18 days ago
-
docs: keep only JevBench Public results, drop architecture figure
-
Link training dataset in introduction
-
Summarize model highlights and compact evaluation table
Rename Task type and reduce table spacing while preserving all reported metrics. -
Update training example to load the released model
-
docs: add GitHub repository badge
-
Update README.md
-
Link released inference, training code and OpenJevData
-
Clarify accuracy and Brier definitions
-
Simplify model summary and remove mathematical notation
-
Add five-model comparisons on MMDM and JevBench Public
-
Document OpenJev architecture, SFT and REINFORCE-Analysis, and evaluation results
-
Upload OpenJev-4B model weights and loading configuration
-
initial commit
