Tencent WorkBuddy Bench: A Multi-Domain Coding-Agent Benchmark with Contamination-Resistant Task Construction

2026-07-23

AI Summary

This work addresses the susceptibility of existing code agent evaluation benchmarks to data contamination and their lack of comprehensive coverage across real-world, multi-domain workflows. To overcome these limitations, we introduce the first open-source, auditable, and end-to-end reproducible benchmark for multi-domain code agents, encompassing four key scenarios: coding, web development, office automation, and security. All tasks are reverse-engineered from authentic developer commits, pull requests, or business operations and presented as natural-language role-based requests to prevent reconstruction via web search. Contamination resistance is achieved through public task disclosure rather than secrecy, supported by containerized execution environments, a unified task schema, a dual-agent evaluation framework (CodeBuddy Code and Claude Code), and domain-specific validation mechanisms. The benchmark enables heterogeneous scoring and cross-model comparison, and includes a multidimensional leaderboard evaluating major model families.

Abstract

We introduce Tencent WorkBuddy Bench, a multi-domain evaluation suite for coding agents; this report documents its construction methodology, scoring protocol, and a cross-model leaderboard. At its core is a unified evaluation framework for constructing and running distribution-informed coding-agent tasks across four work domains - Code, Web, Office, and Security. Rather than adapting public issue text, every task is reverse-engineered from a real commit, pull request, or business scenario and rewritten as a short, colloquial, role-played request, so that a task's prompt is not recoverable by web-searching the underlying issue, pull request, or commit thread. Because the dataset is released openly - task directories, environment images, evaluation harness, tests, and reference solutions - contamination resistance rests on this construction together with dataset versioning rather than on secrecy. The four subsets - repository-level engineering, front-end development, office and business workflows, and red-/blue-team security - probe complementary facets of real work, each with its own verification style. All are packaged in a uniform task-directory format and run, under a uniform and reproducible protocol, on two agent harnesses (CodeBuddy Code and Claude Code); the full open release makes the benchmark reproducible end to end and directly auditable, since any third party can re-run each task and inspect its content. Because each subset uses a different scoring instrument, scores are not comparable across subsets and the suite reports no suite-wide average. We report a cross-model leaderboard across several model families.

Problem

• coding-agent benchmark
• contamination-resistant
• multi-domain evaluation
• task construction
• reproducible benchmark

Innovation

• contamination-resistant
• multi-domain benchmark
• coding agent evaluation
• reverse-engineered tasks
• reproducible benchmarking