Factual Text Generation on LongFact Objects
0.426AURCSelective Redaction
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Selective RedactionModel=GPT-OSS-20B2026.02 | 0.426 | — | |
| Atom-wise Selective AbstractionModel=GPT-OSS-20B2026.02 | 0.3639 | 14.59 | |
| Selective RedactionModel=GPT-OSS-120B2026.02 | 0.3604 | — | |
| Selective RedactionModel=QWEN3-30B-T2026.02 | 0.3224 | — | |
| Selective RedactionModel=Average2026.02 | 0.3051 | — | |
| Atom-wise Selective AbstractionModel=QWEN3-30B-T2026.02 | 0.287 | 10.97 | |
| Atom-wise Selective AbstractionModel=GPT-OSS-120B2026.02 | 0.2847 | 21.02 | |
| Selective RedactionModel=QWEN3-235B-T2026.02 | 0.2785 | — | |
| Atom-wise Selective AbstractionModel=Average2026.02 | 0.2631 | 13.77 | |
| Selective RedactionModel=LLAMA-3.3-70B-I2026.02 | 0.2348 | — | |
| Atom-wise Selective AbstractionModel=QWEN3-235B-T2026.02 | 0.2289 | 17.83 | |
| Atom-wise Selective AbstractionModel=LLAMA-3.3-70B-I2026.02 | 0.2169 | 7.64 | |
| Selective RedactionModel=QWEN3-30B-I2026.02 | 0.2087 | — | |
| Atom-wise Selective AbstractionModel=QWEN3-30B-I2026.02 | 0.1974 | 5.41 |