stella_en_1.5B_v5__ratio-broaden

stella_en_1.5B_v5__ratio-broaden

NovaSearch/stella_en_1.5B_v5 fine-tuned for the BROADEN ideation operation of RATIO (Retrieval Across Typed Ideation Operations): given a scientific query sentence, retrieve a formulation of the query at a broader scope or greater generality.

One of 9 checkpoints in the RATIO release (3 encoders x ADDRESS / BROADEN / SPECIFY); see the dataset card for the benchmark.

Input format (important)

This checkpoint was trained with Stella's built-in s2p_query prompt on queries only ("Instruct: Retrieve semantically similar text.\nQuery: "); candidate sentences are encoded without any prompt. Use the same format at inference:

from sentence_transformers import SentenceTransformer
model = SentenceTransformer("maayans/stella_en_1.5B_v5__ratio-broaden", trust_remote_code=True)
q_emb = model.encode(queries, prompt_name="s2p_query")
d_emb = model.encode(candidates)

Training

Relation-specific contrastive fine-tuning (MultipleNegativesRankingLoss) on the RATIO broaden train split. The benchmark uses a temporal split: all test papers postdate the training data of every evaluated encoder. See the paper for details.

Citation

@misc{sharon2026ratiobenchmarkretrievaltyped,
      title={RATIO: A Benchmark for Retrieval Across Typed Ideation Operations in Scientific Literature}, 
      author={Maayan Sharon and Tom Hope},
      year={2026},
      eprint={2608.27394},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2608.27394}, 
}