Sentiment Analysis on Financial Phrase Bank (test)
0.9824AccuracyFinBERT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FinBERTRegime=Publicly available, pretrained classifier2026.06 | 0.9824 | 97.71 | |
| GPT-4oRegime=Teacher (zero-shot)2026.06 | 0.9735 | 97.57 | |
| BERT Seq-CLSTraining Strategy=individual2023.05 | 0.972 | — | |
| ModernBERTID=(1), Regime=Full2026.06 | 0.9648 | 95.78 | |
| DistilBERTID=(1), Regime=Full2026.06 | 0.9604 | 94.95 | |
| ModernBERTID=(7), Regime=105 + Frz+ES + Bal + Synth (Clust)2026.06 | 0.9515 | 94.63 | |
| ModernBERTID=(6), Regime=105 + Frz+ES + Bal + Synth (Rand)†2026.06 | 0.9457 | 93.29 | |
| DistilBERTID=(7), Regime=105 + Frz+ES + Bal + Synth (Clust)2026.06 | 0.9383 | 92.76 | |
| DistilBERTID=(6), Regime=105 + Frz+ES + Bal + Synth (Rand)†2026.06 | 0.9295 | 90.63 | |
| ModernBERTID=(5), Regime=105 + Frz+ES + Bal (Clust)2026.06 | 0.9251 | 91.25 | |
| DistilBERTID=(5), Regime=105 + Frz+ES + Bal (Clust)2026.06 | 0.9207 | 89.73 | |
| DistilBERTID=(4), Regime=105 + Frz+ES + Bal (Rand)†2026.06 | 0.9163 | 88.81 | |
| ModernBERTID=(2), Regime=1052026.06 | 0.9075 | 89.15 | |
| DistilBERTID=(3), Regime=105 + Frz+ES2026.06 | 0.8943 | 83.4 | |
| ModernBERTID=(3), Regime=105 + Frz+ES2026.06 | 0.8899 | 85.79 | |
| BERT-TinyID=(7), Regime=105 + Frz+ES + Bal + Synth (Clust)2026.06 | 0.8899 | 87.69 | |
| DistilBERTID=(9), Regime=12 + Frz+ES + Bal + Synth2026.06 | 0.8899 | 85.07 | |
| ModernBERTID=(9), Regime=12 + Frz+ES + Bal + Synth2026.06 | 0.8855 | 86.02 | |
| BERT-TinyID=(1), Regime=Full2026.06 | 0.859 | 85.38 | |
| ModernBERTID=(4), Regime=105 + Frz+ES + Bal (Rand)†2026.06 | 0.8399 | 80.37 | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=AG-news2024.05 | 0.8363 | — | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=SST22024.05 | 0.8343 | — | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=QQP2024.05 | 0.8084 | — | |
| BERT-TinyID=(3), Regime=105 + Frz+ES2026.06 | 0.8062 | 79.3 | |
| DistilBERTID=(2), Regime=1052026.06 | 0.8018 | 64.48 | |
| BERT-TinyID=(9), Regime=12 + Frz+ES + Bal + Synth2026.06 | 0.8018 | 74.67 | |
| BERT-TinyID=(5), Regime=105 + Frz+ES + Bal (Clust)2026.06 | 0.8017 | 80.27 | |
| BERT-TinyID=(6), Regime=105 + Frz+ES + Bal + Synth (Rand)†2026.06 | 0.8003 | 75.98 | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=Zero-shot2024.05 | 0.7964 | — | |
| DistilBERTID=(8), Regime=12 + Frz+ES + Bal2026.06 | 0.793 | 67.3 | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=RACE2024.05 | 0.7884 | — | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=BoolQ2024.05 | 0.7784 | — | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=MNLI2024.05 | 0.7725 | — | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=ARC-Easy2024.05 | 0.7605 | — | |
| GPT-2Training Strategy=explicit2023.05 | 0.759 | — | |
| BERT-TinyID=(4), Regime=105 + Frz+ES + Bal (Rand)†2026.06 | 0.7332 | 61.49 | |
| Bi-EncoderTraining Strategy=vanilla2023.05 | 0.717 | — | |
| GPT-2Training Strategy=vanilla2023.05 | 0.711 | — | |
| Binary BERTTraining Strategy=vanilla2023.05 | 0.689 | — | |
| BERT-TinyID=(2), Regime=1052026.06 | 0.6872 | 53.37 | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=Commonsense-QA2024.05 | 0.6866 | — | |
| ModernBERTID=(8), Regime=12 + Frz+ES + Bal2026.06 | 0.674 | 55.54 | |
| Binary BERTTraining Strategy=explicit2023.05 | 0.667 | — | |
| Binary BERTTraining Strategy=implicit2023.05 | 0.66 | — | |
| Bi-EncoderTraining Strategy=implicit2023.05 | 0.64 | — | |
| Bi-EncoderTraining Strategy=explicit2023.05 | 0.636 | — | |
| GPT-2Training Strategy=implicit2023.05 | 0.614 | — | |
| BERT-TinyID=(8), Regime=12 + Frz+ES + Bal2026.06 | 0.5859 | 47.78 | |
| GPT-3Training Strategy=Zero-shot2023.05 | 0.528 | — | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=Conll2003-POS2024.05 | 0.4391 | — | |
| BARTTraining Strategy=Zero-shot2023.05 | 0.402 | — | |
| BERT Seq-CLSTraining Strategy=full2023.05 | 0.249 | — | |
| LLaMA-2 7B Chat (cross-task prompting)Source Task=Conll2003-NER2024.05 | 0.1218 | — |