Study the paper

Attention Is All You Need

Lessons, visuals, quizzes, flashcards, and resources—organized in teaching order.

Empirical Evaluation and Generalization

Table 4: F1 scores for English constituency parsing

Table 4: F1 scores for English constituency parsing data
label	label	source value	label	Provenance
F1 score improvement of semi-supervised Transformer over WSJ-only Transformer	Derived comparison	1.4	F1	Deterministic derivation
Transformer (4 layers) semi-supervised F1	WSJ 23 F1	92.7	F1	Paper evidence
Transformer (4 layers) WSJ only F1	WSJ 23 F1	91.3	F1	Paper evidence