RecipeMatching V1 (Qwen3-Embedding-0.6B)

A checkpoint from the paper Recipe-Matching, Not Equivalence (Ali Habibullah, Mohammad Alshiekh, Yazan Alshoibi, Salman Khan and Naeemullah Khan, 2026), shared so the paper's evaluations of this model can be re-run without retraining. It is not one of the three models the paper releases as its main artefacts. Code, data and results: https://github.com/KAUST-Academy/recipe-matching-not-equivalence.

What it is. The paper's mixed-data model V1: Qwen3-Embedding-0.6B fine-tuned (all weights) on the 30,323-row mixed file data/phase2/mixed_pairs.jsonl, with queries encoded under the model's query prompt. V2 (KAUSTAcademy/RecipeMatching_V2_Qwen3-Embedding-0.6B) trains on the same file with a math-equivalence instruction on its queries. Seed 42; the full training record, command line included, is run_config.json in this repository, and every training file named here is in the GitHub repository.

Results of this checkpoint (seed 42, R@1 / R@5 / R@10; the paper's tables show this run; the paper notes that its medium-tier and cross-language R@1 are not representative of V1's three-seed means):

Evaluation	R@1	R@5	R@10
MathNet-Retrieve easy tier (15,000 queries, 117,088 documents)	20.83	91.96	96.71
MathNet-Retrieve medium tier	3.67	63.57	78.80
MathNet-Retrieve hard tier	0.03	7.67	30.01
Cross-language duplicates (strict)	59.80	83.21	90.84

Usage. Encode queries with the model's query prompt, the setting every number above uses.

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("KAUSTAcademy/RecipeMatching_V1_Qwen3-Embedding-0.6B")
q = model.encode(["Find all real x with x^2 = 2x."], prompt_name="query")
d = model.encode(["Determine every real solution of x^2 - 2x = 0."])
print(model.similarity(q, d))

Citation

@misc{habibullah2026recipematchingequivalence,
      title={Recipe-Matching, Not Equivalence}, 
      author={Ali Habibullah and Mohammad Alshiekh and Yazan Alshoibi and Salman Khan and Naeemullah Khan},
      year={2026},
      eprint={2609.31927},
      archivePrefix={arXiv},
      primaryClass={cs.IR},
      url={https://arxiv.org/abs/2609.31927}, 
}