Image-Text-to-Text
PEFT
Safetensors
English
Chinese
decision-model
judgment
classification
vision-language
jev
systemone
qwen3.5
lora
vev
Eval Results (legacy)
Instructions to use CountingSheep/vev-4b-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use CountingSheep/vev-4b-lora with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
Card: the second generation of the 4B model is CountingSheep/vev-4b-v2
Browse files
README.md
CHANGED
|
@@ -43,15 +43,14 @@ Code, API and full results: [Xiaooolong/vev](https://github.com/Xiaooolong/vev).
|
|
| 43 |
|
| 44 |
<img src="https://raw.githubusercontent.com/Xiaooolong/vev/main/docs/doom.webp" alt="vev-4b playing Doom in real time: the game on the left, the eight slice probabilities and the chosen action on the right.">
|
| 45 |
|
| 46 |
-
vev-4b
|
| 47 |
-
|
| 48 |
|
| 49 |
This repository holds the LoRA adapter of [CountingSheep/vev-4b](https://huggingface.co/CountingSheep/vev-4b) (130
|
| 50 |
MB). `vev serve` applies it to [Qwen/Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B) at load time, so it is the
|
| 51 |
-
smaller download if you already have the base model.
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
second adapter on the merged weights of the stage before. This card describes 0.1.0.
|
| 55 |
|
| 56 |
<img src="https://raw.githubusercontent.com/Xiaooolong/vev/main/docs/checkout.png" width="520" alt="A checkout page with a red banner: Payment failed: your card was declined.">
|
| 57 |
|
|
@@ -69,13 +68,13 @@ questions = {
|
|
| 69 |
# next {"retry": 0.983, "wait": 0.004, "nothing": 0.014}
|
| 70 |
```
|
| 71 |
|
| 72 |
-
|
| 73 |
|
| 74 |
## Use it
|
| 75 |
|
| 76 |
```bash
|
| 77 |
pip install git+https://github.com/Xiaooolong/vev typesafe-sdk
|
| 78 |
-
vev serve --model CountingSheep/vev-4b-lora
|
| 79 |
```
|
| 80 |
|
| 81 |
```python
|
|
@@ -96,7 +95,7 @@ print(resp.answers["urgent"].noul)
|
|
| 96 |
|
| 97 |
Accuracy on human-labelled data, by the kind of question:
|
| 98 |
|
| 99 |
-
| Kind of question | Example | vev-4b-lora
|
| 100 |
|---|---|---|
|
| 101 |
| Jev-style decisions on text | JevBench public subset | 0.766 |
|
| 102 |
| The same, on sources the model has not seen | kev transfer-v4 | 0.776 |
|
|
@@ -111,8 +110,8 @@ Accuracy on human-labelled data, by the kind of question:
|
|
| 111 |
The text rows are public benchmarks; the image rows are judgment sets converted from labelled public data. Jev is
|
| 112 |
more accurate on nimble (by 22 points), JevBench (10) and kev transfer-v4 (8); they tie on judgekit. Jev's API
|
| 113 |
takes text only, so the image rows have no Jev comparison. Check Vev on a labelled sample of your own questions
|
| 114 |
-
before relying on it.
|
| 115 |
-
|
| 116 |
|
| 117 |
## Limitations
|
| 118 |
|
|
@@ -166,7 +165,6 @@ All Vev models and data:
|
|
| 166 |
title = {Vev: Jev-like decision models that can also see images},
|
| 167 |
author = {Wang, Xiaolong},
|
| 168 |
year = {2026},
|
| 169 |
-
url = {https://github.com/Xiaooolong/vev}
|
| 170 |
-
version = {0.2.0}
|
| 171 |
}
|
| 172 |
```
|
|
|
|
| 43 |
|
| 44 |
<img src="https://raw.githubusercontent.com/Xiaooolong/vev/main/docs/doom.webp" alt="vev-4b playing Doom in real time: the game on the left, the eight slice probabilities and the chosen action on the right.">
|
| 45 |
|
| 46 |
+
vev-4b playing Doom in real time through a small harness: 8 yes/no questions per look, one request. Try vev-4b-v2
|
| 47 |
+
in the browser: [CountingSheep/vev](https://huggingface.co/spaces/CountingSheep/vev).
|
| 48 |
|
| 49 |
This repository holds the LoRA adapter of [CountingSheep/vev-4b](https://huggingface.co/CountingSheep/vev-4b) (130
|
| 50 |
MB). `vev serve` applies it to [Qwen/Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B) at load time, so it is the
|
| 51 |
+
smaller download if you already have the base model. The second generation,
|
| 52 |
+
[CountingSheep/vev-4b-v2](https://huggingface.co/CountingSheep/vev-4b-v2), has no adapter repository: its last
|
| 53 |
+
training stage trains a second adapter on merged weights, and only the merged result is published.
|
|
|
|
| 54 |
|
| 55 |
<img src="https://raw.githubusercontent.com/Xiaooolong/vev/main/docs/checkout.png" width="520" alt="A checkout page with a red banner: Payment failed: your card was declined.">
|
| 56 |
|
|
|
|
| 68 |
# next {"retry": 0.983, "wait": 0.004, "nothing": 0.014}
|
| 69 |
```
|
| 70 |
|
| 71 |
+
This is a research preview. The weights are licensed under CC BY-NC 4.0.
|
| 72 |
|
| 73 |
## Use it
|
| 74 |
|
| 75 |
```bash
|
| 76 |
pip install git+https://github.com/Xiaooolong/vev typesafe-sdk
|
| 77 |
+
vev serve --model CountingSheep/vev-4b-lora
|
| 78 |
```
|
| 79 |
|
| 80 |
```python
|
|
|
|
| 95 |
|
| 96 |
Accuracy on human-labelled data, by the kind of question:
|
| 97 |
|
| 98 |
+
| Kind of question | Example | vev-4b-lora |
|
| 99 |
|---|---|---|
|
| 100 |
| Jev-style decisions on text | JevBench public subset | 0.766 |
|
| 101 |
| The same, on sources the model has not seen | kev transfer-v4 | 0.776 |
|
|
|
|
| 110 |
The text rows are public benchmarks; the image rows are judgment sets converted from labelled public data. Jev is
|
| 111 |
more accurate on nimble (by 22 points), JevBench (10) and kev transfer-v4 (8); they tie on judgekit. Jev's API
|
| 112 |
takes text only, so the image rows have no Jev comparison. Check Vev on a labelled sample of your own questions
|
| 113 |
+
before relying on it. Per-set results are in `results/vev-4b/` of the [GitHub
|
| 114 |
+
repository](https://github.com/Xiaooolong/vev).
|
| 115 |
|
| 116 |
## Limitations
|
| 117 |
|
|
|
|
| 165 |
title = {Vev: Jev-like decision models that can also see images},
|
| 166 |
author = {Wang, Xiaolong},
|
| 167 |
year = {2026},
|
| 168 |
+
url = {https://github.com/Xiaooolong/vev}
|
|
|
|
| 169 |
}
|
| 170 |
```
|