AgentCompass Research

Explore research papers from the AgentCompass team, covering agent evaluation infrastructure, benchmarks, and analysis. Each paper page explains its contribution, its relationship to AgentCompass, and where to find the paper, code, data, and relevant documentation.

For an introduction to the framework, start with the AgentCompass paper.

2026

Papers are listed by first submission date, newest first.

First submitted	Paper	Acceptance	Focus
2026-09-08	SWE-Bench Pro Verified: A Reliable Benchmark for Software Engineering Agents	—	Coding agent evaluation quality and reliability
2026-07-15	AgentCompass: A Unified Evaluation Infrastructure for Agent Capabilities	EMNLP 2026 Demo Track	Agent evaluation infrastructure