Factuality Correction on ASKHIST
0.92Mean Factual PrecisionRAC
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RACBase Model=llama-3.3-70b-instruct, Correction Stage=after2026.01 | 0.92 | — | — | — | — | — | |
| FCBase Model=llama-3.3-70b-instruct, Correction Stage=after2026.01 | 0.92 | — | — | — | — | — | |
| FCBase Model=granite-4.0-h-small, Correction Stage=after2026.01 | 0.9 | — | — | — | — | — | |
| CRITICBase Model=llama-3.3-70b-instruct, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| RACBase Model=llama-3.3-70b-instruct, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| LLM1Base Model=llama-3.3-70b-instruct, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| LLM1Base Model=llama-3.3-70b-instruct, Correction Stage=after2026.01 | 0.89 | — | — | — | — | — | |
| LLM2Base Model=llama-3.3-70b-instruct, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| LLM2Base Model=llama-3.3-70b-instruct, Correction Stage=after2026.01 | 0.89 | — | — | — | — | — | |
| FCBase Model=llama-3.3-70b-instruct, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| CRITICBase Model=granite-4.0-h-small, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| RACBase Model=granite-4.0-h-small, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| LLM1Base Model=granite-4.0-h-small, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| LLM2Base Model=granite-4.0-h-small, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| FCBase Model=granite-4.0-h-small, Correction Stage=before2026.01 | 0.89 | — | — | — | — | — | |
| CRITICBase Model=llama-3.3-70b-instruct, Correction Stage=after2026.01 | 0.87 | — | — | — | — | — | |
| FCBase Model=mixtral-8x22b-instruct, Correction Stage=after2026.01 | 0.87 | — | — | — | — | — | |
| RACBase Model=granite-4.0-h-small, Correction Stage=after2026.01 | 0.85 | — | — | — | — | — | |
| CRITICBase Model=mixtral-8x22b-instruct, Correction Stage=before2026.01 | 0.84 | — | — | — | — | — | |
| RACBase Model=mixtral-8x22b-instruct, Correction Stage=before2026.01 | 0.84 | — | — | — | — | — | |
| RACBase Model=mixtral-8x22b-instruct, Correction Stage=after2026.01 | 0.84 | — | — | — | — | — | |
| LLM1Base Model=mixtral-8x22b-instruct, Correction Stage=before2026.01 | 0.84 | — | — | — | — | — | |
| LLM2Base Model=mixtral-8x22b-instruct, Correction Stage=before2026.01 | 0.84 | — | — | — | — | — | |
| FCBase Model=mixtral-8x22b-instruct, Correction Stage=before2026.01 | 0.84 | — | — | — | — | — | |
| LLM1Base Model=granite-4.0-h-small, Correction Stage=after2026.01 | 0.84 | — | — | — | — | — | |
| LLM2Base Model=granite-4.0-h-small, Correction Stage=after2026.01 | 0.83 | — | — | — | — | — | |
| LLM1Base Model=mixtral-8x22b-instruct, Correction Stage=after2026.01 | 0.82 | — | — | — | — | — | |
| LLM2Base Model=mixtral-8x22b-instruct, Correction Stage=after2026.01 | 0.82 | — | — | — | — | — | |
| CRITICBase Model=granite-4.0-h-small, Correction Stage=after2026.01 | 0.82 | — | — | — | — | — | |
| CRITICBase Model=mixtral-8x22b-instruct, Correction Stage=after2026.01 | 0.77 | — | — | — | — | — | |
| RACBase Model=gpt-oss-120b, Correction Stage=after2026.01 | 0.77 | — | — | — | — | — | |
| CRITICBase Model=gpt-oss-120b, Correction Stage=before2026.01 | 0.69 | — | — | — | — | — | |
| RACBase Model=gpt-oss-120b, Correction Stage=before2026.01 | 0.69 | — | — | — | — | — | |
| LLM1Base Model=gpt-oss-120b, Correction Stage=before2026.01 | 0.69 | — | — | — | — | — | |
| LLM2Base Model=gpt-oss-120b, Correction Stage=before2026.01 | 0.69 | — | — | — | — | — | |
| FCBase Model=gpt-oss-120b, Correction Stage=before2026.01 | 0.69 | — | — | — | — | — | |
| CRITICBase Model=gpt-oss-120b, Correction Stage=after2026.01 | 0.67 | — | — | — | — | — | |
| LLM2Base Model=gpt-oss-120b, Correction Stage=after2026.01 | 0.67 | — | — | — | — | — | |
| FCBase Model=gpt-oss-120b, Correction Stage=after2026.01 | 0.67 | — | — | — | — | — | |
| LLM1Base Model=gpt-oss-120b, Correction Stage=after2026.01 | 0.42 | — | — | — | — | — | |
| CRITICBase Model=mixtral-8x22b-instruct2026.01 | — | -0.12 | -0.04 | -0.09 | -0.06 | -0.01 | |
| CRITICBase Model=llama-3.3-70b-instruct2026.01 | — | -0.03 | 0 | -0.01 | -0.01 | -0.01 | |
| CRITICBase Model=granite-4.0-h-small (32b)2026.01 | — | -0.13 | -0.09 | -0.11 | -0.05 | -0.04 | |
| CRITICBase Model=gpt-oss-120b2026.01 | — | -0.03 | -0.01 | -0.02 | 0.01 | -0.01 | |
| FC (ours)Base Model=mixtral-8x22b-instruct2026.01 | — | 0.02 | 0 | 0 | 0.02 | 0 | |
| FC (ours)Base Model=llama-3.3-70b-instruct2026.01 | — | 0.04 | 0.01 | 0.02 | 0.01 | 0.01 | |
| FC (ours)Base Model=granite-4.0-h-small (32b)2026.01 | — | 0 | 0 | 0 | 0.01 | 0 | |
| FC (ours)Base Model=gpt-oss-120b2026.01 | — | 0 | 0 | 0 | 0.01 | 0 | |
| LLM1Base Model=mixtral-8x22b-instruct2026.01 | — | -0.02 | 0 | -0.01 | 0 | 0 | |
| LLM1Base Model=llama-3.3-70b-instruct2026.01 | — | 0 | 0 | 0 | 0 | 0 | |
| LLM1Base Model=granite-4.0-h-small (32b)2026.01 | — | -0.08 | -0.03 | -0.05 | -0.03 | -0.02 | |
| LLM1Base Model=gpt-oss-120b2026.01 | — | -0.49 | -0.26 | -0.41 | -0.19 | -0.08 | |
| LLM2Base Model=mixtral-8x22b-instruct2026.01 | — | -0.03 | 0 | -0.02 | 0 | 0 | |
| LLM2Base Model=llama-3.3-70b-instruct2026.01 | — | 0.01 | 0 | 0.01 | 0 | 0 | |
| LLM2Base Model=granite-4.0-h-small (32b)2026.01 | — | -0.07 | -0.01 | -0.04 | -0.03 | -0.01 | |
| LLM2Base Model=gpt-oss-120b2026.01 | — | -0.03 | -0.01 | -0.03 | 0.03 | -0.04 | |
| RACBase Model=mixtral-8x22b-instruct2026.01 | — | -0.01 | -0.01 | -0.01 | -0.01 | 0 | |
| RACBase Model=llama-3.3-70b-instruct2026.01 | — | 0.03 | 0.01 | 0.02 | 0.02 | 0 | |
| RACBase Model=granite-4.0-h-small (32b)2026.01 | — | -0.06 | -0.01 | -0.04 | -0.01 | -0.02 | |
| RACBase Model=gpt-oss-120b2026.01 | — | 0.12 | 0.03 | 0.08 | 0.06 | 0.02 | |
| SFT (ours)Base Model=mixtral-8x22b-instruct2026.01 | — | -0.09 | 0 | -0.05 | -0.08 | 0.01 | |
| SFT (ours)Base Model=llama-3.3-70b-instruct2026.01 | — | -0.01 | 0 | 0 | 0.02 | 0 | |
| SFT (ours)Base Model=granite-4.0-h-small (32b)2026.01 | — | 0 | 0 | 0 | 0 | 0 | |
| SFT (ours)Base Model=gpt-oss-120b2026.01 | — | -0.28 | -0.06 | -0.19 | -0.09 | -0.05 |