Beyond Description: Cognitively Benchmarking Fine-Grained Action for Embodied Agents
2026.06 🎉 CFG-Bench has been accepted to ECCV 2026!
2026.06 🌟 We released CFG-Bench, a fine-grained cognitive benchmark for embodied agents.
Multimodal Large Language Models (MLLMs) show promising results as decision-making engines for embodied agents operating in complex, physical environments. However, existing benchmarks often prioritize… See the full description on the dataset page:
https://huggingface.co/datasets/CFG-Bench/CFG-Bench.
