ModernBERT-bio-large is the Large variant of our English biomedical encoder, built by continued pretraining of ModernBERT-large using a CLM detour recipe. Instead of standard MLM continued pretraining, we temporarily switch to causal language modeling (CLM) before returning to MLM.
ModernBERT-bio-large achieves 78.7% average F1 across 11 English biomedical benchmarks, the highest overall score, outperforming both the MLM baseline (+0.8pp, 7/11 task wins) and all other models.
You can use this model with the transformers library (v4.48.0+):
pip install -U transformers>=4.48.0
If your GPU supports it, install Flash Attention for best efficiency:
pip install flash-attn
Masked Language Modeling
python
1from transformers import AutoTokenizer, AutoModelForMaskedLM
23model_id ="almanach/ModernBERT-bio-large"4tokenizer = AutoTokenizer.from_pretrained(model_id)5model = AutoModelForMaskedLM.from_pretrained(model_id)67text ="The patient was diagnosed with [MASK] and started on antibiotics."8inputs = tokenizer(text, return_tensors="pt")9outputs = model(**inputs)1011masked_index = inputs["input_ids"][0].tolist().index(tokenizer.mask_token_id)12predicted_token_id = outputs.logits[0, masked_index].argmax(axis=-1)13predicted_token = tokenizer.decode(predicted_token_id)14print("Predicted token:", predicted_token)
Fine-tuning (Classification, NER, etc.)
python
1from transformers import AutoTokenizer, AutoModel
23model_id ="almanach/ModernBERT-bio-large"4tokenizer = AutoTokenizer.from_pretrained(model_id)5model = AutoModel.from_pretrained(model_id)67text ="The patient presented with acute myocardial infarction and was treated with percutaneous coronary intervention."8inputs = tokenizer(text, return_tensors="pt", max_length=8192, truncation=True)9outputs = model(**inputs)10# outputs.last_hidden_state: [batch, seq_len, 1024]
Note: ModernBERT-bio does not use token type IDs. You can omit the token_type_ids parameter.
Training
Data
Corpus
Proportion
Description
PubMed
60%
Biomedical abstracts
Med-Inst
20%
Medical instructions
MIMIC
20%
Clinical notes
Total
50B tokens
Methodology
ModernBERT-bio-large is trained in two phases, initialized from ModernBERT-large:
Phase 1 (CLM detour, 50B tokens): The bidirectional attention mask is replaced with a causal mask, and the model is trained with next-token prediction. This dense training signal (100% of positions) deeply modifies early transformer layers for domain adaptation.
Phase 2 (MLM decay, 5B tokens): Bidirectional attention is restored, and the model is trained with masked language modeling at 15% masking. The learning rate decays from peak to 10% following a 1-sqrt schedule.
Both phases use the same data mix (55B tokens total). Training used AdamW (lr=2e-4, beta1=0.9, beta2=0.98), bf16 mixed precision, global batch size of 384 sequences (~3.1M tokens), on 4× H100 80GB GPUs with Composer.
Why a CLM Detour?
CLM supervises every token position, producing dense gradient updates that deeply modify early transformer layers. These changes persist through the MLM decay phase, even when the decay matches the CLM phase in length. The Large model retains 67.2% CKA divergence from its MLM counterpart (vs 56.5% for Base), showing that the effect scales with model capacity. The CLM benefit also widens at Large scale: +0.8pp (Large) vs +0.3pp (Base). See our paper for the full mechanistic analysis.
Evaluation
English biomedical benchmark results (11 tasks, 5 seeds per model):
Clinical Tasks
Model
Ctx
ChemProt
Phenotype
COS
Social Hist.
DEID
Avg
ModernBERT-bio-large
8192
90.4
61.3
94.7
56.5
84.2
77.4
MLM baseline Large (ours)
8192
90.5
61.0
94.9
55.0
82.3
76.7
BioClinical-ModernBERT-base
8192
90.0
60.7
94.8
56.0
81.8
76.7
PubMedBERT
512
90.2
52.0
95.0
48.7
80.4
73.3
BigBIO Tasks
Model
Ctx
AnatEM
BC5CDR
JNLPBA
NCBI
GAD
HoC
Avg
ModernBERT-bio-large
8192
83.2
89.8
75.3
81.7
79.7
69.3
79.8
MLM baseline Large (ours)
8192
82.0
89.4
75.5
81.8
76.4
67.8
78.8
BioClinical-ModernBERT-base
8192
79.2
88.7
74.8
78.7
75.8
67.0
77.4
PubMedBERT
512
83.3
89.7
74.9
82.1
79.3
71.0
80.1
Overall
Model
Clinical
BigBIO
Overall
ModernBERT-bio-large
77.4
79.8
78.7
MLM baseline Large (ours)
76.7
78.8
77.9
ModernBERT-bio-base
76.9
78.9
78.0
BioClinical-ModernBERT-base
76.7
77.4
77.0
PubMedBERT
73.3
80.1
77.0
ModernBERT-bio-large achieves the highest overall score (78.7%), with the CLM benefit widening at Large scale (+0.8pp vs +0.3pp for Base). The model sets new state-of-the-art on DEID (84.2%) and is competitive with the best baselines on the remaining tasks.
Intended Use
This model is designed for English biomedical and clinical NLP tasks:
Named entity recognition (diseases, chemicals, genes, anatomy)
Information extraction from PubMed abstracts and clinical reports
The 8,192-token context is important for long clinical documents. The Large size provides improved performance over Base, particularly on NER tasks (AnatEM, DEID, GAD), at the cost of higher compute requirements.
Trained on English biomedical text; not suitable for other languages without further adaptation. See ModernCamemBERT-bio for French.
Encoder model: produces contextualized representations, does not generate text.
Clinical text may contain sensitive patterns; users are responsible for compliance with applicable regulations (HIPAA, etc.).
Training data includes MIMIC clinical notes, which are de-identified but derived from real patient records.
License
Apache 2.0
Citation
bibtex
1@misc{touchent2026causallanguagemodelingdetour,
2 title={A Causal Language Modeling Detour Improves Encoder Continued Pretraining},
3 author={Rian Touchent and Eric de la Clergerie},
4 year={2026},
5 eprint={2605.12438},
6 archivePrefix={arXiv},
7 primaryClass={cs.CL},
8 url={https://arxiv.org/abs/2605.12438},
9}
Acknowledgments
This work was performed using HPC resources from GENCI-IDRIS (Grant 2024-AD011014393R2).
