Downloads · 30 days
0
0% of all-time downloads
SaifPunjwani/mroot-q17-scoreband-checkpoints-v1
mroot-q17-scoreband-checkpoints-v1 is a text generation model from SaifPunjwani. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as apache-2.0.
Use one of the immutable, root-loadable release tags below. The repository's main branch contains staging history and is not the model revision to load.
Downloads · 30 days
0
0% of all-time downloads
All-time downloads
9
Public
Repo size
20.7 GB
Likes
0
Public
Click a slice to open those files.
.safetensors20.6 GB · 100%
From the Hugging Face model README
Use one of the immutable, root-loadable release tags below. The repository's
main branch contains staging history and is not the model revision to load.
| Profile | Immutable tag | Five-benchmark hard mean |
|---|---|---|
| Lower | scoreband-lower-v2-hard5 | 49.8316 |
| Upper | scoreband-upper-v2-hard5 | 49.6403 |
Hard mean is the unweighted arithmetic mean of AIME24, AIME25, MATH500,
Minerva, and AIME26 benchmark-level avg@n values. AIME26 is included as a
first-class component. The earlier scoreband-upper-v1 tag is retained for
audit history but reports the obsolete four-benchmark mean.
These are standalone Qwen3-1.7B checkpoints. They require no adapter, router, teacher, retrieval system, or custom inference code.
These artifacts were score-selected for calibration. They were not trained with kNN novelty or RepExp and are not valid evidence for either method or for an unbiased algorithm comparison. Each release tag includes the complete measured evaluation, normality report, and provenance record.