xiamoent/Agent-G2-webshop-7b

Agent-G2 WebShop 7B

Agent-G2 WebShop 7B is a WebShop-specialized language-agent checkpoint initialized from Qwen2.5-7B-Instruct and post-trained with Agent-G2: Gaussian Guidance for Agentic Reinforcement Learning.

Agent-G2 samples an expert-prefix depth for each task from an adaptive Gaussian distribution. The distribution is updated from rollout statistics already collected for policy optimization, without additional probe rollouts or a learned depth predictor.

Project Page · Code · Model Collection · Training Data

Important: This checkpoint is designed for research in the sandboxed WebShop simulator. It is not a general-purpose chat model or a system for autonomous real-world purchases.

Model Details

Item	Description
Base model	Qwen/Qwen2.5-7B-Instruct
Architecture	Qwen2ForCausalLM
Checkpoint format	BF16 Safetensors
Configured context length	32,768 tokens
Target environment	WebShop
Post-training	Agent-G2 with GRPO
Language	English
Required action format	<think>...</think><action>search[...]</action> or <think>...</think><action>click[...]</action>

Although the tokenizer metadata contains a larger generic maximum length, the model configuration declares 32,768 positions. The public 1.5B reference recipe uses substantially shorter prompts and responses.

Evaluation

The Agent-G2 project reports the following results for this 7B WebShop checkpoint:

Benchmark	Metric	Result
WebShop	Reward Score (0–100)	92.3
WebShop	Final-purchase Success	84.4%

Agent-G2 uses expert-prefix guidance as a training mechanism rather than an inference-time dependency. The project evaluates the learned policy without supplying an expert trajectory.

These results are reported by the Agent-G2 repository and have not been independently reproduced in this model card. Evaluation variance is not currently available. Results may vary with the WebShop product corpus, prompt template, action history, random seed, and decoding configuration.

Intended Use

This checkpoint is intended for:

• reproducing Agent-G2 results in the WebShop simulator;
• research on long-horizon language agents and agentic reinforcement learning;
• studying adaptive expert-prefix guidance;
• evaluating action selection over an environment-provided admissible action set.

For faithful evaluation, use the WebShop environment, prompt template, action parser, and rollout loop provided by the Agent-G2 repository. A standalone generation only demonstrates that the checkpoint loads successfully; it does not reproduce the interactive benchmark.

Environment Interface

At every environment step, provide the shopping goal, current observation, recent history, and admissible actions. The released parser expects English output containing both reasoning and exactly one action:

<think>Reason about the observation and admissible actions.</think>
<action>search[keywords]</action>

or

<think>Reason about the observation and admissible actions.</think>
<action>click[value]</action>

Missing tags, malformed actions, unsupported action types, or outputs containing Chinese characters are marked invalid by the released WebShop parser. Only actions from the current environment-provided admissible action set should be executed.

Quick Start

pip install -U transformers accelerate torch

The following example performs one WebShop-style generation step. Replace the placeholders with state supplied by the environment:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "xiamoent/Agent-G2-webshop-7b"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
)
model.eval()

task_description = "<shopping instruction>"
current_observation = "<current WebShop observation>"
available_actions = ["<admissible action 1>", "<admissible action 2>"]
actions_text = "\n".join(available_actions)

prompt = f"""
You are an expert autonomous agent operating in the WebShop e-commerce environment.
Your task is to: {task_description}.
Your current observation is: {current_observation}.
Your admissible actions of the current situation are:
[
{actions_text}
].

Now take one action for the current step. Enclose your reasoning within
<think> </think> tags, then present one admissible action within <action> </action>
tags.
""".strip()

inputs = tokenizer.apply_chat_template(
    [{"role": "user", "content": prompt}],
    tokenize=True,
    add_generation_prompt=True,
    return_dict=True,
    return_tensors="pt",
).to(model.device)

with torch.inference_mode():
    output_ids = model.generate(
        **inputs,
        max_new_tokens=512,
        do_sample=True,
        temperature=0.4,
        top_p=0.8,
        top_k=20,
        repetition_penalty=1.05,
    )

new_tokens = output_ids[0, inputs["input_ids"].shape[-1]:]
response = tokenizer.decode(new_tokens, skip_special_tokens=True)
print(response)

The released checkpoint's generation_config.json defaults to temperature 0.7. The example uses temperature 0.4 to match the project's evaluation setting.

Training

Agent-G2 uses expert WebShop trajectories as prefix guidance during training, followed by policy rollouts and GRPO updates. The guidance depth is sampled per task from a Gaussian distribution estimated online from existing rollout statistics. Prefix guidance is not required at inference time.

The associated data release includes 5,855 WebShop expert trajectories with unique IDs and action lengths from 3 to 10. See the Agent-G2 repository and training data for the public implementation and expert-prefix store.

The repository currently publishes paper-locked WebShop training commands for the 1.5B setup, but not a separate 7B hyperparameter file or the exact checkpoint-selection rule for this Hub artifact. This card therefore does not attribute the 1.5B-specific hyperparameters to the 7B checkpoint.

Limitations

• The model is specialized for the text-based WebShop simulator and may not generalize to other websites, interfaces, or product corpora.
• It can produce malformed or inadmissible actions; environment-side validation is required.
• Performance is sensitive to prompt formatting, observation history, decoding settings, random seed, and environment configuration.
• The reported evaluation does not include variance across repeated runs.
• The model may inherit factual errors, biases, and other limitations from the base model and training data.
• This checkpoint must not be used to make autonomous real-world purchases or other consequential transactions without strong safeguards and explicit human approval.

Citation

If you find this checkpoint useful, please cite Agent-G2:

@misc{wang2026agentg2gaussianguidanceagentic,
      title={Agent-G$^2$: Gaussian Guidance for Agentic Reinforcement Learning},
      author={Zixuan Wang and Yanrui Miao and Zhengxi Lu and Teng Pan and Yiwen Qiu and Hongxing Li and Peng Qiu and Ruiqing Zhang and Yongliang Shen},
      year={2026},
      eprint={2608.23318},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2608.23318},
}

The paper has been accepted to the EMNLP 2026 Main Conference. A public paper link will be added when available.

Acknowledgements

Agent-G2 builds on verl-agent, veRL, and WebShop.