arXiv subjects

Arjun Guha

Publications and source records attributed to Arjun Guha.

At least 19 recordsLinked to original sources

arXiv2026

The Best Programming Language for Tokenmaxxing: An Investigation of Coding Agent Behavior Across Programming Languages

Although coding agents are now very effective in a variety of programming languages, this paper first shows that the cost (in tokens) can very significantly by programming language. We evaluate five recent models on programming problems in Python, Java, Rust, and OCaml. We carefully control for problem difficulty, and show that there can be stark variation in token consumption that is consistent across models. To understand why, we analyze both the structure and content of agent trajectories. First, we re-execute every intermediate solution and abstract each trajectory as a sequence of test-outcome vectors, then label the work between successive solutions. This reveals agents repeatedly producing noncompiling solutions in unfamiliar languages and revising solutions that already pass. Second, we analyze trajectory text, finding that agents plan solutions in code comments, distrust the provided tests in favor of inputs they invent, and sidestep unfamiliar target languages by prototyping in Python. Our results show that by-language token efficiency is a metric that should be considered when benchmarking and developing multilingual agents, and, for the tokenmaxxer, a guide to the most expensive language to work in.

cs.SE↗

arXiv2026

Learning Reasoning World Models for Parallel Code

Large language models have shown remarkable ability in serial code generation, but they still struggle with parallel code for which training data is comparatively scarce. A common remedy is to use coding agents that interact with external tools, but tool calls can be costly and sometimes impractical, e.g., for partially written code. We propose Parallel-Code World Models (PCWMs), reasoning LLMs that aim to predict tool outcomes directly from parallel source code. To train PCWMs, we design a novel exploration and data generation pipeline that samples diverse parallel-coding problems and candidate implementations across multiple domains, then executes them via tools to record data races and performance profiles. From these, we synthesize reasoning traces that causally connect source code to observed tool outcomes. Fine-tuning on the resulting data yields noticeable gains, with a 7B-parameter world model improving from 64.3% to 72.8% accuracy in race-outcome prediction, while an 8B-parameter model improves in a performance profiling task from 49.3% to 58.6% accuracy. Furthermore, when open-weight models were tasked with fixing data races, world-model feedback improved their race-fixing rates relative to self-feedback by 2.7%-9.1% using our 7B-parameter world model and by 6.1%-11.1% using our 14B-parameter world model. Our results suggest that reasoning world models may have the potential to serve alongside external tool calls in parallel-coding agents.

cs.SE↗

arXiv2026

Steering Code LLMs with Activation Directions for Language and Library Control

Code LLMs often default to particular programming languages and libraries under neutral prompts. We investigate whether these preferences are encoded as approximately linear directions in activation space that can be manipulated at inference time. Using a difference-in-means method, we estimate layer-wise steering vectors for five language/library pairs and add them to model hidden states during generation. Across three open-weight code LLMs, these interventions substantially increase generation toward the target ecosystem under neutral prompts and often remain effective even when prompts explicitly request the opposite choice. Steering strength varies by model and target, with common ecosystems easier to induce than rarer alternatives, and overly strong interventions can reduce output quality. Overall, our results suggest that code-style preferences in LLMs are partly represented by compact, steerable structure in activation space.

cs.LG↗

arXiv2025

AgentPack: A Dataset of Code Changes, Co-Authored by Agents and Humans

Fine-tuning large language models for code editing has typically relied on mining commits and pull requests. The working hypothesis has been that commit messages describe human intent in natural language, and patches to code describe the changes that implement that intent. However, much of the previously collected data is noisy: commit messages are terse, human-written commits commingle several unrelated edits, and many commits come from simple, rule-based bots. The recent adoption of software engineering agents changes this landscape. Code changes \emph{co-authored} by humans and agents are often accompanied by substantially more explicit natural-language descriptions of intent and rationale. Moreover, when these changes land in public repositories, they are implicitly filtered by humans: maintainers discard low-quality commits to their projects. We present AgentPack, a corpus of 1.8M code edits co-authored by Claude Code, OpenAI Codex, and Cursor Agent across public GitHub projects up to early October 2025. We describe the identification and curation pipeline, quantify adoption trends of these agents, and analyze the structural properties of the edits. Finally, we show that models fine-tuned on AgentPack can outperform models trained on prior human-only commit corpora, highlighting the potential of using public data from software engineering agents to train future code-editing models.

cs.SE↗

arXiv2025

Bridging the Gap Between Binary and Source Based Package Management in Spack

Binary package managers install software quickly but they limit configurability due to rigid ABI requirements that ensure compatibility between binaries. Source package managers provide flexibility in building software, but compilation can be slow. For example, installing an HPC code with a new MPI implementation may result in a full rebuild. Spack, a widely deployed, HPC-focused package manager, can use source and pre-compiled binaries, but lacks a binary compatibility model, so it cannot mix binaries not built together. We present splicing, an extension to Spack that models binary compatibility between packages and allows seamless mixing of source and binary distributions. Splicing augments Spack's packaging language and dependency resolution engine to reuse compatible binaries but maintains the flexibility of source builds. It incurs minimal installation-time overhead and allows rapid installation from binaries, even for ABI-sensitive dependencies like MPI that would otherwise require many rebuilds.

cs.SE↗

arXiv2025

Agnostics: Learning to Code in Any Programming Language via Reinforcement with a Universal Learning Environment

Large language models (LLMs) already excel at writing code in high-resource languages such as Python and JavaScript, yet stumble on low-resource languages that remain essential to science and engineering. Besides the obvious shortage of pre-training data, post-training itself is a bottleneck: every new language seems to require new datasets, test harnesses, and reinforcement-learning (RL) infrastructure. We introduce Agnostics, a language-agnostic post-training pipeline that eliminates this per-language engineering. The key idea is to judge code solely by its externally observable behavior, so a single verifier can test solutions written in any language. Concretely, we (i) use an LLM to rewrite existing unit-test datasets into an I/O format, (ii) supply a short configuration that tells the verifier how to compile and run a target language, and (iii) apply reinforcement learning with verifiable rewards (RLVR) in a robust code execution environment. Applied to five low-resource languages--Lua, Julia, R, OCaml, and Fortran--Agnostics (1) improves Qwen-3 4B to performance that rivals other 16B-70B open-weight models; (2) scales cleanly to larger and diverse model families (Qwen-3 8B, DeepSeek Coder 6.7B Instruct, Phi 4 Mini); and (3) for ${\le} 16$B parameter models, sets new state-of-the-art pass@1 results on MultiPL-E and a new multi-language version of LiveCodeBench that we introduce. We release the language-agnostic training datasets (Ag-MBPP-X, Ag-Codeforces-X, Ag-LiveCodeBench-X), training code, and ready-to-use configurations, making RL post-training in any programming language as simple as editing a short YAML file.

cs.LG↗

arXiv2025

More Than a Score: Probing the Impact of Prompt Specificity on LLM Code Generation

State-of-the-art Large Language Models (LLMs) achieve high pass@1 on general benchmarks like HumanEval but underperform on specialized suites such as ParEval. Is this due to LLMs missing domain knowledge or insufficient prompt detail is given? To answer this, we introduce PartialOrderEval, which augments any code generation benchmark with a partial order of prompts from minimal to maximally detailed. Applying it to HumanEval and both serial and OpenMP subsets of ParEval, we measure how pass@1 scales with prompt specificity. Our experiments with Llama-3.x and Qwen2.5-Coder demonstrate varying degrees of prompt sensitivity across different tasks, and a qualitative analysis highlights explicit I/O specifications, edge-case handling, and stepwise breakdowns as the key drivers of prompt detail improvement.

cs.CL↗

arXiv2025

Leveraging AI for Productive and Trustworthy HPC Software: Challenges and Research Directions

We discuss the challenges and propose research directions for using AI to revolutionize the development of high-performance computing (HPC) software. AI technologies, in particular large language models, have transformed every aspect of software development. For its part, HPC software is recognized as a highly specialized scientific field of its own. We discuss the challenges associated with leveraging state-of-the-art AI technologies to develop such a unique and niche class of software and outline our research directions in the two US Department of Energy--funded projects for advancing HPC Software via AI: Ellora and Durban.

cs.SE↗

arXiv2025

"I Would Have Written My Code Differently'': Beginners Struggle to Understand LLM-Generated Code

Large language models (LLMs) are being increasingly adopted for programming work. Prior work shows that while LLMs accelerate task completion for professional programmers, beginning programmers struggle to prompt models effectively. However, prompting is just half of the code generation process -- when code is generated, it must be read, evaluated, and integrated (or rejected). How accessible are these tasks for beginning programmers? This paper measures how well beginners comprehend LLM-generated code and explores the challenges students face in judging code correctness. We compare how well students understand natural language descriptions of functions and LLM-generated implementations, studying 32 CS1 students on 160 task instances. Our results show a low per-task success rate of 32.5\%, with indiscriminate struggles across demographic populations. Key challenges include barriers for non-native English speakers, unfamiliarity with Python syntax, and automation bias. Our findings highlight the barrier that code comprehension presents to beginning programmers seeking to write code with LLMs.

cs.SE↗

arXiv2025

ReasoningWeekly: A General Knowledge and Verbal Reasoning Challenge for Large Language Models

Existing benchmarks for frontier models often test specialized, "PhD-level" knowledge that is difficult for non-experts to grasp. In contrast, we present a benchmark with 613 problems based on the NPR Sunday Puzzle Challenge that requires only general knowledge. Our benchmark is challenging for both humans and models; however correct solutions are easy to verify, and models' mistakes are easy to spot. As LLMs are more widely deployed in society, we believe it is useful to develop benchmarks for frontier models that humans can understand without the need for deep domain expertise. Our work reveals capability gaps that are not evident in existing benchmarks: OpenAI o1 significantly outperforms other reasoning models on our benchmark, despite being on par with other models when tested on benchmarks that test specialized knowledge. Furthermore, our analysis of reasoning outputs uncovers new kinds of failures. DeepSeek R1, for instance, often concedes with "I give up" before providing an answer that it knows is wrong. R1 can also be remarkably "uncertain" in its output and in rare cases, it does not "finish thinking," which suggests the need for techniques to ``wrap up'' before the context window limit is reached. We also quantify the effectiveness of reasoning longer to identify the point beyond which more reasoning is unlikely to improve accuracy on our benchmark.

cs.AI↗

arXiv2024

SelfCodeAlign: Self-Alignment for Code Generation

Instruction tuning is a supervised fine-tuning approach that significantly improves the ability of large language models (LLMs) to follow human instructions. We propose SelfCodeAlign, the first fully transparent and permissive pipeline for self-aligning code LLMs without extensive human annotations or distillation. SelfCodeAlign employs the same base model for inference throughout the data generation process. It first extracts diverse coding concepts from high-quality seed snippets to generate new tasks. It then samples multiple responses per task, pairs each with test cases, and validates them in a sandbox environment. Finally, passing examples are selected for instruction tuning. In our primary experiments, we use SelfCodeAlign with CodeQwen1.5-7B to generate a dataset of 74k instruction-response pairs. Finetuning on this dataset leads to a model that achieves a 67.1 pass@1 on HumanEval+, surpassing CodeLlama-70B-Instruct despite being ten times smaller. Across all benchmarks, this finetuned model consistently outperforms the original version trained with OctoPack, the previous state-of-the-art method for instruction tuning without human annotations or distillation. Additionally, we show that SelfCodeAlign is effective across LLMs of various sizes, from 3B to 33B, and that the base models can benefit more from alignment with their own data distribution. We further validate each component's effectiveness in our pipeline, showing that SelfCodeAlign outperforms both direct distillation from GPT-4o and leading GPT-3.5-based distillation methods, such as OSS-Instruct and Evol-Instruct. SelfCodeAlign has also led to the creation of StarCoder2-Instruct, the first fully transparent, permissively licensed, and self-aligned code LLM that achieves state-of-the-art coding performance.

cs.CL↗

arXiv2024

Creating and Repairing Robot Programs in Open-World Domains

Using Large Language Models (LLMs) to produce robot programs from natural language has allowed for robot systems that can complete a higher diversity of tasks. However, LLM-generated programs may be faulty, either due to ambiguity in instructions, misinterpretation of the desired task, or missing information about the world state. As these programs run, the state of the world changes and they gather new information. When a failure occurs, it is important that they recover from the current world state and avoid repeating steps that they they previously completed successfully. We propose RoboRepair, a system which traces the execution of a program up until error, and then runs an LLM-produced recovery program that minimizes repeated actions. To evaluate the efficacy of our system, we create a benchmark consisting of eleven tasks with various error conditions that require the generation of a recovery program. We compare the efficiency of the recovery program to a plan built with an oracle that has foreknowledge of future errors.

cs.RO↗

arXiv2024

Substance Beats Style: Why Beginning Students Fail to Code with LLMs

Although LLMs are increasing the productivity of professional programmers, existing work shows that beginners struggle to prompt LLMs to solve text-to-code tasks. Why is this the case? This paper explores two competing hypotheses about the cause of student-LLM miscommunication: (1) students simply lack the technical vocabulary needed to write good prompts, and (2) students do not understand the extent of information that LLMs need to solve code generation tasks. We study (1) with a causal intervention experiment on technical vocabulary and (2) by analyzing graphs that abstract how students edit prompts and the different failures that they encounter. We find that substance beats style: a poor grasp of technical vocabulary is merely correlated with prompt failure; that the information content of prompts predicts success; that students get stuck making trivial edits; and more. Our findings have implications for the use of LLMs in programming education, and for efforts to make computing more accessible with LLMs.

cs.CY↗

arXiv2024

NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals

We introduce NNsight and NDIF, technologies that work in tandem to enable scientific study of the representations and computations learned by very large neural networks. NNsight is an open-source system that extends PyTorch to introduce deferred remote execution. The National Deep Inference Fabric (NDIF) is a scalable inference service that executes NNsight requests, allowing users to share GPU resources and pretrained models. These technologies are enabled by the Intervention Graph, an architecture developed to decouple experimental design from model runtime. Together, this framework provides transparent and efficient access to the internals of deep neural networks such as very large language models (LLMs) without imposing the cost or complexity of hosting customized models individually. We conduct a quantitative survey of the machine learning literature that reveals a growing gap in the study of the internals of large-scale AI. We demonstrate the design and use of our framework to address this gap by enabling a range of research methods on huge models. Finally, we conduct benchmarks to compare performance with previous approaches. Code, documentation, and tutorials are available at https://nnsight.net/.

cs.LG↗

arXiv2024

Robo-Instruct: Simulator-Augmented Instruction Alignment For Finetuning Code LLMs

Code LLMs have shown promising results with converting tasks in natural language to programs that can be executed by service robots. We are interested in finetuning small, specialized LLMs for this purpose, but collecting datasets of task-program pairs specific to each robot is time-consuming and expensive. While approaches such as SELF-INSTRUCT and EVOL-INSTRUCT are capable of generating novel tasks given a few examples, they are unable to provide the corresponding programs that correctly abide by physical-world and robot-constraints using the provided programming interface. Using a simulator is a natural potential solution to checking for such constraints, but building simulation environments that can handle arbitrary tasks and their necessary objects and locations, is challenging. To address these challenges, we introduce ROBO-INSTRUCT, which synthesizes task-specific simulation environments on the fly during program execution, by opportunistically inferring entity properties and enforcing corresponding constraints based on how the entities are used in the task program. Additionally, ROBO-INSTRUCT integrates an LLM-aided post-processing procedure to refine instructions for better alignment with robot programs. We demonstrate the effectiveness of ROBO-INSTRUCT across multiple LLMs, showing that our fine-tuned models outperform all baseline methods and even match or surpass the performance of several larger and proprietary models.

cs.CL↗

arXiv2024

Understanding How CodeLLMs (Mis)Predict Types with Activation Steering

Large Language Models (LLMs) are widely used by software engineers for programming tasks. However, research shows that LLMs often lack a deep understanding of program semantics. Even minor changes to syntax, such as renaming variables, can significantly degrade performance across various tasks. In this work, we examine the task of type prediction: given a partially typed program, can a model predict a missing type annotations such that the resulting program is more typed? We construct a dataset of adversarial examples where models initially predict the correct types, but begin to fail after semantically irrelevant edits. This is problematic, as models should ideally generalize across different syntactic forms of semantically equivalent code. This lack of robustness suggests that models may have a shallow understanding of code semantics. Despite this, we provide evidence that LLMs do, in fact, learn robust mechanisms for type prediction-though these mechanisms often fail to activate in adversarial scenarios. By using activation steering, a method that manipulates a model's internal activations to guide it toward using latent knowledge, we restore accurate predictions on adversarial inputs. We show that steering successfully activates a type prediction mechanism that is shared by both Python and TypeScript, and is more effective than prompting with in-context examples. Across five different models, our comprehensive evaluation demonstrates that LLMs can learn generalizable representations of code semantics that transfer across programming languages.

cs.CL↗

arXiv2024

StarCoder 2 and The Stack v2: The Next Generation

The BigCode project, an open-scientific collaboration focused on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder2. In partnership with Software Heritage (SWH), we build The Stack v2 on top of the digital commons of their source code archive. Alongside the SWH repositories spanning 619 programming languages, we carefully select other high-quality data sources, such as GitHub pull requests, Kaggle notebooks, and code documentation. This results in a training set that is 4x larger than the first StarCoder dataset. We train StarCoder2 models with 3B, 7B, and 15B parameters on 3.3 to 4.3 trillion tokens and thoroughly evaluate them on a comprehensive set of Code LLM benchmarks. We find that our small model, StarCoder2-3B, outperforms other Code LLMs of similar size on most benchmarks, and also outperforms StarCoderBase-15B. Our large model, StarCoder2- 15B, significantly outperforms other models of comparable size. In addition, it matches or outperforms CodeLlama-34B, a model more than twice its size. Although DeepSeekCoder- 33B is the best-performing model at code completion for high-resource languages, we find that StarCoder2-15B outperforms it on math and code reasoning benchmarks, as well as several low-resource languages. We make the model weights available under an OpenRAIL license and ensure full transparency regarding the training data by releasing the SoftWare Heritage persistent IDentifiers (SWHIDs) of the source code data.

cs.SE↗

arXiv2024

How Beginning Programmers and Code LLMs (Mis)read Each Other

Generative AI models, specifically large language models (LLMs), have made strides towards the long-standing goal of text-to-code generation. This progress has invited numerous studies of user interaction. However, less is known about the struggles and strategies of non-experts, for whom each step of the text-to-code problem presents challenges: describing their intent in natural language, evaluating the correctness of generated code, and editing prompts when the generated code is incorrect. This paper presents a large-scale controlled study of how 120 beginning coders across three academic institutions approach writing and editing prompts. A novel experimental design allows us to target specific steps in the text-to-code process and reveals that beginners struggle with writing and editing prompts, even for problems at their skill level and when correctness is automatically determined. Our mixed-methods evaluation provides insight into student processes and perceptions with key implications for non-expert Code LLM use within and outside of education.

cs.HC↗