List your company

EEG-to-text decoding on THINGS EEG cross-exemplar evaluation suite v2

22.5ROUGE-1SENSE
8.87612.41315.9519.487May 27, 2026

Evaluation results

MethodLinks
SENSEAblation Configuration=Base, Decoder Model=Gemini 2.5 Flash Lite2026.0522.54.221.213.22.415.188.24.71
SYNAPSEAblation Configuration=B1 (Full framework), Decoder Model=Qwen2.5-7B2026.0521.8318.919.42.916.787.44.21
SYNAPSEAblation Configuration=B2 (Full framework w/o Exemplars), Decoder Model=Qwen2.5-7B2026.0521.52.518.119.42.716.8873.91
SYNAPSEAblation Configuration=B1 (Full framework), Decoder Model=GPT-4o-mini2026.0521.42.518.319.32.716.587.44.81
SYNAPSEAblation Configuration=B2 (Full framework w/o Exemplars), Decoder Model=Gemini 2.5 Flash Lite2026.0521.42.418.418.52.716.387.24.31
SYNAPSEAblation Configuration=B1 (Full framework), Decoder Model=Gemini 2.5 Flash Lite2026.0521.22.818.217.62.715.987.44.41
SENSEAblation Configuration=Base, Decoder Model=GPT-4o-mini2026.0521.22.719.413.42.114.188.54.81
SYNAPSEAblation Configuration=B2 (Full framework w/o Exemplars), Decoder Model=GPT-4o-mini2026.0521.12.118.118.42.515.987.44.71
SYNAPSEAblation Configuration=B1 (Full framework), Decoder Model=LLaMa-3-8B2026.0520.33.918.317.8317.187.44.51
SENSEAblation Configuration=Base, Decoder Model=LLaMa-3-8B2026.0519.94.217.117.53.117.886.24.11
SYNAPSEAblation Configuration=B2 (Full framework w/o Exemplars), Decoder Model=LLaMa-3-8B2026.0515.21.812.413.32.114.5874.21
SENSEAblation Configuration=Base, Decoder Model=Qwen2.5-7B2026.059.41.28.53.60.74.384.721