Source record

Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2609.26952.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

PLLM+ is a Python dependency-repair pipeline that runs cheap deterministic steps (AST-based interpreter inference, replay of known-working configurations, live PyPI validation) before falling back to an LLM Proposer/Critic loop. On the HG2.9K benchmark it solves 1,500/2,891 snippets versus 1,169 for the baseline, and 1,495 of those fixes come from the replay step, not the LLM.

Quantified evidence that deterministic reuse of validated solutions outperformed the LLM fallback 299:1, a useful counterpoint to default-first LLM agent design.

Cited by

No citations on record.