OmniGAIA

OmniGAIA: Omni-Modal General AI Assistant Benchmark 📄 Paper • 💻 Code & Demo • 🤗 Dataset & Model • 📈 Leaderboard OmniGAIA is a benchmark for Omni-Modal General AI Assistants that jointly reason over vision, audio, and language with external tools. It is designed to evaluate long-horizon, multi-hop, open-form problem solving in realistic settings rather than short perception-only QA. Benchmark Construction The OmniGAIA construction… See the full description on the dataset page:

Source: Hugging Face Hub (RUC-NLPIR/OmniGAIA). Metadata imported from the dataset’s Hub tags.