
-357 training, 33 validation, 138 initial confirmation and 113 reserved selected
-recordings, plus 478 real overlap exclusions. Raw sensor values are obtained
-from the separately attributed upstream source, not silently duplicated here.
-Filename groups are anonymized public IDs, not authenticated physical-well IDs.
+The validation-selected detector covered 24 of 55 confirmation episodes and caught 3 of
+12 observable onsets within 300 seconds. It raised 42 false alarms over 9.124 normal
+days and spent 64.392% of normal time in alarm. Use the tables to study transfer,
+abstention and alarm burden, not barrier integrity, remaining life or deployment
+readiness.
 
-## Known failure and intended use
+The study used 357 training, 33 validation, 138 confirmation and 113 reserved recordings,
+and excluded 478 overlapping real recordings. Models and the full report are in
+[well-production-event-causal](https://huggingface.co/sankalpsthakur/well-production-event-causal).
 
-The validation-selected primary covers 24/55 confirmation episodes, with 42
-false alarms over 9.124 eligible normal days and 64.392% normal alarm occupation.
-Only 3/12 observable onsets are covered within 300 seconds; most covered episodes
-already had an active alarm at onset. Normal numeric eligibility is 59.812%.
-Use these tables to inspect group transfer, abstention, onset censoring and alarm
-burden, not to claim physical barrier integrity, RUL or deployment readiness.
+## Source and attribution
 
-All 31 frozen fixtures passed in the actual included CPU run. Its final JSON
-report failed serialization after traces/weights were saved; a separate report
-was recovered without fitting/retuning. The provider failure remains explicit.
-The original confirmation set is consumed; eight additional groups stay reserved
-for a separately reviewed/frozen future comparison. Public-pretraining exposure
-cannot be excluded by study-level group separation.
+Derived from Petrobras 3W 2.0 under CC BY 4.0. The
+[source descriptor](https://doi.org/10.1038/s41597-026-07225-z) was published in April
+2026; the pinned release predates it. See [ATTRIBUTION.md](ATTRIBUTION.md),
+[table schemas](TABLES.json), [protocol](benchmark-protocol.json) and
+[manifest](benchmark-manifest.json). The previous card is in [HISTORY.md](HISTORY.md).