Qwen2.5-3B-Instruct-alfworld

This model is a fine-tuned version of Qwen/Qwen2.5-3B-Instruct using the verl-agent reinforcement learning framework on the ALFWorld embodied AI environment.

Performance

Metric	Score
Test accuracy	0.9843

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "zhangdw/Qwen2.5-3B-Instruct-alfworld"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")

Training

Trained with verl-agent on the ALFWorld environment.