Clarify version comparison: v3.0 strict vs v3.1 normalized scoring
Browse files
README.md
CHANGED
|
@@ -41,11 +41,13 @@ A 360M parameter model fine-tuned for high-accuracy JSON extraction. v1.2 introd
|
|
| 41 |
|
| 42 |
### Version Comparison
|
| 43 |
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
|
| 47 |
-
|
| 48 |
-
|
|
|
|
|
|
|
|
| 49 |
|
| 50 |
### vs Baselines
|
| 51 |
|
|
|
|
| 41 |
|
| 42 |
### Version Comparison
|
| 43 |
|
| 44 |
+
> **Note:** v1.1+ use EdgeJSON v3.1 normalized scoring (case-insensitive keys). This is stricter on simple/medium schemas but fairer overall. v1.2 is the first sub-400M model to solve complex schemas under the standardized benchmark.
|
| 45 |
+
|
| 46 |
+
| Version | EdgeJSON Eval | Overall (JSONExact) | Complex | Notes |
|
| 47 |
+
|---------|---------------|---------------------|---------|-------|
|
| 48 |
+
| v1.0 | v3.0 (strict) | 55.1% | 4.0% | Original release |
|
| 49 |
+
| v1.1 | v3.1 (normalized) | **60.1%** | 0.0% | Best simple/medium under fair scoring |
|
| 50 |
+
| **v1.2** | v3.1 (normalized) | 58.9% | **4.0%** | **Complex breakthrough under fair scoring** |
|
| 51 |
|
| 52 |
### vs Baselines
|
| 53 |
|