Piccolo-4B is a decision model for tool review and context selection. This repository contains the complete Safetensors checkpoint.
hf download agentscope-ai/Piccolo-4B --local-dir Piccolo-4B
Use the included reference implementation for bidirectional attention and candidate scoring.
python -m pip install -r Piccolo-4B/requirements.txt
python Piccolo-4B/piccolo_inference.py --model Piccolo-4B --request request.json --device cuda:0
Use the request example to prepare request.json.
Use Piccolo locally for tool review and context selection with the QwenPaw setup guide.
| HarnessJevBench Hard ↑ | DataJevBench ↑ | JevBench Public ↑ | Decision Index Public ↑ |
|---|---|---|---|
| 62.03% | 70.60% | 87.01% | 58.00 |
Benchmark results are for the full checkpoint.
This model is also available on ModelScope.
model-00001-of-00002.safetensors 5.3 GBmodel-00002-of-00002.safetensors 3.1 GBmodel-frozen-auxiliary.safetensors 908 MB