Text Classification
Transformers
Safetensors
English
qwen3_5
image-text-to-text
decision-model
system-one
calibration
typed-decisions
Instructions to use oraculumai/Manchego with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use oraculumai/Manchego with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="oraculumai/Manchego")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("oraculumai/Manchego") model = AutoModelForMultimodalLM.from_pretrained("oraculumai/Manchego", device_map="auto") - Notebooks
- Google Colab
- Kaggle
File size: 1,038 Bytes
f82e029 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 | {
"schema": "manchego-temperature-map/2",
"model": "Manchego v3 candidate v3m100b (release candidate rc1)",
"model_sha256": "2ee838433bfe278a226dc644667ad4a99ece82cc47325c7645a7dae723c1863b",
"contract": "semif",
"temperatures": {
"choice": 1.5,
"noul": 0.2,
"score": 1.0
},
"how_applied": "p = softmax(option-code logits / T[question type]); the chosen option is the largest logit, so no temperature changes it",
"fitted_on": "TMAP-V15 (research repository amendment A-V4-20, registered in commit 08c67a5 before the fit; results commit a18545d): v3m100b's SemIf records on 28,182 development rows the model never trained on (choice 10,485, noul 16,050, score 1,647); no benchmark item",
"rule": "per type, the temperature from the grid 0.2..2.0 that maximises the estimated JevBench v1.5 composite (headline A: noul answers with 0.20 < P(yes) < 0.80 count as wrong; typed calibration; equal types); ties to the temperature closest to 1",
"fit_sha256": "065b2ea824ac9cbcaff9ca0c1690fb98168e81568ca21ecd0b0e9a9798a1c9e2"
}
|