Claw-Eval-Live: A Live Agent Benchmark for Evolving Real-World Workflows

2026-04-30

AI Summary

Current benchmarks for large-model agents are static and fail to capture the dynamic nature of real-world tasks, while also lacking verifiability of intermediate reasoning processes. This work proposes the first dynamically updated, end-to-end evaluation framework that decouples a refreshable requirement signal layer from timestamped snapshots, enabling continuous assessment of workflow agents in authentic business and local environments. Introducing a novel “dual grounding” paradigm, the framework integrates external dynamic demands with verifiable execution traces, leveraging structured task families and a multidimensional scoring mechanism to move beyond reliance on final outputs alone. By fusing deterministic checks with structured LLM-based semantic evaluation over execution trajectories, audit logs, service state snapshots, and workspace artifacts, the benchmark evaluates 13 state-of-the-art models across 105 tasks, revealing that even the best model achieves only 66.7% success—highlighting HR, management, and multi-system coordination as critical bottlenecks and underscoring the insufficiency of leaderboard scores in fully characterizing agent capabilities.

Abstract

LLM agents are expected to complete end-to-end units of work across software tools, business services, and local workspaces. Yet many agent benchmarks freeze a curated task set at release time and grade mainly the final response, making it difficult to evaluate agents against evolving workflow demand or verify whether a task was executed. We introduce Claw-Eval-Live, a live benchmark for workflow agents that separates a refreshable signal layer, updated across releases from public workflow-demand signals, from a reproducible, time-stamped release snapshot. Each release is constructed from public workflow-demand signals, with ClawHub Top-500 skills used in the current release, and materialized as controlled tasks with fixed fixtures, services, workspaces, and graders. For grading, Claw-Eval-Live records execution traces, audit logs, service state, and post-run workspace artifacts, using deterministic checks when evidence is sufficient and structured LLM judging only for semantic dimensions. The release contains 105 tasks spanning controlled business services and local workspace repair, and evaluates 13 frontier models under a shared public pass rule. Experiments reveal that reliable workflow automation remains far from solved: the leading model passes only 66.7% of tasks and no model reaches 70%. Failures are structured by task family and execution surface, with HR, management, and multi-system business workflows as persistent bottlenecks and local workspace repair comparatively easier but unsaturated. Leaderboard rank alone is insufficient because models with similar pass rates can diverge in overall completion, and task-level discrimination concentrates in a middle band of tasks. Claw-Eval-Live suggests that workflow-agent evaluation should be grounded twice, in fresh external demand and in verifiable agent action.

Problem

• LLM agents
• workflow evaluation
• live benchmark
• task execution verification
• evolving workflows

Innovation

• live benchmark
• workflow agents
• execution trace
• deterministic evaluation
• refreshable signal layer