library_name: transformers
pipeline_tag: text-generation
datasets:
- m-a-p/TerminalTraj
TerminalTraj-14B
This is the 14B model for the paper Large-Scale Terminal Agentic Trajectory Generation from Dockerized Environments.
TerminalTraj is a scalable pipeline designed to generate high-quality terminal trajectories that capture realistic long-horizon interactions across diverse domains. It addresses the challenges of executability and verifiability by (i) filtering high-quality repositories to construct Dockerized execution environments, (ii) generating Docker-aligned task instances, and (iii) synthesizing agent trajectories with executable validation code.
The model is based on the Qwen2.5-Coder backbone and demonstrates significant performance improvements on terminal-based agentic tasks (TerminalBench).
- GitHub Repository: multimodal-art-projection/TerminalTraj
- Paper: Large-Scale Terminal Agentic Trajectory Generation from Dockerized Environments
- Dataset: m-a-p/TerminalTraj
Usage
You can use this model with the transformers library:
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
model_id = "m-a-p/TerminalTraj-14B"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype=torch.float16, # 14B建议用fp16或bf16
device_map="auto" # 自动分配GPU
)
Citation
BibTeX:
@misc{wu2026largescaleterminalagentictrajectory,
title={Large-Scale Terminal Agentic Trajectory Generation from Dockerized Environments},
author={Siwei Wu and Yizhi Li and Yuyang Song and Wei Zhang and Yang Wang and Riza Batista-Navarro and Xian Yang and Mingjie Tang and Bryan Dai and Jian Yang and Chenghua Lin},
year={2026},
eprint={2602.01244},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2602.01244},
}