---
license: mit
pretty_name: EC²-Bench (EC2-Bench)
language:
- en
- zh
size_categories:
- n<1K
task_categories:
- text-generation
tags:
- agents
- llm-agents
- benchmark
- edge-cloud-collaboration
- edge-ai
- on-device
- llm-routing
- privacy
- openclaw
configs:
- config_name: default
  data_files:
  - split: test
    path: data/tasks.jsonl
---

# EC²-Bench (EC2-Bench)

**Empowering Edge Agents: A Systematic Analysis of Edge-Cloud Collaboration in LLM Agent Execution**

[📄 Paper (arXiv 2610.17816)](https://arxiv.org/abs/2610.17816) · [💻 Code (GitHub)](https://github.com/OpenBMB/EC2-Bench) · [🌐 Project Page](https://passionate11.github.io/ec2-bench-project-page/)

EC2-Bench is a benchmark for edge-cloud collaboration in long-horizon LLM agents, jointly measuring task performance, resource usage and privacy exposure.

![EC²-Bench overview](https://raw.githubusercontent.com/OpenBMB/EC2-Bench/main/assets/overview.png)

Edge-side models keep data local and cost little, but are weak at demanding planning and tool use. Cloud-hosted frontier models are far more capable, but expensive, and they see everything they are sent. In long-horizon agent execution, cloud participation is no longer a one-shot routing decision per query: it is **interleaved throughout the trajectory**, so every cloud call shapes later steps, accumulates cost, and decides how much workspace information leaves the device.

## What is in this dataset

| Path | What it is |
| --- | --- |
| `data/tasks.jsonl` | Index of all **128 tasks** (shown in the viewer above): prompt, expected behavior, grading criteria, category, source benchmark, language, and the number of annotated sensitivity units |
| `workspace/ACE_Bench.tar.gz` | All task workspaces in one tarball (files, mock services, and privacy ground truth under `gt/sensitive_items.json`) |
| `workspace/ACE_Bench/` | The same workspaces, unpacked, for browsing |
| `Images/acebench-openclaw-v1.0.tar.gz` | Prebuilt Docker image (`docker save` tarball) bundling the OpenClaw harness and per-task mock services |

The executable task definitions (with inline `grade()` verifiers) and the evaluation code live in the [GitHub repo](https://github.com/OpenBMB/EC2-Bench). Task IDs keep the benchmark's earlier working name (`ACE_Bench`).

**At a glance**

- **128 executable digital-workspace tasks** in 6 categories (office & daily tasks, information search, safety & security, development & operations, data analysis, automation), in English and Chinese.
- **100 tasks with fine-grained privacy annotations**: 3,385 sensitivity units covering personal and organizational information.
- Tasks are adapted from Claw-Eval, WildClawBench, QwenClawBench, LiveClawBench, PinchBench and ClawBench, plus 7 newly curated tasks, and standardized to one OpenClaw runtime.

## Evaluation axes

- **Task performance**: weighted completion score (rule-based checks, environment-state auditing, rubric-based LLM judge) and Pass³.
- **Resource usage**: cloud input / cache-read / output tokens with USD cost, plus edge-side FLOPs.
- **Privacy exposure**: fraction of annotated sensitivity units that become visible to the cloud anywhere along the trajectory (lower is better).

Four collaboration strategies are studied: **Query Routing**, **Per-step Routing**, **Edge-controlled Delegation** and **Edge-controlled Advisory**, against edge-only and cloud-only baselines.

## Quick start

```bash
# task workspaces
hf download chengpingan/EC2-Bench workspace/ACE_Bench.tar.gz --repo-type dataset --local-dir .
tar -xzf workspace/ACE_Bench.tar.gz

# prebuilt Docker image (Hugging Face is not a Docker registry: download, then docker load)
hf download chengpingan/EC2-Bench Images/acebench-openclaw-v1.0.tar.gz --repo-type dataset --local-dir .
docker load -i Images/acebench-openclaw-v1.0.tar.gz
```

```python
from datasets import load_dataset
tasks = load_dataset("chengpingan/EC2-Bench", split="test")
```

See the [GitHub README](https://github.com/OpenBMB/EC2-Bench#quick-start) for running edge-only, cloud-only and collaborative agents and for grading.

## Key findings

1. **Collaboration works for long-horizon agents.** Partial cloud participation improves on edge-only execution at a fraction of cloud-only cost and exposure.
2. **Stronger edges close the performance gap, not the privacy gap.** As the edge model gets stronger, completion differences across strategies shrink, while privacy exposure stays strongly strategy-dependent.
3. **The best collaboration mode shifts with edge capability**, from external routing and direct cloud execution toward edge-controlled invocation and cloud advisory. Deciding when to invoke the cloud and following cloud guidance are the two edge-side abilities that matter.

Full results: [paper](https://arxiv.org/abs/2610.17816) and [leaderboard](https://github.com/OpenBMB/EC2-Bench#leaderboard).

## Citation

```bibtex
@article{huang2026ec2bench,
  title   = {Empowering Edge Agents: A Systematic Analysis of Edge-Cloud Collaboration in LLM Agent Execution},
  author  = {Huang, Pengcheng and Liu, Zhenghao and Ren, Shixin and Yu, Shi and Yao, Sijia and Yan, Yukun and Xiao, Chaojun and Sun, Maosong},
  journal = {arXiv preprint arXiv:2610.17816},
  year    = {2026}
}
```

## License

MIT, same as the [code repository](https://github.com/OpenBMB/EC2-Bench/blob/main/LICENSE). Adapted tasks are used under the licenses of their source benchmarks: Claw-Eval, WildClawBench, QwenClawBench, LiveClawBench and PinchBench are MIT-licensed, and ClawBench is Apache-2.0.