Model card: BIRD dev EX re-scored with --meta_time_out 3600 (SQLite 3.45.3), Spider test EX, checkpoint provenance
Browse files
README.md
CHANGED
|
@@ -10,8 +10,14 @@ FINER-SQL 3B SQL-Writer continually trained with GRPO (format + execution reward
|
|
| 10 |
counterexample-enriched database environment of *Small Language Models for Text-to-SQL against Spurious Correctness*.
|
| 11 |
Checkpoint step 1000 (dev-selected). Code: the `core-db-sql` GitHub repository.
|
| 12 |
|
| 13 |
-
| BIRD dev EX | Spider dev EX | Spider dev TS |
|
| 14 |
-
|---:|---:|---:|
|
| 15 |
-
| 68.
|
| 16 |
|
| 17 |
-
k = 30 samples, T = 1.0, vote-after-verify selection, official evaluators
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
counterexample-enriched database environment of *Small Language Models for Text-to-SQL against Spurious Correctness*.
|
| 11 |
Checkpoint step 1000 (dev-selected). Code: the `core-db-sql` GitHub repository.
|
| 12 |
|
| 13 |
+
| BIRD dev EX | Spider dev EX | Spider dev TS | Spider test EX |
|
| 14 |
+
|---:|---:|---:|---:|
|
| 15 |
+
| 68.84 (1056/1534) | 86.75 (897/1034) | 81.53 (843/1034) | 87.66 (1882/2147) |
|
| 16 |
|
| 17 |
+
k = 30 samples, T = 1.0, seed 42, at most 2048 new tokens, vote-after-verify selection, official evaluators (BIRD:
|
| 18 |
+
evaluation_bird_ex.py with SQLite 3.45.3 and `--meta_time_out 3600`; Spider: taoyds/test-suite-sql-eval @e97acc54,
|
| 19 |
+
default flags; TS on the distilled Spider dev test suites). Prompts: `thanhdath/core-db-evaluation-prompts`
|
| 20 |
+
(`bird_dev`, `spider_dev`; Spider test: `spider_test_full_schema`).
|
| 21 |
+
|
| 22 |
+
Provenance: these weights are byte-identical to `thanhdath/3b-core-noam-ict` @ `0c74daf264c89d9e4bdf1378bcf065cb76b1ef0f`, sub-folder `checkpoints/step01000/checkpoint-1000` (training step 1000).
|
| 23 |
+
The saved predictions and the re-scoring script are in `scripts/rq2/` of the code repository.
|