latent-lab/smaller-than-truth-bitnet-2b

lmprobe: Linear Probe on bitnet-b1.58-2B-4T

Truth probe for 'X is smaller than Y' statements. Near-perfect accuracy (99.0%) — structural/relational knowledge survives ternary quantization.

Classes

• 0: false_statement
• 1: true_statement

Usage

from lmprobe import LinearProbe

probe = LinearProbe.from_hub("latent-lab/smaller-than-truth-bitnet-2b", trust_classifier=True)
predictions = probe.predict(["your text here"])

Probe Details

• Base model: microsoft/bitnet-b1.58-2B-4T
• Model revision: 04c3b9ad9361b824064a1f25ea60a8be9599b127
• Layers: all (0–29, 30 layers)
• Pooling: last_token
• Classifier: logistic_regression
• Task: classification
• Random state: 42

Evaluation

Metric	Value
accuracy	0.9899
auroc	0.9995
f1	0.9899
precision	0.9899
recall	0.9899

Training Data

• Positive examples: 792

• Negative examples: 792

• Positive hash: sha256:55f43cad0c06e9599839603d3c513543de4ca71eed2b22cccb16542930f6e2b8

• Negative hash: sha256:d9f02b07f2025712a6dd175310f7af58ee761c73b40792e10cf06335953628c0

• Evaluation samples: 396

• Evaluation hash: sha256:546af552622d90b1e54d88d3111f68aa201178f0458267da51ecbe393443e821

Reproducibility

• lmprobe version: 0.5.8
• Python: 3.12.3
• PyTorch: 2.10.0+cu128
• scikit-learn: 1.8.0
• transformers: 5.3.0