SEP partitions LLM code candidates into functional equivalence classes using symbolic execution on public examples and selects the dominant class, improving accuracy from 0.754 to 0.826 on HumanEval+ and 0.565 to 0.647 on LiveCodeBench at N=10.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Inference-Time Code Selection via Symbolic Equivalence Partitioning
SEP partitions LLM code candidates into functional equivalence classes using symbolic execution on public examples and selects the dominant class, improving accuracy from 0.754 to 0.826 on HumanEval+ and 0.565 to 0.647 on LiveCodeBench at N=10.