api-service-sac/s1-code-v2

s1-code v2

Earlier version of s1-code. For use, prefer v3. A System One decision model (322M, CPU) that returns the probability that a Python function answers a search in English or Spanish. Fine-tuned from convaiinnovations/laya (Apache 2.0).

Same input format as v3: the state is the file path, the function name, an empty line and the first 1,500 characters of the source; the question is the noul This code answers the search: <your search>.

Results (new held-out test, 197 questions, 25 candidates from Qwen3-Embedding)

Training

About 81,000 questions with exact English/Spanish parity from 306 public repositories, commit messages (CommitPackFT), issues (SWE-bench, SWE-Gym, SWE-smith), CoSQA and CoSQA+ searches, and filtered private functions; one positive and three granite-mined negatives per question; checkpoint taken after the first of two epochs (the second overfitted).

Known issue: the CoSQA+ part of its data had false negatives (generated code that also answers the query). They were removed for v3.

Licence and attribution

Apache 2.0. Fine-tuned from Laya by Convai Innovations. Training data includes CoSQA+ (CC-BY-4.0, Gong et al.) and CoSQA (MIT). Private training data is not released.