Improve model card

README.md

@@ -1,9 +1,10 @@
 ---
-license: apache-2.0
-language:
-- en
 base_model:
 - ByteDance-Seed/UI-TARS-1.5-7B
+language:
+- en
+library_name: transformers
+license: apache-2.0
 pipeline_tag: image-text-to-text
 tags:
 - gui-agent
@@ -15,24 +16,23 @@ tags:
 - robustness
 - reinforcement-learning
 - grpo
-library_name: transformers
 ---
 
 # AgentHijack-Agent
 
-**AgentHijack-Agent** is the action-generation model released with the paper
-[*AgentHijack: Benchmarking Computer Use Agent Robustness to Common Environment Corruptions*](https://AgentHijack.github.io) (ICML 2026).
+**AgentHijack-Agent** is the action-generation model released with the paper 
+[*AgentHijack: Benchmarking Computer Use Agent Robustness to Common Environment Corruptions*](https://huggingface.co/papers/2605.25707) (ICML 2026).
 
 It is fine-tuned from [`UI-TARS-1.5-7B`](https://huggingface.co/ByteDance-Seed/UI-TARS-1.5-7B) (Qwen2.5-VL architecture) using **Data-Augmented Group Relative Policy Optimization (DA-GRPO)** on the AgentHijack benchmark, with the goal of producing a computer-use agent that remains reliable under *common environment corruptions* (pop-ups, resolution changes, UI marks, subtitles, multi-apps, accidental touches, app minimization, network errors, and verification prompts).
 
 The same checkpoint serves a dual role in the AgentHijack-Agent framework:
-
 1. **Action generator** — produces the next GUI action from screenshots + history.
 2. **Onlooker** — summarizes behavioral changes between consecutive screenshots and performs an initial environment check before execution.
 
-- 📄 **Paper:** *AgentHijack: Benchmarking Computer Use Agent Robustness to Common Environment Corruptions* (ICML 2026)
-- 🌐 **Project page:** https://AgentHijack.github.io
-- 🧩 **Base model:** `ByteDance-Seed/UI-TARS-1.5-7B` (Qwen2.5-VL-7B architecture)
+- 📄 **Paper:** [AgentHijack: Benchmarking Computer Use Agent Robustness to Common Environment Corruptions](https://huggingface.co/papers/2605.25707)
+- 🌐 **Project page:** [https://AgentHijack.github.io](https://AgentHijack.github.io)
+- 💻 **Code:** [https://github.com/tmlr-group/AgentHijack](https://github.com/tmlr-group/AgentHijack)
+- 👥 **Authors:** Jingwei Sun, Jianing Zhu, Yuanyi Li, Tongliang Liu, Xia Hu, and Bo Han
 - 🏛️ **Affiliations:** TMLR Group, Hong Kong Baptist University
 
 ---
@@ -46,8 +46,6 @@ Compared with the base `UI-TARS-1.5-7B`, AgentHijack-Agent:
 - Recovers from **unexpected operations** (accidental touch, app minimization) via behavioral summarization.
 - Detects **environment errors** (network failure, login/verification prompts) up-front instead of looping on meaningless attempts.
 
-See Table 2 and Figure 8 of the paper for full results and qualitative trajectories.
-
 ---
 
 ## Model details
@@ -59,17 +57,14 @@ See Table 2 and Figure 8 of the paper for full results and qualitative trajector
 | Precision | `bfloat16` |
 | Context length | 128k tokens |
 | Image resolution | 1920 × 1080 (native, paper default) |
-| Sharding | 4 × `safetensors` shards |
 | Tokenizer | Inherited from UI-TARS-1.5-7B / Qwen2.5-VL |
 
 ### Training
 
-- **Algorithm:** Data-Augmented GRPO (DA-GRPO), an extension of GRPO that rolls out the same instruction across *different corrupted environments* drawn from a corruption set `C`, instead of a single clean environment.
+- **Algorithm:** Data-Augmented GRPO (DA-GRPO), an extension of GRPO that rolls out the same instruction across *different corrupted environments* drawn from a corruption set `C`.
 - **Framework:** [VERL](https://github.com/volcengine/verl).
-- **Data:** 128 tasks sampled from the AgentHijack benchmark (built on top of OSWorld with 9 configurable corruption types, 3,321 tasks total).
-- **Schedule:** 15 epochs.
-- **Reward:** `r = r_success + r_format`, with an experience-replay buffer (following ARPO) to mitigate sparse-reward batches.
-- **Optimization:** clip range [0.2, 0.3], KL loss disabled to encourage exploration.
+- **Data:** 128 tasks sampled from the AgentHijack benchmark (built on top of OSWorld).
+- **Reward:** `r = r_success + r_format`, with an experience-replay buffer to mitigate sparse-reward batches.
 
 ---
 
@@ -93,48 +88,23 @@ wait()
 finished(content='xxx')
 ```
 
-### Prompt template (action generator)
-
-```
-You are a GUI agent. You are given a task and your action history, with
-screenshots. You need to perform the next action to complete the task.
-
-## Output Format
-```
-Thought: ...
-Action: ...
-```
-
-## Action Space
-
-{action_space}
-
-## Note
-- Use {language} in `Thought` part.
-- Write a small plan and finally summarize your next action (with its target
-  element) in one sentence in `Thought` part.
-
-## User Instruction
-{instruction}
-```
-
 ### Minimal inference example
 
 ```python
 from transformers import AutoProcessor, AutoModelForImageTextToText
 import torch
 
-model_id = "<your-username>/AgentHijack-Agent"
+model_id = "TMLR-Group-HF/AgentHijack-Agent"
 processor = AutoProcessor.from_pretrained(model_id)
 model = AutoModelForImageTextToText.from_pretrained(
     model_id, torch_dtype=torch.bfloat16, device_map="auto"
 )
 
-# Build a chat with screenshot(s) + the action-generator prompt above,
+# Build a chat with screenshot(s) + the action-generator prompt,
 # then run model.generate(...) as usual.
 ```
 
-For the full agent framework (action generator + onlooker + environment checking), please refer to the code at [AgentHijack.github.io](https://AgentHijack.github.io).
+For the full agent framework (action generator + onlooker + environment checking), please refer to the code at [GitHub](https://github.com/tmlr-group/AgentHijack).
 
 ---
 
@@ -156,4 +126,4 @@ If you use this model or the AgentHijack benchmark, please cite:
 
 ## Acknowledgements
 
-This model is built on top of [UI-TARS-1.5-7B](https://huggingface.co/ByteDance-Seed/UI-TARS-1.5-7B) and the [Qwen2.5-VL](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct) family, with training infrastructure based on [VERL](https://github.com/volcengine/verl). The benchmark environment extends [OSWorld](https://os-world.github.io/).
+This model is built on top of [UI-TARS-1.5-7B](https://huggingface.co/ByteDance-Seed/UI-TARS-1.5-7B) and the [Qwen2.5-VL](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct) family, with training infrastructure based on [VERL](https://github.com/volcengine/verl). The benchmark environment extends [OSWorld](https://os-world.github.io/).