AI Paper Atlas

Research map / Benchmarks / PHOENIX14

PHOENIX14 leaderboard

13 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on PHOENIX14. Best WER (%) (dev): 16.9 (HyperSign, AAAI 2026).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

PHOENIX14 · WER (%) (dev)

12 papers, each with its best reported number for its own method; lower is better.

#WER (%) ↓MethodPaperTopic
116.9HyperSign
ResNet34
HyperSign: Saliency-Aware Spatial Graphs and Temporal Hypergraphs for Continuous Sign Language Recognition
AAAI 2026 · Table 1
Knowledge and temporal graphs
217.1OLMD
ResNet34
OLMD: Orientation-aware Long-term Motion Decoupling for Continuous Sign Language Recognition
AAAI 2025 · Table 1
Video segmentation and tracking
317.4KD-MSLRTKD-MSLRT: Lightweight Sign Language Recognition Model Based on Mediapipe and 3D to 1D Knowledge Distillation
AAAI 2025 · Table 1
Language, safety and interpretability
418.1TCNet
ResNet-18
TCNet: Continuous Sign Language Recognition from Trajectories and Correlated Regions
AAAI 2024 · Table 1
Video segmentation and tracking
518.1CSGC
ResNet18
Cross-Sentence Gloss Consistency for Continuous Sign Language Recognition
AAAI 2024 · Table 7
Language, safety and interpretability
618.2MultiSignGraphSignGraph: A Sign Sequence is Worth Graphs of Nodes
CVPR 2024 · Table 4
Knowledge and temporal graphs
718.8CorrNet
ResNet18
Continuous Sign Language Recognition With Correlation Network
CVPR 2023 · Table: Comparison with state-o
Video segmentation and tracking
819.7CoSign-2sCoSign: Exploring Co-occurrence Signals in Skeleton-based Continuous Sign Language Recognition
ICCV 2023 · Table 1
Video segmentation and tracking
919.7TLP
ResNet18
Temporal Lift Pooling for Continuous Sign Language Recognition
ECCV 2022 · Table 3
Action recognition
1020.8SMKD (ours)
ResNet18+BiLSTM
Self-Mutual Distillation Learning for Continuous Sign Language Recognition
ICCV 2021 · Table 5
Video segmentation and tracking
2 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

PHOENIX14 · WER (%)

10 papers, each with its best reported number for its own method; lower is better.

#WER (%) ↓MethodPaperTopic
117.1HyperSign
ResNet34
HyperSign: Saliency-Aware Spatial Graphs and Temporal Hypergraphs for Continuous Sign Language Recognition
AAAI 2026 · Table 1
Knowledge and temporal graphs
217.2OLMD
ResNet34
OLMD: Orientation-aware Long-term Motion Decoupling for Continuous Sign Language Recognition
AAAI 2025 · Table 1
Video segmentation and tracking
318.9TCNet
ResNet-18
TCNet: Continuous Sign Language Recognition from Trajectories and Correlated Regions
AAAI 2024 · Table 1
Video segmentation and tracking
419CSGC
ResNet18
Cross-Sentence Gloss Consistency for Continuous Sign Language Recognition
AAAI 2024 · Table 7
Language, safety and interpretability
519.4CorrNet
ResNet18
Continuous Sign Language Recognition With Correlation Network
CVPR 2023 · Table: Comparison with state-o
Video segmentation and tracking
620.1CoSign-2sCoSign: Exploring Co-occurrence Signals in Skeleton-based Continuous Sign Language Recognition
ICCV 2023 · Table 1
Video segmentation and tracking
720.8TLP
ResNet18
Temporal Lift Pooling for Continuous Sign Language Recognition
ECCV 2022 · Table 3
Action recognition
821SEN
ResNet18
Self-Emphasizing Network for Continuous Sign Language Recognition
AAAI 2023 · Table: Comparison with state-o
Video segmentation and tracking
921SMKD (ours)
ResNet18+BiLSTM
Self-Mutual Distillation Learning for Continuous Sign Language Recognition
ICCV 2021 · Table 5
Video segmentation and tracking
1022.3Baseline+VAC
ResNet18
Visual Alignment Constraint for Continuous Sign Language Recognition
ICCV 2021 · Table 5
Video segmentation and tracking

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.