WebLINX: Real-World Website Navigation with Multi-Turn Dialogue

Xing Han Lù , Zdeněk Kasner , Siva Reddy

🏛 Institutions: Mila , McGill University , Institute of Formal and Applied Linguistics , Charles University , Facebook CIFAR AI Chair
📅 Date: February 8, 2024
📑 Publisher: ICML 2024
💻 Env: Web
🔑 Keywords: benchmark dataset WebLINX conversational web navigation HTML pruning

TLDR

WebLINX is a large-scale benchmark for conversational web navigation with 100K interactions drawn from 2,300 expert demonstrations across more than 150 real websites. The paper also studies a retrieval-inspired HTML-pruning model and shows that even finetuned multimodal models still struggle to generalize to unseen websites.

Open paper Report issue