Searcharxiv⌕ Search

arXiv · 2611.15797

Early Reflections on How Large Language Models are Reshaping Software Engineering Research

Abstract

Large language models (LLMs) are radically reshaping software engineering (SE) research. Yet, their steep financial and computational costs raise questions about who participates in that change, and who does not. By analysing all 1,918 papers published in the ICSE 2019-2026 Research and SEIP tracks, we uncover a shifting and highly concentrated research landscape. China and the US dominate LLM-based SE research, appearing in 74% of LLM-based publications between 2023-2026, yet they diverge sharply in their use of commercial versus open-weight models. Beneath this national dominance lies extreme institutional concentration: just 5 institutions in China and 7 in the US appear in more than half of their respective nations' LLM-based output. In the US, only half of the unique institutions active in 2023-2026 participated in LLM-based research, whereas Singapore displays broader usage, with 11 of 15 active institutions participating in LLM-based work. Furthermore, across 2023-2026 Research-track papers, industry involvement is more prevalent in LLM-based work (23.4% vs 16.9%), and academia-only papers make up a smaller share (63.4% vs 70.9%). Crucially, we expose a gap in model usage across industry and academia. SEIP track papers lean strongly towards commercial LLMs, with 77.4% featuring them and 40.5% using them exclusively. In contrast, only 21.2% of Research-track work relies strictly on commercial models, making greater use of open-weight models that typically support more sustainable and reproducible experiments. However, this divide warrants reflection on the extent to which SE research aligns with industry practice.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

David Williams, Michail Kyrkos, Max Hort, Maria Kechagia, Justyna Petke, Federica Sarro. 2026-11-11. Early Reflections on How Large Language Models are Reshaping Software Engineering Research. https://searcharxiv.com/abs/2611.15797

Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Searcharxiv by Indiciae: BibTeX · RIS.

KEEP EXPLORING

Related papers

AgentLens: Revealing The Lucky Pass Problem in SWE-Agent Evaluation

Evaluation of software engineering (SWE) agents is dominated by a binary signal: whether the final patch passes the tests. This outcome-only view treats a principled solution and a chaotic trial-and-error process as equivalent. We show that this equivalence is empirically false. We evaluate 2,614 OpenHands trajectories from eight model backends on 60 SWE-bench Verified tasks. Of these, 47 have enough passing trajectories to construct task-level process references, yielding a 1,815-trajectory evaluation subset. Among passing trajectories in this subset, 10.7% exhibit behavior we call a Lucky Pass: regression cycles, blind retries, missing verification, or temporally disordered exploration, implementation, and verification. We introduce AgentLens, a framework for process-level assessment of SWE-agent trajectories, and release AgentLens-Bench, a dataset of 1,815 trajectories annotated with quality scores, waste signals, divergence points, and 47 task-level Prefix Tree Acceptor (PTA) references. AgentLens builds PTA references by merging multiple passing solutions for the same task, and uses a context-sensitive intent labeler to assign actions to Exploration, Implementation, Verification, or Orchestration based on trajectory history rather than tool identity alone. On AgentLens-Bench, the quality score separates passing trajectories into Lucky, Solid, and Ideal tiers and further decomposes Lucky Passes into five recurring mechanisms. Across the eight model backends, Lucky rates range from 0.5% to 23.2%, and some models move by as many as five rank positions when ranked by quality score instead of pass rate. We publicly release AgentLens-Bench and the AgentLens SDK at https://github.com/agentlens-framework/agentlens, with a project page at https://agentlens-framework.github.io/agentlens/.

cs.SE↗

Where Is the Cost of Third-Party API Routers in Agentic Software Development?

Third-party API routers have become a common layer that unifies access across increasingly diverse LLM providers. In coding-agent workflows, high-autonomy operation is widely adopted because it reduces interaction overhead. As a result, a third-party API router, which sits between the agent and the upstream provider, inevitably occupies the trusted path. It can inspect and modify every request and response, yet no mechanism verifies alignment between the provider's output and the repository-level actions ultimately executed by the agent. Consequently, client-side permission mechanisms may become ineffective in practice. Whether this control gap produces real, hard-to-detect effects on software development tasks remains empirically unmeasured. In this paper, we conduct an empirical study of router-side injection in coding agents, examining four intervention levels of increasing subtlety: Response Substitution (L1), Response Append (L2), LLM-Polished Injection (L3), and LLM-Polished with Distribution Alignment Injection (L4). Moreover, we develop SIDEL, a framework for trace recording, replay, injection, and defense evaluation, with a curated dataset of 400 samples. We evaluate four representative coding agents, and further evaluate whitelist-based execution control and LLM review. Router-side intervention substantially alters repository-level actions and remains difficult for existing client-side safeguards to detect. Without additional mitigations, all evaluated agents achieved a defense success rate of 0 percent across all injection levels. Client-side mitigations and reactive reviews improve resistance but do not fully restore end-to-end control, motivating provider-side output-integrity guarantees. Our code is available at https://github.com/Riyasushin/SIDEL.

cs.SE↗

A Grey Literature Review of Graduate Students' Research Practices and Skills in the LLM Era

Context: Graduate students in computing now use Large Language Models (LLMs) throughout their research, from literature reviews and coding to data analysis and writing. Yet there is little shared understanding of which skills competent use of these tools requires, so graduate programs struggle to teach, mentor, or assess it, and unverified outputs can propagate into the scientific record. Objective: We investigate which LLM-based practices graduate students employ in their research and which skills they need to benefit from these practices. Method: We conducted a grey literature review of Reddit discussions. We retrieved 1,383 posts from five research-oriented subreddits, screened them against explicit inclusion and exclusion criteria, and analyzed the 54 included posts with thematic analysis, in which two authors independently coded each excerpt. Results: We identified seven categories of practices and nine categories of skills. Commenters locate the problem in accepting generated output without inspection rather than in using LLMs, and warn that delegating the effortful parts of a task, such as debugging, reading, or validating sources, delivers the output without building the skill the task was meant to develop. Traditional skills, such as programming and critical thinking, are what allow students to verify what an LLM produces. Conclusions: Our categories give graduate programs a concrete basis to teach students to work with LLMs without eroding the skills that research training should build, and they add research-specific skills, such as data sharing and scientific reproducibility, to existing AI competency frameworks.

cs.SE↗