SciEmbed-CTX-8192

Long-context variant (max_seq_length=8192). Recommended for long scientific inputs; best within-recipe scores on the Body-Fact Retrieval probe and LongEmbed.

A 149M-parameter ModernBERT-base scientific document embedder trained with citation-context sentences as the primary contrastive signal. Part of the SciEmbed release (Findings of EMNLP 2026).

Usage

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("J0nasW/sciembed-ctx-8192")
emb = model.encode(["citation-context supervision for scientific embeddings"],
                   normalize_embeddings=True)

• Context length: 8192 tokens
• Pooling: mean · Output dim: 768 (Matryoshka-truncatable to 512/256/128)
• License: MIT

Citation

See the repository README. Paper: SciEmbed: Citation-Context Supervision for Scientific Document Embeddings, Findings of the Association for Computational Linguistics: EMNLP 2026.