Empathetic Response Generation research benchmarks

Benchmarks

Dataset NameSOTA methodMetricTrendResultsLast Updated

AvaMERG (test)

ES4R

0.285BLEU-1

Jun 12, 2026

EmpatheticDialogues (test)

Blender

0.54Coverage

Feb 26, 2026

ERG

Ola-Omni-7B

98.9Distinct-1 Score

Mar 4, 2026

MSP-PODCAST

RE-LLM (w/o dim-emo. aux)

1.889Emotional Reaction Score

Feb 26, 2026

ESD

RE-LLM (w/o dim-emo. aux)

1.851Emotional Reaction

Feb 26, 2026

IEMOCAP

RE-LLM (w/o dim-emo. aux)

1.856Emotional Reaction

Feb 26, 2026

HumDial Challenge Track 1 Task 3-en (dev)

Our Model

4.36LLM Score (0-5)

Feb 26, 2026

HumDial Challenge Track 1 Task 3-zh (dev)

Stepaudio2-mini

4.53LLM Score (0-5)

Feb 26, 2026

ReddiSupp User Study (test)

MindTailor

3.98Empathy

Jun 23, 2026

ReddiSupp (Expert Evaluation)

MindTailor

4.2Empathy

Jun 23, 2026

MELD

ES4R

9.9ROUGE-L

Feb 26, 2026

Empathetic Response Generation (ERG)

Vanilla CoT

0.006BLEU

Feb 26, 2026

EQ-Bench 3

—

—Primary metric

Feb 26, 2026