Skip to content

Dataset · ByteDance-Seed

Autonomous agent training benchmark

ByteDance-Seed/EdgeBench

134 real-world tasks where agents learn in executable environments via multi-level feedback over 12+ hours.

What's inside

EdgeBench is a benchmark of 134 real-world tasks for evaluating how autonomous AI agents learn from their environment. Agents are placed in executable environments with realistic multi-level feedback and allowed to iterate for over 12 hours per task, tracking the full improvement trajectory. 51 tasks are publicly open; language is English.

What you can do with it

  • Evaluating long-horizon agentic systems
  • Comparing agent training strategies via feedback
  • Testing LLM agents in executable environments
Tasks
Text generation
Languages
English
Size
up to 1K rows
Downloads this month
5.9K
Likes
86

How to load it

from datasets import load_dataset

ds = load_dataset("ByteDance-Seed/EdgeBench")