CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling

Yifan Zhang , Yutong Dai , Viraj Prabhu , Zhiyuan Hu , Ran Xu , Zeyuan Chen

🏛 Institutions: Salesforce AI Research
📅 Date: October 5, 2026
📑 Publisher: arXiv
💻 Env: Web
🔑 Keywords: CLIFT conformal prediction self-verification test-time scaling WebArena Mind2Web

TLDR

Presents CLIFT, a framework applying conformal prediction to provide calibrated step-level self-verification signals for web agents without relying on costly external LLM judges. Used for RL credit assignment and test-time trajectory search, CLIFT yields consistent success rate gains on WebArena and Mind2Web.

Open paper Report issue