Laya coding decider v2 (SWE-Gym fine-tune)

Fine-tuned from convaiinnovations/laya on ~9k coding-decision rows derived from SWE-Gym/SWE-Gym (real GitHub issues + gold fix diffs): keep/drop file relevance, priority score, injection noul.

v2 held-out results (1150 rows)

  • Overall accuracy: 0.833 | ECE: 0.147 | Brier: 0.086
  • Relevance-only accuracy: 0.907 on 1150 items (majority-drop baseline ~0.64)
  • Score MAE: 0.411 | p50 latency: 126.4 ms

Usage

import laya
agent = laya.load("evryvilanislemn/laya-decider-v2")
result = agent.predict(state, questions)
print(result["answers"])
Downloads last month

-

Downloads are not tracked for this model. How to track
Safetensors
Model size
0.4B params
Tensor type
F16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support