MM-OCEAN: A Multimodal Benchmark for Grounded Personality Reasoning

MM-OCEAN is a benchmark for evaluating Multimodal Large Language Models on Grounded Personality Reasoning. The benchmark contains 1,104 test videos sourced from the ChaLearn First Impressions V2 dataset. It was created by anonymous-mm-ocean and last updated on Hugging Face in May 2026.

Use Cases

• Benchmarking AI models on personality trait inference based on video evidence.
• Evaluating the grounding of personality judgments in observable behavior.
• Testing multimodal reasoning across tasks of increasing cognitive depth.

Strengths

• Contains 1,104 test videos for evaluation.
• Focuses on a specific, structured task: Grounded Personality Reasoning based on the Big Five model.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Video clips are sourced from the ChaLearn First Impressions V2 dataset.

Collection Method

Likely curated and annotated for the specific benchmark tasks.

Freshness

Last updated 2026-05-22 02:45:13; freshness should be verified.