Agents

Explore the browser agents supported by LexBench platform, view configuration and integration guide

Agent-TARS

v1.0

ByteDance·AvailableMultimodal

ByteDance's open-source multimodal browser agent supporting complex web interaction tasks

Key Capabilities

Multimodal understanding

Complex web interactions

Visual element recognition

Auto screenshot analysis

Supported Data

LexBench-BrowserOnline-Mind2WebBrowseComp

browser-use

v0.1

Community·AvailableAutomation

Lightweight browser automation agent, easy to integrate and extend

Key Capabilities

Lightweight design

Fast response

Easy to extend

Low resource usage

Supported Data

LexBench-BrowserOnline-Mind2Web

Skyvern

v1.0

Skyvern AI·AvailableAutomation

基于计算机视觉的浏览器自动化 Agent，对 UI 变化具有强鲁棒性

Key Capabilities

Computer vision-based

Robust to UI changes

Multi-step workflows

Enterprise-ready

Supported Data

LexBench-Browser

SeeClick

v1.0

Community·AvailableMultimodal

视觉-语言多模态 GUI 模型，支持屏幕理解和点击预测

Key Capabilities

Vision-language model

GUI grounding

Click prediction

Screen understanding

Supported Data

LexBench-Browser

How to Integrate New Agent

Follow these steps to integrate your agent into LexBench platform

Implement Interface

Implement the Agent interface specification defined by LexBench

Add Configuration

Add agent configuration file in the agents directory

Submit Evaluation

Run evaluation scripts and submit results

Quick Start

# Run Agent-TARS evaluation
python scripts/run.py \
  --agent Agent-TARS \
  --data LexBench-Browser \
  --split no_login \
  --mode first_n \
  --count 10

# Run browser-use evaluation
python scripts/run.py \
  --agent browser-use \
  --data LexBench-Browser \
  --split no_login \
  --mode first_n \
  --count 10

# Run Agent-TARS evaluation python scripts/run.py \ --agent Agent-TARS \ --data LexBench-Browser \ --split no_login \ --mode first_n \ --count 10 # Run browser-use evaluation python scripts/run.py \ --agent browser-use \ --data LexBench-Browser \ --split no_login \ --mode first_n \ --count 10