Relation Extraction2017en

TAC Relation Extraction Dataset

Large supervised relation extraction dataset from TAC KBP, 106K examples covering 41 relation types. Standard sentence-level RE benchmark evaluated by Micro-F1 (excluding no_relation).

Paper / Website
Current State of the Art

LUKE

Studio Ousia

72.7

f1

TACRED — f1

3 results · 1 SOTA advances · higher is better

All results
SOTA frontier
7071727320262027f1LUKE

f1 Progress Over Time

Showing 3 breakthroughs from Jul 2019 to Oct 2020

70.671.271.872.372.9Jul 2019Feb 2020Oct 2020f1Date

Key Milestones

Jul 2019
SpanBERT

SpanBERT-large Micro-F1 on TACRED test. seed — verify

70.8
Source
Nov 2019
KEPLER

KEPLER Micro-F1 on TACRED test. seed — verify

71.7
+1.3%
Source
Oct 2020
LUKECurrent SOTA

LUKE Micro-F1 on TACRED test. seed — verify

72.7
+1.4%
Source
Total Improvement
2.7%
Time Span
1y 3m
Breakthroughs
3
Current SOTA
72.7

Top Models Performance Comparison

Top 3 models ranked by f1

f11LUKE72.7100.0%2KEPLER71.798.6%3SpanBERT70.897.4%0%25%50%75%100%% of best
Best Score
72.7
Top Model
LUKE
Models Compared
3
Score Range
1.9

f1Primary

#ModelScorePaper / CodeDate
1
LUKEOpen Source
Studio Ousia		72.7
editorial
Apr 2026
2
KEPLEROpen Source
Tsinghua		71.7
editorial
Apr 2026
3
SpanBERTOpen Source
Facebook AI		70.8
editorial
Apr 2026