HumDial-EIBench: Human-Recorded Multi-Turn Emotional Intelligence Benchmark

HumDial-EIBench is a benchmark built from authentic human-recorded dialogues sourced from the ICASSP 2026 HumDial Challenge. It is designed to evaluate whether audio language models truly understand emotion in speech, rather than relying on text transcription shortcuts. The dataset was created by ASLP-lab and was last updated on Hugging Face in May 2026.

Use Cases

• Benchmarking model performance on emotional understanding based on authentic human-recorded audio dialogues.
• Evaluating multi-turn conversational reasoning capabilities of audio language models.
• Training models to recognize and respond to emotional cues in speech, as described in the benchmark's four evaluation tasks.

Strengths

• Built from authentic human-recorded dialogues, likely providing realistic emotional data.
• Specifically designed to test for genuine emotional understanding beyond text transcription, as stated in the description.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.
• Freshness should be verified as the last update was 2026-05-28.

Provenance

Source

ASLP-lab via Hugging Face.

Collection Method

Authentic human-recorded dialogues from the ICASSP 2026 HumDial Challenge.

Freshness

Last updated 2026-05-28 09:41:20.