Hallucination Correction on Hallucination
15.15Error Rate (ERR)SoLA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SoLABackbone=GPT2-XL2026.03 | 15.15 | 1.01 | 7.35 | |
| GRACEBackbone=GPT2-XL2026.03 | 15.84 | 7.14 | 10 | |
| ELDERBackbone=GPT2-XL2026.03 | 16.12 | 5.87 | 8.42 | |
| MELOBackbone=GPT2-XL2026.03 | 17.45 | 1.04 | 2.66 | |
| ROMEBackbone=GPT2-XL2026.03 | 30.28 | 103.82 | 14.02 | |
| MENDBackbone=GPT2-XL2026.03 | 1,369.8 | 1,754.9 | 2,902.5 | |
| CMRBackbone=GPT2-XL2026.03 | 1,449.3 | 28.14 | 107.76 | |
| EWCBackbone=GPT2-XL2026.03 | 1,485.7 | 29.24 | 109.59 | |
| CLEARBackbone=GPT2-XL2026.03 | 2,394.3 | 35.34 | 195.82 | |
| SERACBackbone=GPT2-XL2026.03 | 8,183.7 | 133.3 | 10.04 |