Datasets:
Tasks:
Text Classification
Formats:
parquet
Languages:
English
Size:
1K - 10K
Tags:
structured-decisions
calibration
probabilistic-classification
system-one
workflow-evaluation
Synthetic
License:
Add d1-2B and d1-2B-OptiQ-4bit (zero-shot)
#54
by sandeep85 - opened
LiquidAI/d1-2B scored with Liquid's own torch runner (float32): 0.637. mlx-community/d1-2B-OptiQ-4bit (MLX, M3 Max): 0.641. Full 400-case test split, one request per case with all five questions, the benchmark's own metrics. Chart regenerated.
Merged.
codelion changed pull request status to merged