This page outlines the current progress of Collective Memory against several WebArena benchmarks for autonomous tasks, passing through the gpt-4-1106-preview model from OpenAI.

Success rates may be artificially low as we begin to fill our benchmark records.