Relation Extraction on SciERC
53.19Relation Strict F1HGNet
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HGNetLearning Paradigm=Supervised2026.03 | 53.19 | — | — | 50.96 | 55.41 | — | |
| HGNetLearning Paradigm=Supervised, Loss Configuration=w/o LDHL2026.03 | 51.68 | — | — | 48.52 | 55.37 | — | |
| HGNetLearning Paradigm=Supervised, Loss Configuration=w/o LCAF Loss2026.03 | 47.33 | — | — | 42.7 | 52.14 | — | |
| ProUIEParameters=4B2026.04 | 46.76 | — | — | — | — | — | |
| GATLearning Paradigm=Supervised2026.03 | 46.21 | — | — | 40.37 | 49.11 | — | |
| GCNLearning Paradigm=Supervised2026.03 | 45.62 | — | — | 40.13 | 48.78 | — | |
| GATLearning Paradigm=Supervised, Loss Configuration=w/o LDHL2026.03 | 45.48 | — | — | 38.96 | 49.25 | — | |
| InstructUIE2023.04 | 45.15 | — | — | — | — | — | |
| InstructUIEParameters=11B2026.04 | 45.15 | — | — | — | — | — | |
| GCNLearning Paradigm=Supervised, Loss Configuration=w/o LDHL2026.03 | 44.98 | — | — | 38.46 | 48.51 | — | |
| KnowCoder-Xsupervised evaluation=true2024.11 | 44.93 | — | — | — | — | — | |
| KnowCoder-XEvaluation Mode=Supervised2024.11 | 44.93 | — | — | — | — | — | |
| HGERELearning Paradigm=Supervised2026.03 | 43.86 | — | — | 37.72 | 47.35 | — | |
| IEPILEsupervised evaluation=true2024.11 | 43.53 | — | — | — | — | — | |
| IEPILEEvaluation Mode=Supervised2024.11 | 43.53 | — | — | — | — | — | |
| PL-MarkerLearning Paradigm=Supervised2026.03 | 41.63 | — | — | 35.6 | 44.97 | — | |
| YAYI-UIEsupervised evaluation=true2024.11 | 40.94 | — | — | — | — | — | |
| YAYI-UIEEvaluation Mode=Supervised2024.11 | 40.94 | — | — | — | — | — | |
| KnowCoderModel=7B, Evaluation Setting=supervised2024.03 | 40 | — | — | — | — | — | |
| KnowCoderEvaluation Mode=Supervised2024.11 | 40 | — | — | — | — | — | |
| KnowCoderParameters=7B2026.04 | 40 | — | — | — | — | — | |
| E2H-largeModel Category=Unified Models, Model Scale=large, Large-scale continued pre-training=false2023.05 | 39 | — | — | — | — | — | |
| InstructUIE2023.04 | 38.4 | — | — | — | — | — | |
| UniEXNumber of Parameters=372M, Backbone=RoBERTa-large2023.05 | 38 | — | — | — | — | — | |
| USMEvaluation Setting=supervised2024.03 | 37.4 | — | — | — | — | — | |
| InstructUIEEvaluation Mode=Supervised2024.11 | 37.4 | — | — | — | — | — | |
| USM2023.04 | 37.36 | — | — | — | — | — | |
| USM2026.04 | 37.36 | — | — | — | — | — | |
| UniREModel Category=Specialized Models, Large-scale continued pre-training=false2023.05 | 36.9 | — | — | — | — | — | |
| PUREModel Category=Specialized Models, Large-scale continued pre-training=false2023.05 | 36.8 | — | — | — | — | — | |
| PURELearning Paradigm=Supervised2026.03 | 36.78 | — | — | 34.39 | 38.46 | — | |
| UIEBackbone=T5-v1.1-large, Pre-training=Large-scale structure-text data2022.03 | 36.53 | — | — | — | — | — | |
| UIE*Model Category=Unified Models, Large-scale continued pre-training=true2023.05 | 36.53 | — | — | — | — | — | |
| UIE2023.04 | 36.53 | — | — | — | — | — | |
| UIE2026.04 | 36.53 | — | — | — | — | — | |
| Zhong and Chen (2021)Type=Comparable SOTA2022.03 | 35.6 | — | — | — | — | — | |
| T5-largeModel Category=Unified Models, Model Scale=large, Large-scale continued pre-training=false2023.05 | 35.44 | — | — | — | — | — | |
| E2H-baseModel Category=Unified Models, Model Scale=base, Large-scale continued pre-training=false2023.05 | 35.06 | — | — | — | — | — | |
| SELBackbone=T5-v1.1-large, Pre-training=None2022.03 | 33.36 | — | — | — | — | — | |
| UIENumber of Parameters=770M, Backbone=T5-large2023.05 | 33.36 | — | — | — | — | — | |
| T5-baseModel Category=Unified Models, Model Scale=base, Large-scale continued pre-training=false2023.05 | 33.13 | — | — | — | — | — | |
| llama-3.3-70b-versatileLearning Paradigm=Zero-Shot2026.03 | 22.39 | — | — | 22.15 | 22.53 | — | |
| APIEModel=Deepseek-v3-660B2025.08 | 21.45 | — | — | — | — | — | |
| openai/gpt-oss-120bLearning Paradigm=Zero-Shot2026.03 | 20.45 | — | — | 19.68 | 21.27 | — | |
| RSLModel=Deepseek-v3-660B2025.08 | 20.35 | — | — | — | — | — | |
| ZSLModel=Deepseek-v3-660B2025.08 | 20.3 | — | — | — | — | — | |
| Active-PromptModel=Gemma-3-12B2025.08 | 19.25 | — | — | — | — | — | |
| RSLModel=Qwen-2.5-14B2025.08 | 19.23 | — | — | — | — | — | |
| Active-PromptModel=Deepseek-v3-660B2025.08 | 18.9 | — | — | — | — | — | |
| qwen/qwen3-32bLearning Paradigm=Zero-Shot2026.03 | 18.2 | — | — | 16.57 | 19.33 | — | |
| ZSLModel=Qwen-2.5-14B2025.08 | 18.08 | — | — | — | — | — | |
| APIEModel=Gemma-3-12B2025.08 | 16.85 | — | — | — | — | — | |
| GPT-3.5 TurboLearning Paradigm=Zero-Shot2026.03 | 14.98 | — | — | 14.97 | 15.02 | — | |
| Qwen3-32Bcheckpoint=open-source base, fine-tuning=none, parameters=32B2026.06 | 14.85 | — | — | — | — | — | |
| ZSLModel=Gemma-3-12B2025.08 | 14.53 | — | — | — | — | — | |
| RSLModel=Gemma-3-12B2025.08 | 14.43 | — | — | — | — | — | |
| KD SortModel=Deepseek-v3-660B2025.08 | 14.2 | — | — | — | — | — | |
| APIEModel=Qwen-2.5-14B2025.08 | 14.11 | — | — | — | — | — | |
| llama-3.1-8b-instantLearning Paradigm=Zero-Shot2026.03 | 13.92 | — | — | 13.3 | 14.27 | — | |
| Active-PromptModel=Qwen-2.5-14B2025.08 | 13.74 | — | — | — | — | — | |
| APIEModel=Deepseek-R1-14B2025.08 | 13.58 | — | — | — | — | — | |
| Agents-K1 (4B)training=GRPO on IEPile, parameters=4B2026.06 | 12.7 | — | — | — | — | — | |
| LC-ICLParadigm=ICL, Backbone=Llama-3.3-70B-Instruct2026.06 | 12.21 | — | — | — | — | — | |
| BCLModel=Llama-3.1-8b, Token Cost=5012026.06 | 12.06 | — | — | — | — | — | |
| Qwen3-4Bcheckpoint=open-source base, fine-tuning=none, parameters=4B2026.06 | 11.66 | — | — | — | — | — | |
| Active-PromptModel=Deepseek-R1-14B2025.08 | 11.48 | — | — | — | — | — | |
| RSLModel=Deepseek-R1-14B2025.08 | 11.3 | — | — | — | — | — | |
| SMADE-IEBackbone=GPT-3.5-Turbo-0125, Prompting Strategy=multi-agent debate2026.06 | 11.2 | — | — | — | — | 21.47 | |
| ZSLModel=Deepseek-R1-14B2025.08 | 11.06 | — | — | — | — | — | |
| AEiOBackbone=GPT-3.5-Turbo-0125, Prompting Strategy=monolithic prompting2026.06 | 10.91 | — | — | — | — | 19.98 | |
| KD SortModel=Deepseek-R1-14B2025.08 | 10.75 | — | — | — | — | — | |
| BCLModel=Pixtral-12B, Token Cost=5012026.06 | 10.34 | — | — | — | — | — | |
| KD SortModel=Gemma-3-12B2025.08 | 9.84 | — | — | — | — | — | |
| Qwen3-8Bcheckpoint=open-source base, fine-tuning=none, parameters=8B2026.06 | 9.65 | — | — | — | — | — | |
| BCLModel=Qwen-2.5-7b, Token Cost=5012026.06 | 9.57 | — | — | — | — | — | |
| KD SortModel=Qwen-2.5-14B2025.08 | 9.55 | — | — | — | — | — | |
| PEParadigm=ICL, Backbone=Llama-3.3-70B-Instruct2026.06 | 9.35 | — | — | — | — | — | |
| One-StepBackbone=GPT-3.5-Turbo-0125, Prompting Strategy=each-type prompting2026.06 | 8.94 | — | — | — | — | 18.56 | |
| CROSSAGENTIEBackbone=GPT-3.5-Turbo-0125, Prompting Strategy=multi-agent debate2026.06 | 8.79 | — | — | — | — | 14.61 | |
| BCLModel=Qwen-2.5-3b, Token Cost=5012026.06 | 8.28 | — | — | — | — | — | |
| One-shotModel=Pixtral-12B, Token Cost=3852026.06 | 7.86 | — | — | — | — | — | |
| One-shotModel=Llama-3.1-8b, Token Cost=3852026.06 | 7.18 | — | — | — | — | — | |
| CodexPrompt Type=code, Shots=12023.05 | 6.05 | — | 30.11 | — | — | — | |
| G&OBackbone=GPT-3.5-Turbo-0125, Prompting Strategy=each-type prompting2026.06 | 5.58 | — | — | — | — | 13.07 | |
| One-shotModel=Qwen-2.5-7b, Token Cost=3852026.06 | 5.54 | — | — | — | — | — | |
| One-shotModel=Qwen-2.5-3b, Token Cost=3852026.06 | 5.01 | — | — | — | — | — | |
| UIE-largePrompt Type=text, Shots=12023.05 | 4.65 | — | — | — | — | — | |
| ChatIEModel=Pixtral-12B, Token Cost=9422026.06 | 4.4 | — | — | — | — | — | |
| LC-ICLParadigm=ICL, Backbone=Llama-3.1-8B-Instruct2026.06 | 4.18 | — | — | — | — | — | |
| DIParadigm=w/o ICL, Backbone=Llama-3.3-70B-Instruct2026.06 | 2.78 | — | — | — | — | — | |
| PEParadigm=ICL, Backbone=Llama-3.1-8B-Instruct2026.06 | 2.02 | — | — | — | — | — | |
| ChatIEModel=Llama-3.1-8b, Token Cost=9422026.06 | 1.03 | — | — | — | — | — | |
| ChatIEModel=Qwen-2.5-7b, Token Cost=9422026.06 | 0.88 | — | — | — | — | — | |
| ChatIEModel=Qwen-2.5-3b, Token Cost=9422026.06 | 0 | — | — | — | — | — | |
| CodeIEModel=Qwen-2.5-3b, Token Cost=11722026.06 | 0 | — | — | — | — | — | |
| CodeIEModel=Qwen-2.5-7b, Token Cost=11722026.06 | 0 | — | — | — | — | — | |
| CodeIEModel=Llama-3.1-8b, Token Cost=11722026.06 | 0 | — | — | — | — | — | |
| CodeIEModel=Pixtral-12B, Token Cost=11722026.06 | 0 | — | — | — | — | — | |
| DIParadigm=w/o ICL, Backbone=Llama-3.1-8B-Instruct2026.06 | 0 | — | — | — | — | — | |
| Bert LargeParam=355M, Data=16G, FLOPS=9.07E192023.05 | — | 81.37 | — | — | — | — |