summarization benchmark
QMSum
QMSum is a benchmark for query-based multi-domain meeting summarization consisting of 1,808 query-summary pairs over 232 meetings across academic, product, and committee domains. The dataset enables models to select and summarize relevant spans of meetings in response to specific queries. Published at NAACL 2021, QMSum presents significant challenges in long meeting summarization where models must identify and summarize relevant content based on user queries.
Updated Aug 7, 2026
Published models2
Registry coverage2
MetricScore
EvidenceB
QMSum leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | MI | 21.3% | 100.0% | 2 | C | |
| 2 | MI | 19.9% | 0.0% | 2 | C |
Score distribution
Top published rows on the benchmark's original scale.
QMSum highlights
The top published results on this benchmark's own scale.
What is QMSum?
Definition and scoring fields from the benchmark registry.
QMSum is a benchmark for query-based multi-domain meeting summarization consisting of 1,808 query-summary pairs over 232 meetings across academic, product, and committee domains. The dataset enables models to select and summarize relevant spans of meetings in response to specific queries. Published at NAACL 2021, QMSum presents significant challenges in long meeting summarization where models must identify and summarize relevant content based on user queries.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- QMSum
- Modality
- text
- Primary category
- summarization
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- qmsum|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about QMSum.
Which model scores highest on QMSum?
Phi-3.5-mini-instruct is currently ranked first with 21.3%.
What does QMSum measure?
QMSum is a benchmark for query-based multi-domain meeting summarization consisting of 1,808 query-summary pairs over 232 meetings across academic, product, and committee domains. The dataset enables models to select and summarize relevant spans of meetings in response to specific queries. Published at NAACL 2021, QMSum presents significant challenges in long meeting summarization where models must identify and summarize relevant content based on user queries.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
2 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.