Benchmark Almanac

ServiceNow Research

We track 7 benchmarks that ServiceNow Research made or helped make.

0 in use 2 with public code 3 with a leaderboard here

Benchmarks

Name Kind Released Best score Status
UI-Vision Element Grounding Models locate interface elements from instructions. The score is the published overall grounding accuracy. Finding things on screen Mar 2025 8.27% Not in cards
AgentRewardBench Models judge whether web-agent trajectories completed their tasks. Their judgments are compared with expert labels across 1,302 trajectories from five benchmarks, using precision, recall, and F1. Web browsing Apr 2025 82% Not in cards
WorkArena++ Web agents complete 682 compositional knowledge-work tasks in ServiceNow, which need planning, reasoning and memory across many steps. Web browsing Jul 2024 1.90% Not in cards
UI-Vision Layout Grounding Models identify interface layout regions. The score is the published intersection over union percentage. Finding things on screen Mar 2025 24.3% Not in cards
UI-Vision Computer use agents are tested offline on human desktop demonstrations from 83 applications. Three tasks go from element grounding to layout grounding to next-action prediction for clicks, drags and key presses. Finding things on screen Mar 2025 Not in cards
UI-Vision Action Prediction Models predict the next desktop interface action. The score is the published step success rate. Finding things on screen Mar 2025 Not in cards
WorkArena Web agents do routine knowledge-work tasks in a live ServiceNow instance, such as filling forms and filtering lists. WorkArena++ adds 682 compositional tasks at levels 2 and 3. Web browsing Mar 2024 Not in cards