Hallucination Correction on SelfCheckGPT
62,500,000,000Reliability ScoreMEMIT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MEMITBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=1002026.06 | 62,500,000,000 | 1 | |
| MEMITBase Model=Mistral-7B, Sequential edits (τ)=6002026.06 | 3,330,000 | 0 | |
| MEMITBase Model=Mistral-7B, Sequential edits (τ)=1002026.06 | 90,300 | 0 | |
| ROMEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=1002026.06 | 50,000 | 2 | |
| MEMITBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=6002026.06 | 37,500 | 5 | |
| ROMEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=6002026.06 | 28,900 | 1 | |
| AlphaEditBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=6002026.06 | 9,040 | 5 | |
| ROMEBase Model=Mistral-7B, Sequential edits (τ)=6002026.06 | 7,870 | 1 | |
| ROMEBase Model=Mistral-7B, Sequential edits (τ)=1002026.06 | 6,770 | 3 | |
| MEMITBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=102026.06 | 1,840 | 93 | |
| MEMITBase Model=Mistral-7B, Sequential edits (τ)=102026.06 | 313 | 93 | |
| WISEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=6002026.06 | 263 | 99 | |
| GRACEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=6002026.06 | 75.4 | 100 | |
| GRACEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=1002026.06 | 71.7 | 100 | |
| DEFERBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=1002026.06 | 71.2 | 100 | |
| DEFERBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=6002026.06 | 71.2 | 100 | |
| GRACEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=102026.06 | 71 | 100 | |
| DEFERBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=102026.06 | 68.6 | 100 | |
| AlphaEditBase Model=Mistral-7B, Sequential edits (τ)=6002026.06 | 64.3 | 87 | |
| ROMEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=102026.06 | 47.4 | 61 | |
| GRACEBase Model=Mistral-7B, Sequential edits (τ)=6002026.06 | 32.5 | 100 | |
| DEFERBase Model=Mistral-7B, Sequential edits (τ)=102026.06 | 32.2 | 99 | |
| DEFERBase Model=Mistral-7B, Sequential edits (τ)=1002026.06 | 32.2 | 100 | |
| DEFERBase Model=Mistral-7B, Sequential edits (τ)=6002026.06 | 32.2 | 100 | |
| GRACEBase Model=Mistral-7B, Sequential edits (τ)=1002026.06 | 32.2 | 100 | |
| GRACEBase Model=Mistral-7B, Sequential edits (τ)=102026.06 | 32.1 | 100 | |
| DEFERBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=12026.06 | 23.4 | 80 | |
| DEFERBase Model=Mistral-7B, Sequential edits (τ)=12026.06 | 17.6 | 92 | |
| WISEBase Model=Mistral-7B, Sequential edits (τ)=102026.06 | 14.3 | 98 | |
| WISEBase Model=Mistral-7B, Sequential edits (τ)=6002026.06 | 10 | 92 | |
| FTBase Model=Mistral-7B, Sequential edits (τ)=6002026.06 | 7.26 | 20 | |
| AlphaEditBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=1002026.06 | 5.04 | 96 | |
| WISEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=1002026.06 | 3.54 | 99 | |
| WISEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=102026.06 | 3.47 | 97 | |
| AlphaEditBase Model=Mistral-7B, Sequential edits (τ)=1002026.06 | 2.96 | 98 | |
| ROMEBase Model=Mistral-7B, Sequential edits (τ)=102026.06 | 2.58 | 91 | |
| FTBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=6002026.06 | 2.02 | 45 | |
| ROMEBase Model=Mistral-7B, Sequential edits (τ)=12026.06 | 1.98 | 99 | |
| FTBase Model=Mistral-7B, Sequential edits (τ)=1002026.06 | 1.94 | 22 | |
| GRACEBase Model=Mistral-7B, Sequential edits (τ)=12026.06 | 1.92 | 100 | |
| AlphaEditBase Model=Mistral-7B, Sequential edits (τ)=102026.06 | 1.88 | 99 | |
| WISEBase Model=Mistral-7B, Sequential edits (τ)=1002026.06 | 1.74 | 96 | |
| AlphaEditBase Model=Mistral-7B, Sequential edits (τ)=12026.06 | 1.71 | 100 | |
| MEMITBase Model=Mistral-7B, Sequential edits (τ)=12026.06 | 1.59 | 100 | |
| AlphaEditBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=102026.06 | 1.49 | 99 | |
| FTBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=1002026.06 | 1.44 | 65 | |
| WISEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=12026.06 | 1.44 | 97 | |
| FTBase Model=Mistral-7B, Sequential edits (τ)=102026.06 | 1.41 | 44 | |
| GRACEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=12026.06 | 1.4 | 100 | |
| ROMEBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=12026.06 | 1.39 | 97 | |
| MEMITBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=12026.06 | 1.32 | 99 | |
| AlphaEditBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=12026.06 | 1.29 | 100 | |
| LOKIBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=6002026.06 | 1.14 | 85 | |
| LOKIBase Model=Mistral-7B, Sequential edits (τ)=6002026.06 | 1.13 | 87 | |
| FTBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=102026.06 | 1.12 | 84 | |
| WISEBase Model=Mistral-7B, Sequential edits (τ)=12026.06 | 1.08 | 100 | |
| LOKIBase Model=Mistral-7B, Sequential edits (τ)=1002026.06 | 1.04 | 94 | |
| LOKIBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=1002026.06 | 1.03 | 94 | |
| FTBase Model=Mistral-7B, Sequential edits (τ)=12026.06 | 1.02 | 77 | |
| FTBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=12026.06 | 1.01 | 95 | |
| LOKIBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=12026.06 | 1.01 | 98 | |
| LOKIBase Model=Llama-3-8B-Instruct, Sequential edits (τ)=102026.06 | 1.01 | 98 | |
| LOKIBase Model=Mistral-7B, Sequential edits (τ)=12026.06 | 1.01 | 99 | |
| LOKIBase Model=Mistral-7B, Sequential edits (τ)=102026.06 | 1.01 | 99 |