Papers for
software integration engineers
Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.
LLM method guarantees valid imports in generated tests without errors
No More Import Errors: Constrained Decoding for Guaranteed Valid Imports in LLM-Generated Tests
Abstract: One of the most common failure reasons of LLM-based test generation pipelines is import errors. Grammar constrained decoding can modify the probabilities of the LLM's final layer during token generation to guarantee valid imports. However, we find that it can lead to degenerate solutions, i.e., repetitive valid imports (import streaks) that exhaust the token budget before a test is generated. We propose Constrained Beam Decoding with Lookup (CBDL), a novel method that guarantees import validity while addressing import streaks by avoiding premature commitment to tokens that lead to repetitive imports. Because it operates at the decoding stage, CBDL can be integrated into existing LLM-based test generation pipelines. We implement CBDL for Python and Java and evaluate it on two test generation tasks, showing that generated tests have 100% valid imports, outperforming both static (RAG and LIBRO) and dynamic (Agentic Repair) import-validity approaches that leave 45%, 10%, and 43% of import errors unresolved. CBDL converts on average 14% of initial import errors into functionally correct tests, increasing overall test correctness by 1-4 percentage points over unconstrained generation. Even when import validity does not directly lead to a correct test, it allows agentic repair pipelines to allocate the repair budget to remaining errors, increasing correct tests by 2-7 percentage points for Python and 1-6 for Java. Finally, we show that a hybrid pipeline, where a small local model equipped with CBDL generates imports while the frontier GPT-5.2 model generates the test body, outperforms generating both with GPT-5.2 alone by 1-3 percentage points. Overall, CBDL removes a major bottleneck in LLM-based test generation: by guaranteeing import validity with a small overhead (16-21 seconds per test), it improves test correctness and enables cost-effective hybrid generation.
Llm agents improve firmware coding through iterative testing and repair
Closed-loop evaluation of LLM agents for embedded software development
Abstract: Large language models (LLMs) are increasingly deployed as coding agents that edit files, run builds and tests, inspect execution results, and repair software iteratively. Embedded firmware is a demanding target because correctness depends on closed-loop behavior under sensing, timing, and safety constraints, not only on static source quality. Yet embedded-agent evaluation remains limited and often emphasizes one-shot synthesis or offline correctness. We present a benchmark for closed-loop evaluation of embedded coding agents. Each task provides a plain-text engineering description, constrained workspace, and visible build-and-runtime surface. The agent must translate requirements into implementation and self-verification steps, then iterate until the required device behavior is achieved. The suite contains five embedded-control tasks and four feedback scenarios: one-shot generation, realistic self-verification, CI-style red/green feedback, and oracle-style detailed feedback. The implementation targets simulated ESP32 firmware for reproducibility. We evaluate seven GPT-family and Qwen-family configurations across five tasks and four scenarios, with three repetitions per condition for 420 runs. gpt-5.4 has the highest pass rate among evaluated configurations but does not saturate the benchmark; qwen3.5-27B is the strongest observed local model; and smaller local models degrade sharply in pass rate and search efficiency. These results suggest that capable local embedded coding agents are emerging.