LLM-generated text detection on DetectRL Out-of-Domain Multi-LLM 1.0 (test) benchmark leaderboard · SOTA2 Research

LLM-generated text detection on DetectRL Out-of-Domain Multi-LLM 1.0 (test)

90.6Average Performance ScoreBinoculars + kNNProxy

Evaluation Results

Method	Links
Binoculars + kNNProxy2026.04		90.6	—	67.9	96.6	98.1
Binoculars + kNNProxy2026.04		90.6	95.4	—	96.2	97.9
Binoculars + kNNProxy2026.04		90.6	96.5	71.2	—	98.2
Binoculars + kNNProxy2026.04		90.6	97.1	75.9	96.5	—
Fast-DetectGPT + kNNProxy2026.04		78.3	—	45.3	81.8	93.6
Fast-DetectGPT + kNNProxy2026.04		78.3	90.2	—	80.8	91.8
Fast-DetectGPT + kNNProxy2026.04		78.3	90.8	47.1	—	92.3
Fast-DetectGPT + kNNProxy2026.04		78.3	91.7	51.5	82.8	—
Binoculars2026.04		73.7	—	39.7	88	91.7
Binoculars2026.04		73.7	82.2	—	87.9	91.2
Binoculars2026.04		73.7	77.2	30.3	—	91
Binoculars2026.04		73.7	81.4	35.5	88.6	—
Fast-DetectGPT2026.04		33.2	—	13	59.6	69.9
Fast-DetectGPT2026.04		33.2	0.2	—	0	1.2
Fast-DetectGPT2026.04		33.2	55.8	8.2	—	68.4
Fast-DetectGPT2026.04		33.2	56.3	8.7	57.7	—