KDLAI

/

KDL-Frontier-Parser-nano-v2

Like 1

Follow

KoreaDeepLearning 26

Submitting one complete author-measured official ParseBench pass for KDLAI/KDL-Frontier-Parser-nano-v2, following the dataset submission instructions.

Model revision: 766505d578be5d99f2358ec9c95224023cd28d5f
Harness: 7de04fc11c6a692dfc5e1a86c7be817f3f0af8de
Dataset: 2805a1d940f95a203e0ae4b88be9934f7765b3fc
Overall: 78.959352
Coverage: 2078/2078 inference successes, 0 inference failures; all five evaluation groups completed with 0 failed cases.

Full results and reproducibility instructions: https://huggingface.co/KDLAI/KDL-Frontier-Parser-nano-v2/tree/5b5e0edda3c019563a2f6205ec49095288fafeb8/evaluation

This request does not claim maintainer reproduction or official leaderboard acceptance. It is a separate v2 model ID; the original v1 model and row stay unchanged.