Tag
This paper introduces LoopCoder-v2, a 7B code model that benefits most from a single rethinking loop; additional loops degrade performance, challenging the assumption that more test-time compute always helps.
LoopCoder-v2 proposes Parallel Loop Transformers (PLT) for efficient test-time computation scaling in code generation, showing that two loops yield significant gains while more loops cause diminishing returns and positional mismatch costs.