---
license: apache-2.0
language:
- en
library_name: transformers
pipeline_tag: image-text-to-text
tags:
- multimodal
- mllm
- knowledge-distillation
- mobilevlm
- mobilellama
base_model: mtgv/MobileLLaMA-1.4B-Chat
---

# Cosine-Beta-KD-Task

A 1.7B multimodal LLM checkpoint distilled with **Cosine-KD + Beta-KD (Task-level uncertainty weighting)**,
built on top of MobileVLM with
[`MobileLLaMA-1.4B-Chat`](https://huggingface.co/mtgv/MobileLLaMA-1.4B-Chat) as
the language backbone.

This checkpoint corresponds to the **`Beta-KD (Task)`** row of the model
zoo in [Beta-KD: Uncertainty-Aware Knowledge Distillation for Multimodal Large
Language Models](https://arxiv.org/abs/2603.21426).

## Model Details

| Item | Value |
|------|-------|
| Architecture | MobileVLM (CLIP visual encoder + LDP projector + MobileLLaMA LLM) |
| Language model | MobileLLaMA 1.4B |
| Distillation losses | Cosine-KD (logit alignment) + Beta-KD task-level uncertainty loss |
| Training step | `checkpoint-18000` |
| Total params | ~1.7B |
| Precision | fp16 |

## Evaluation

Evaluated on six standard multimodal benchmarks (no beam search, greedy
decoding to match the chat-demo behavior).

| Method | LLM | MME<sup>P</sup> | MME<sup>A</sup> | GQA | VQA<sup>T</sup> | POPE | MMB<sup>dev</sup> | SQA<sup>I</sup> | Avg. |
|--------|-----|------|------|------|------|------|------|------|------|
| Cosine-KD baseline | MobileLLaMA 1.4B | 1308.4 | 65.4 | 59.9 | 52.2 | 84.6 | 57.1 | 61.3 | 63.4 |
| **+ Beta-KD (Task)** *(this model)* | MobileLLaMA 1.4B | **1352.0** | **67.6** | 60.8 | 53.9 | 85.4 | 59.1 | 61.2 | 64.7 |
| + Beta-KD (Instance) | MobileLLaMA 1.4B | 1350.3 | 67.5 | **61.2** | **54.2** | **86.0** | **60.2** | **62.9** | **65.3** |

## Usage

```python
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

repo_id = "jsun39/Cosine-Beta-KD-Task"

tokenizer = AutoTokenizer.from_pretrained(repo_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    repo_id,
    torch_dtype=torch.float16,
    trust_remote_code=True,
).cuda()
```

For full inference (image + text), please follow the inference example in the
[Beta-KD repo](https://github.com/Jingchensun/beta-kd) — the visual encoder /
projector loading, image preprocessing, and chat template are described there.

## Files

This repo contains only the files needed for inference:

- `pytorch_model.bin` — fp16 weights
- `config.json`, `generation_config.json`
- `tokenizer.model`, `tokenizer_config.json`, `special_tokens_map.json`

DeepSpeed optimizer / RNG / trainer states are intentionally **not** uploaded.

## Citation

```bibtex
@article{sun2026betakd,
  title   = {Beta-KD: Uncertainty-Aware Knowledge Distillation for Multimodal
             Large Language Models},
  author  = {Sun, Jingchen and Han, Shaobo and Patel, Deep and Kohno, Wataru and Jin, Can and Chen, Changyou},
  journal = {CVPR},
  year    = {2026}
}
```

## License

Released under the Apache-2.0 license, inheriting from MobileVLM and
MobileLLaMA. The visual encoder and any third-party data follow their original
licenses.