FoundryFoundry
HomeLeaderboardResearch

Research & Insights
from the Foundry Team

Deep dives into browser agent evaluation, benchmarking methodologies, and the latest developments in AI web automation.

SDRBench: A Simulated, Resettable RL Gym for Realistic Sales Development Representative Workflows
5 min read
September 22, 2025By Foundry Team

SDRBench: A Simulated, Resettable RL Gym for Realistic Sales Development Representative Workflows

SDRBench is a benchmark and RL gym for browser agents, designed to model realistic enterprise SaaS workflows across apps like Gmail, Sheets, LinkedIn-like prospecting, Apollo, and Salesforce. It provides 50 deterministic tasks with cross-application state persistence, reproducible evaluation via Foundry’s Agent Web Engine, and leaderboard-ready baselines, enabling research on generalization, credit assignment, and error recovery in complex browser-native environments.

Read full article
Foundry
Research•Contact Us•Privacy Policy
© 2025 Foundry•Backed by YC