Recent studyBlind TTS Elo is live. Compare two anonymous voice samples, vote after listening, and help separate real preference signal from noise.Vote in the study ->
Codesota · Benchmark · FUNSDHome/Leaderboards/FUNSD
Unknown

FUNSD.

199 fully annotated forms. Tests semantic entity labeling and linking.

Paper ↗Leaderboard ↓Lineage
§ 01 · SOTA history

Year over year.

§ 02 · Leaderboard

Results by metric.

Found a wrong score or missing run?
Use row edits to send a sourced correction into moderation.
Add / edit result ↗Report issue ↗

f1

F1 is the reported evaluation metric for FUNSD. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for f1verifiedpapervendorcommunityunverified
RankModelTrustScoreYearLinksEdit
01LayoutLMv3-large
LayoutLMv3-large. Table 1 in paper. ACM MM 2022. SOTA at time of publication.
verified92.082022Source ↗Edit result
02UDOP
UDOP (Unified Document Processing). Table 3 in paper. CVPR 2023. Single generative model for all document tasks.
verified91.622023Source ↗Edit result
03LayoutLMv3-base
LayoutLMv3-base. Table 1 in paper. ACM MM 2022.
verified90.292022Source ↗Edit result
04DocFormerv2-large
DocFormerv2-large. Table 5 in paper. ICCV 2023.
verified88.892023Source ↗Edit result
05LiLT[EN-R2]-base
LiLT with English RoBERTa backbone (EN-R2), base size. Table 2 in paper. ACL 2022. Best monolingual FUNSD result.
verified88.412022Source ↗Edit result
06DocFormerv2-base
DocFormerv2-base. Table 5 in paper. ICCV 2023.
verified88.372023Source ↗Edit result
07StructuralLM
StructuralLM (large). Table 1 in paper. ACL 2021. Precision 83.52, Recall 86.81.
verified85.142021Source ↗Edit result
08FormNet
FormNet. Table 1 in paper. ACL 2022. Uses rich structural encoding via graph neural network.
verified84.692022Source ↗Edit result
09BROS-large
BROS-large on FUNSD entity extraction. Table 3 in paper. AAAI 2022.
verified84.522022Source ↗Edit result
10LayoutLMv2-large
LayoutLMv2-large. Table 6 in paper. ACL 2021.
verified84.22021Source ↗Edit result
11LayoutLMv2-base
LayoutLMv2-base. Table 6 in paper. ACL 2021.
verified82.762021Source ↗Edit result
12LayoutLMv1-base
LayoutLM-base with text+layout+image embeddings, 11M docs. Best base variant. Table 1 in paper. ACL 2020.
verified79.272020Source ↗Edit result
13LayoutLMv1-large
LayoutLM-large, text+layout, MVLM, 11M docs 1 epoch. Table 1 in paper. ACL 2020.
verified77.892020Source ↗Edit result
Lineage

FUNSD in context.

See full ocr benchmarks lineage →
This benchmark (1)
saturated2019-05
FUNSD
Successors (1)
superseded2023-05
OCRBench
Once VLMs could read at all, evaluation needed to span more than forms. OCRBench bundled scene text, document VQA, KIE and handwritten math into one composite — the first VLM-era OCR benchmark.
§ 04 · Submit a result

Add to the leaderboard.

← Back to Leaderboards