---
license: apache-2.0
language: en
tags:
  - credence
  - concept-bottleneck-model
  - uncertainty-quantification
  - pytorch
  - text-classification
  - cebab
  - cebab
datasets:
  - cebab
pipeline_tag: text-classification
library_name: pytorch
arxiv: 2604.24170
---

# credence-phi-3-mini-4k-instruct-20251231-152729-cebab

CREDENCE checkpoint for **cebab** (cebab) with backbone `see checkpoint config` (5 concept heads).

Paper: [https://huggingface.co/papers/2604.24170](https://huggingface.co/papers/2604.24170) · arXiv: [2604.24170](https://arxiv.org/abs/2604.24170)

Training run folder: `phi-3-mini-4k-instruct_20251231_152729`

## Files

- `credence_checkpoint.pt` — PyTorch checkpoint (`model_state_dict`, `config`, `metadata`, optimizer state)

## Load with huggingface_hub

```python
from huggingface_hub import hf_hub_download
import torch

path = hf_hub_download(repo_id="tankiit/credence-phi-3-mini-4k-instruct-20251231-152729-cebab", filename="credence_checkpoint.pt")
ckpt = torch.load(path, map_location="cpu", weights_only=False)
state = ckpt["model_state_dict"]
config = ckpt["config"]
```

Load into your CREDENCE model implementation (see project `credence.py`).

## Citation

```bibtex
@article{mukherjee2026credence,
  title={Credal Concept Bottleneck Models for Epistemic--Aleatoric Uncertainty Decomposition},
  author={Mukherjee, et al.},
  journal={arXiv preprint arXiv:2604.24170},
  year={2026}
}
```