List your company

Biomedical domain adaptation on Open Medical-LLM leaderboard macro-average

72.3Macro Average ScoreOriginal tokenization
50.66856.28461.967.516May 26, 2025

Evaluation results

MethodLinks
Original tokenizationBackbone=Llama3.1-8B-i2025.0572.3
Token DistillationBackbone=Llama3.1-8B-i2025.0571
Token Distillation + αNTPBackbone=Llama3.1-8B-i2025.0570.9
Original tokenizationBackbone=Llama3-8B-i2025.0570.6
Token Distillation + NTPBackbone=Llama3.1-8B-i2025.0570.6
Original tokenizationBackbone=Llama3-8B2025.0569.8
Original tokenizationBackbone=Llama3.1-8B2025.0569.2
NTP (Lampinen et al., 2018)Backbone=Llama3.1-8B-i2025.0568.9
Token Distillation + αNTPBackbone=Llama3-8B-i2025.0567.8
Token DistillationBackbone=Llama3-8B-i2025.0567.6
Token Distillation + αNTPBackbone=Llama3-8B2025.0567.6
Token Distillation + αNTPBackbone=Llama3.1-8B2025.0567.4
Token DistillationBackbone=Llama3-8B2025.0567.3
Token DistillationBackbone=Llama3.1-8B2025.0567.3
Token Distillation + NTPBackbone=Llama3-8B2025.0567.2
Token Distillation + NTPBackbone=Llama3.1-8B2025.0567.2
Mean (Gee et al., 2022)Backbone=Llama3.1-8B-i2025.0566.9
NTP (tune all embeddings)Backbone=Llama3.1-8B-i2025.0566.8
Token Distillation + NTPBackbone=Llama3-8B-i2025.0566.7
Original tokenizationBackbone=Avg2025.0566.5
ZeTT (Minixhofer et al., 2024)Backbone=Llama3-8B-i2025.0566.3
ZeTT (Minixhofer et al., 2024)Backbone=Llama3-8B2025.0566.1
NTP (Lampinen et al., 2018)Backbone=Llama3.1-8B2025.0566
NTP (Lampinen et al., 2018)Backbone=Llama3-8B2025.0565.6
NTP (Lampinen et al., 2018)Backbone=Llama3-8B-i2025.0565.3
Token Distillation + αNTPBackbone=Avg2025.0564.7
Token DistillationBackbone=Avg2025.0564.6
Original tokenizationBackbone=Mistral-7B2025.0564.5
Original tokenizationBackbone=Llama3.2-3B-i2025.0564.4
Token Distillation + NTPBackbone=Avg2025.0564.3
NTP (tune all embeddings)Backbone=Llama3-8B2025.0563.8
NTP (tune all embeddings)Backbone=Llama3-8B-i2025.0563.8
Mean (Gee et al., 2022)Backbone=Llama3-8B2025.0563.8
Mean (Gee et al., 2022)Backbone=Llama3.1-8B2025.0563.7
Mean (Gee et al., 2022)Backbone=Llama3-8B-i2025.0563.4
Token DistillationBackbone=Llama3.2-3B-i2025.0563.1
NTP (Lampinen et al., 2018)Backbone=Avg2025.0563
Token Distillation + NTPBackbone=Mistral-7B2025.0563
NTP (tune all embeddings)Backbone=Llama3.1-8B2025.0562.9
Token DistillationBackbone=Mistral-7B2025.0562.8
Token Distillation + αNTPBackbone=Mistral-7B2025.0562.8
ZeTT (Minixhofer et al., 2024)Backbone=Mistral-7B2025.0562.7
Random (Hewitt, 2021)Backbone=Llama3.1-8B-i2025.0562.6
Token Distillation + αNTPBackbone=Llama3.2-3B-i2025.0562.5
Token Distillation + NTPBackbone=Llama3.2-3B-i2025.0562.2
Original tokenizationBackbone=OLMo2-7B-i2025.0561.3
NTP (Lampinen et al., 2018)Backbone=Llama3.2-3B-i2025.0561.3
NTP (Lampinen et al., 2018)Backbone=Mistral-7B2025.0561.2
Token DistillationBackbone=OLMo2-7B-i2025.0561.2
Token Distillation + αNTPBackbone=OLMo2-7B-i2025.0561.2
Mean (Gee et al., 2022)Backbone=Avg2025.0560.8
Random (Hewitt, 2021)Backbone=Llama3-8B-i2025.0560.6
Original tokenizationBackbone=Llama3.2-3B2025.0560.1
NTP (tune all embeddings)Backbone=Avg2025.0559.8
Token Distillation + NTPBackbone=OLMo2-7B-i2025.0559.8
Random (Hewitt, 2021)Backbone=Llama3.1-8B2025.0559.3
Random (Hewitt, 2021)Backbone=Llama3-8B2025.0558.8
NTP (Lampinen et al., 2018)Backbone=OLMo2-7B-i2025.0558.7
Mean (Gee et al., 2022)Backbone=Llama3.2-3B-i2025.0558.4
Mean (Gee et al., 2022)Backbone=Mistral-7B2025.0558.3
NTP (tune all embeddings)Backbone=OLMo2-7B-i2025.0558
Mean (Gee et al., 2022)Backbone=OLMo2-7B-i2025.0558
Token Distillation + αNTPBackbone=Llama3.2-3B2025.0557.9
Token Distillation + NTPBackbone=Llama3.2-3B2025.0557.6
Random (Hewitt, 2021)Backbone=Avg2025.0557.5
Random (Hewitt, 2021)Backbone=Mistral-7B2025.0557
NTP (Lampinen et al., 2018)Backbone=Llama3.2-3B2025.0556.6
Token DistillationBackbone=Llama3.2-3B2025.0556.2
Random (Hewitt, 2021)Backbone=Llama3.2-3B-i2025.0555.8
NTP (tune all embeddings)Backbone=Mistral-7B2025.0555.7
NTP (tune all embeddings)Backbone=Llama3.2-3B-i2025.0555.5
Mean (Gee et al., 2022)Backbone=Llama3.2-3B2025.0554.2
Random (Hewitt, 2021)Backbone=OLMo2-7B-i2025.0553.9
Random (Hewitt, 2021)Backbone=Llama3.2-3B2025.0551.6
NTP (tune all embeddings)Backbone=Llama3.2-3B2025.0551.5