Tencent Hunyuan releases WorkForge for training work agents
The dual programmatic-plus-semantic verifier design addresses the hardest part of agent training: rewarding long-horizon tasks without human grading. Teams building enterprise work agents can reuse the recipe to generate their own verifiable task suites from internal workflows rather than hand-crafted benchmarks. It sits alongside agent-environment efforts like task synthesis pipelines from other labs but anchors everything in expert-derived, evidence-backed tasks.
Tencent Hunyuan introduced WorkForge, a framework that builds verifiable training environments for long-horizon work agents from real-world expert workflows. It constructs authentic workspaces, extracts checkable facts, and derives tasks with dual verifiers combining programmatic and semantic checks grounded in observable evidence. The team scaled it to 16.7K environments across 40 professional domains covering 60 file types. According to the announcement, agents show consistent gains on long-horizon work benchmarks as data volume and interaction horizon increase. A paper accompanies the release.
📢 Introducing WorkForge: a scalable framework that constructs verifiable training environments for long-horizon work agents from real-world resources.
Starting from expert workflows, it builds authentic workspaces, extracts checkable facts, and derives tasks plus dual (programmatic + semantic) verifiers grounded in observable evidence.
🚀 We scale to 16.7K environments across 40 professional domains covering 60 file types, enabling consistent gains on long-horizon work benchmarks as data volume and interaction horizon increase.
📚Paper: https://t.co/HjswwiHhlv @zhang_leo54389 @ZoeyyyZhououou @Be1ong1
#Tencent Hy#AIAgents #WorkAgents #AgentTraining
