VisualWebArena benchmark demonstrates that state-of-the-art multimodal agents still exhibit significant limitations on visually grounded web tasks.
PREVIOUS ACTION: None example_assistant: Let’s think step-by-step
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks
VisualWebArena benchmark demonstrates that state-of-the-art multimodal agents still exhibit significant limitations on visually grounded web tasks.