Out-of-Distribution detection on OOD Detection Source LLM GPT-4o (test)
98.7XSUM ScoreREADER
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| READERevaluation_config=per-domain2026.05 | 98.7 | 98 | 95.7 | 97.3 | 97.4 | |
| RADARevaluation_config=per-domain2026.05 | 92.3 | 76.7 | 79.3 | 78 | 81.6 | |
| BISCOPEtraining_variant=target-adapted, evaluation_config=per-domain2026.05 | 89.3 | 88.3 | 86 | 82 | 86.4 | |
| IMBDtraining_variant=fixed-training, evaluation_config=per-domain2026.05 | 87 | 95.3 | 96 | 93.3 | 92.9 | |
| BISCOPEtraining_variant=fixed-training, evaluation_config=per-domain2026.05 | 86 | 89 | 91 | 90 | 89 | |
| BINOCULARSevaluation_config=per-domain2026.05 | 83.3 | 90 | 94 | 94.3 | 90.4 | |
| FAST-DETECTGPTsurrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 83.3 | 90.3 | 94.3 | 94.3 | 90.6 | |
| ADADETECTGPTtraining_variant=fixed-training, surrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 82.7 | 89.7 | 94 | 94 | 90.1 | |
| IMBDtraining_variant=target-adapted, evaluation_config=per-domain2026.05 | 73 | 86 | 82 | 69.3 | 77.6 | |
| DETECTGPTsurrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 67.7 | 73.3 | 67 | 83.3 | 72.8 | |
| NPRsurrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 63.3 | 76.3 | 63.3 | 83.7 | 71.7 | |
| ADADETECTGPTtraining_variant=target-adapted, surrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 55 | 61.7 | 60.3 | 57.3 | 58.6 | |
| ENTROPYsurrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 54.7 | 65.3 | 68 | 67.3 | 63.8 | |
| DNAGPTsurrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 54.7 | 76.7 | 87.3 | 86.7 | 76.3 | |
| ROBERTABASEtraining_variant=target-adapted, evaluation_config=per-domain2026.05 | 54.3 | 54.7 | 58.7 | 58.7 | 56.6 | |
| LIKELIHOODsurrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 54 | 71.3 | 76.7 | 72.3 | 68.6 | |
| LRRsurrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 54 | 59.7 | 65 | 64.3 | 60.7 | |
| LOGRANKsurrogate_llm=Gemma, evaluation_config=per-domain2026.05 | 53 | 68.7 | 76 | 72 | 67.4 | |
| ROBERTALARGEtraining_variant=target-adapted, evaluation_config=per-domain2026.05 | 52.3 | 58.3 | 52.7 | 63.3 | 56.7 |