arXiv subjects

Sumit Gulwani

Publications and source records attributed to Sumit Gulwani.

At least 19 recordsLinked to original sources

arXiv2026

Back to the Future: A workbook time machine for spread sheet creation benchmarks

We introduce the workbook time machine, a pipeline that automatically creates benchmarks evaluating the ability of language models to create derived objects in spreadsheets (formulas, charts, pivot tables, and conditional formatting). Applied to public workbook corpora, it produces wtmcorpus--a collection of (input workbook, output workbook, query) triples spanning four artifact types and varying complexity. From this corpus we curate wtmbench, a 150-task evaluation benchmark with queries at three levels of specificity. We evaluate existing spreadsheet manipulation agents and baselines on wtmbench across artifact types, step complexity, and instruction granularity. Our evaluations show that query specificity, agent orchestration, and interface API used to control spreadsheets play a big role in LLM performance on Excel tasks.

cs.AI↗

arXiv2026

Reason Wide, Not Deep: Amortizing the Reasoning Premium into Distilled Skills

Reasoning modes of language models outperform their non-reasoning counterparts on multi-step agentic tasks, but pay a 3-6x premium in output tokens on every episode -- much of it spent re-deriving procedures that are shared across episodes of the same domain. We show this recurring cost can be amortized: a coding agent analyses a small corpus of existing trajectories from a training split and compiles a compact natural-language skill that is injected into the non-reasoning model's system prompt. Across four agentic benchmarks (ALFWorld, tau$^2$-bench telecom and retail, and SpreadsheetBench-Verified), skills recover 55%-100%+ of the reasoning gap for GPT-5.4-mini on held-out tasks -- exceeding the reasoning mode outright on two of four -- while emitting 2.7-6x fewer output tokens and zero reasoning tokens. Notably, reasoning traces are not a prerequisite: skills distilled from non-reasoning trajectories alone remain competitive with skills distilled from paired reasoning/non-reasoning corpora, with domain-dependent differences between the two sources. We interpret these results through a search lens: test-time reasoning is deep search inside a single episode, re-paid at every deployment, while corpus distillation is wide search across episodes, paid once. The two recover overlapping procedural knowledge, and width over cheap trajectories is often the better buy -- with the residual gap on some domains (telecom, SpreadsheetBench) delineating where genuinely per-instance deep search remains necessary.

cs.AI↗

arXiv2026

Plans Work in Mysterious Ways: Evaluating a Plan Mode for Spreadsheet Agents

Plan Modes have become standard features in agentic programming tools, allowing users to gain transparency and control by working with the agent to develop a plan before task execution. However, it remains unclear whether the benefits of this feature translate to end-user programming environments such as spreadsheets. Since spreadsheet programmers tend to work iteratively and care less about technical correctness, upfront planning may not fit into their workflows as easily. In this paper, we build a prototype of a Plan Mode for spreadsheet programming and evaluate it against a non-planning baseline through a within-subjects user study (N=24). We found that despite similar task outcomes with both tools, using Plan Mode led to a reduction in refinement and a better perception of the tool across dimensions of creativity support and human-machine collaboration. We discuss the implications of these results for the future design of Plan Modes, and for the broader role of human-AI planning in end-user programming.

cs.HC↗

arXiv2026

A Benchmark and Framework for Evaluating Next Action Predictions in Spreadsheets

Predictive code completion greatly accelerates how quickly developers work. In spreadsheets, despite being much more common, such auto-completion features are virtually non-existent. To address this gap, we introduce a benchmark for systems that observe a sequence of user actions in a spreadsheet and predict future actions. Two challenges are (1) the absence of edit histories in public spreadsheet corpora and (2) the complex space of spreadsheet actions (spatial, temporal, composite). To address (1), we manually curate 52 sequences of 12K actions that recreate spreadsheets from public corpora, seeded by parametrized heuristics and LLM refinement. To address (2), we propose an online evaluation that expects a prediction after each user action, accepts or rejects that prediction, updates the future actions upon acceptance, and repeats this until the target spreadsheet is obtained. We use multiple baseline predictors (including zero-shot LLMs, fine-tuned SLMs, and classical models) and analyze different properties that our benchmark teaches us, including but not limited to: properties of saved actions and false positives, efficiency, effect of user profiles, effect of triggers, and effect of context.

cs.SE↗

arXiv2026

SkillAxe: Sharpening LLM-Authored Agent Skills Through Evaluation-Guided Self-Refinement

Skill documents, structured natural-language instructions that guide Large Language Model (LLM) agents, are critical to modern agent frameworks, yet LLMs struggle to write skills that actually work. On SkillsBench, human-authored skills improve pass rates by 16.2 percentage points, while LLM-authored skills provide no measurable gain. We introduce SkillAxe, a fully unsupervised framework that enables LLMs to iteratively diagnose and refine their own skills. SkillAxe decomposes skill quality into four interpretable dimensions (quality impact, trigger precision, instruction compliance with fault attribution, and solution-path coverage), producing structured improvement briefs that require no ground-truth labels, test suites, or environment rewards. On SkillsBench, SkillAxe improves pass rates by 28\% relative over unimproved LLM skills and closes 47--67\% of the gap to human-authored skills. We validate the approach as a continuous improvement engine in the wild on SpreadsheetBench, where a SkillAxe-built skill library learns from past agent trajectories and raises pass rate from 16.0\% to 52.0\% using only 22 skills.

cs.MA↗

arXiv2026

Auditing and Controlling AI Agent Actions in Spreadsheets

Advances in AI agent capabilities have outpaced users' ability to meaningfully oversee their execution. AI agents can perform sophisticated, multi-step knowledge work autonomously from start to finish, yet this process remains effectively inaccessible during execution, often buried within large volumes of intermediate reasoning and outputs: by the time users receive the output, all underlying decisions have already been made without their involvement. This lack of transparency leaves users unable to examine the agent's assumptions, identify errors before they propagate, or redirect execution when it deviates from their intent. The stakes are particularly high in spreadsheet environments, where process and artifact are inseparable. Each decision the agent makes is recorded directly in cells that belong to and reflect on the user. We introduce Pista, a spreadsheet AI agent that decomposes execution into auditable, controllable actions, providing users with visibility into the agent's decision-making process and the capacity to intervene at each step. A formative study (N = 8) and a within-subjects summative evaluation (N = 16) comparing Pista to a baseline agent demonstrated that active participation in execution influenced not only task outcomes but also users' comprehension of the task, their perception of the agent, and their sense of role within the workflow. Users identified their own intent reflected in the agent's actions, detected errors that post-hoc review would have failed to surface, and reported a sense of co-ownership over the resulting output. These findings indicate that meaningful human oversight of AI agents in knowledge work requires not improved post-hoc review mechanisms, but active participation in decisions as they are made.

cs.HC↗

arXiv2026

An Empirical Investigation of Robustness in Large Language Models under Tabular Distortions

We investigate how large language models (LLMs) fail when tabular data in an otherwise canonical representation is subjected to semantic and structural distortions. Our findings reveal that LLMs lack an inherent ability to detect and correct subtle distortions in table representations. Only when provided with an explicit prior, via a system prompt, do models partially adjust their reasoning strategies and correct some distortions, though not consistently or completely. To study this phenomenon, we introduce a small, expert-curated dataset that explicitly evaluates LLMs on table question answering (TQA) tasks requiring an additional error-correction step prior to analysis. Our results reveal systematic differences in how LLMs ingest and interpret tabular information under distortion, with even SoTA models such as GPT-5.2 model exhibiting a drop of minimum 22% accuracy under distortion. These findings raise important questions for future research, particularly regarding when and how models should autonomously decide to realign tabular inputs, analogous to human behavior, without relying on explicit prompts or tabular data pre-processing.

cs.AI↗

arXiv2025

Exploration vs. Fixation: Scaffolding Divergent and Convergent Thinking for Human-AI Co-Creation with Generative Models

Generative AI has democratized content creation, but popular chatbot-based interfaces often prioritize execution, generating fully rendered artifacts right away. This issue can lead to premature convergence and design fixation, where users are being anchored to initial outputs. Recent works have proposed new interfaces to address this issue by supporting exploration, though typically constrained to be semantically close to a user's initial task framing, potentially limiting the creativity of the outcomes. We examine an approach grounded in the Geneplore model of creative cognition and instantiate it in a human-AI co-creation system, HAICo, for creative image generation. HAICo explicitly structures the creative process into two switchable modes: DIVERGENT mode scaffolds the broad exploration of remote conceptual ideas; CONVERGENT mode supports a targeted refinement of selected ideas. Through a within-subjects study (N=24) on a poster image creation task, we demonstrate that HAICo outperforms ChatGPT across multiple dimensions of creativity and usability. Our results highlight the critical need to shift from pure execution-focused chatbots to scaffolded co-creation systems that actively guide exploration and foster the creative process.

cs.HC↗

arXiv2025

IndiMathBench: Autoformalizing Mathematical Reasoning Problems with a Human Touch

Reliable autoformalization remains challenging even in the era of large language models (LLMs). The scarcity of high-quality training data is a major bottleneck. Expert annotation requires substantial time and deep expertise in both mathematics and theorem proving. We introduce IndiMathBench, a human-verified benchmark designed to evaluate mathematical theorem proving, curated using an AI-powered human-assisted pipeline for formalizing natural language problems in Lean. IndiMathBench is composed of 312 formal Lean 4 theorems paired with their corresponding informal problem statements, sourced from Indian Mathematics Olympiads. Through category-based retrieval, iterative compiler feedback, and multi-model ensembles, our pipeline generates candidate formalizations that experts efficiently validate via an interactive dashboard with automated quality summaries. Evaluation across multiple frontier models demonstrates that autoformalization remains challenging, with substantial gaps between syntactic validity and semantic correctness, while theorem proving success rates remain low even with iterative refinement, demonstrating that \benchmark~presents a challenging testbed for mathematical reasoning. IndiMathBench is available at https://github.com/prmbiy/IndiMathBench.

cs.AI↗

arXiv2025

Improving Language Agents through BREW: Bootstrapping expeRientially-learned Environmental knoWledge

Large Language Model (LLM)-based agents are increasingly capable of complex, multi-step tasks such as GUI automation, tool use, and data manipulation, yet they cannot learn from experience: each new session rediscovers solutions from scratch. We introduce BREW (Bootstrapping expeRientially-learned Environmental knoWledge), a framework that distills an agent's past interaction trajectories into a structured, retrievable knowledge base (KB) of natural-language recipes, concept-level procedural documents that capture what to do, when it applies, and what to watch out for. Drawing on the principle of library learning from program synthesis, BREW decomposes agent memory into modular, concept-localized documents and formalizes KB construction as a state-space search problem. To navigate this space, we introduce Expand-and-Gather Monte Carlo Tree Search (EG-MCTS), a reward-guided algorithm that jointly optimizes recipe accuracy and retrievability across parallel, per-concept search trees. We further adapt hindsight relabeling to convert near-miss trajectories into positive demonstrations, surfacing latent agent competencies as reusable knowledge. On three domain-grounded benchmarks, OSWorld, tau^2-Bench, and SpreadSheetBench, BREW achieves 10-20% gains in task success and 10-15% fewer execution steps over base agents, while consistently outperforming existing memory-augmented baselines that can degrade below memoryless performance. The resulting KB is inspectable, modular, and extensible, providing a transparent and controllable substrate for agent optimization.

cs.AI↗

arXiv2025

Training Emergent Joint Associations: A Reinforcement Learning Approach to Creative Thinking in Language Models

Associative thinking--the ability to connect seemingly unrelated ideas--is a foundational element of human creativity and problem-solving. This paper explores whether reinforcement learning (RL) guided by associative thinking principles can enhance a model's performance across diverse generative tasks, including story writing, code generation, and chart creation. We introduce a reinforcement learning framework that uses a prompt-based evaluation mechanism, incorporating established divergent thinking metrics from creativity research. A base language model is fine-tuned using this framework to reward outputs demonstrating higher novelty through higher degrees of conceptual connectivity. Interestingly, the experimental results suggest that RL-based associative thinking-trained models not only generate more original and coherent stories but also exhibit improved abstraction and flexibility in tasks such as programming and data visualization. Our findings provide initial evidence that modeling cognitive creativity principles through reinforcement learning can yield more adaptive and generative AI.

cs.AI↗

arXiv2025

Scaling Competence, Shrinking Reasoning: Cognitive Signatures in Language Model Learning

We analyze reasoning in language models during task-specific fine-tuning and draws parallel between reasoning tokens--intermediate steps generated while solving problem and the human working memory. Drawing from cognitive science, we align training dynamics with the Four Stages of Competence: models initially produce incorrect outputs without reasoning, then begin reasoning (but still fail), eventually reason effectively, and finally solve tasks without explicit reasoning. We find that reasoning token length expands as performance improves, peaks at the stage of conscious competence, then declines as the model internalizes the task. Notably, after training, models retain performance even when reasoning is removed--suggesting it scaffolded learning but is no longer needed. This progression offers actionable insights: reasoning token dynamics can serve as a signal for diagnosing training stage, identifying convergence, and guiding early stopping. We propose metrics to track this trajectory and argue that reasoning behavior is valuable for understanding and optimizing reasoning model training.

cs.CL↗

arXiv2025

SWE-Sharp-Bench: A Reproducible Benchmark for C# Software Engineering Tasks

AI coding agents have shown great progress on Python software engineering benchmarks like SWE-Bench, and for other languages like Java and C in benchmarks like Multi-SWE-Bench. However, C# -- a prominent enterprise language ranking #5 in the TIOBE index -- remains absent from such benchmarks. We introduce SWE-Sharp-Bench, a reproducible software engineering benchmark for C# featuring 150 instances from 17 repositories. Evaluating identical model-agent configurations across languages reveals a significant performance gap: while 70% of Python tasks in SWE-Bench Verified are solved, only 40% of our C# tasks are resolved. We open-source SWE-Sharp-Bench and our entire curation pipeline.

cs.SE↗

arXiv2025

Closing the Loop: An Instructor-in-the-Loop AI Assistance System for Supporting Student Help-Seeking in Programming Education

Timely and high-quality feedback is essential for effective learning in programming courses; yet, providing such support at scale remains a challenge. While AI-based systems offer scalable and immediate help, their responses can occasionally be inaccurate or insufficient. Human instructors, in contrast, may bring more valuable expertise but are limited in time and availability. To address these limitations, we present a hybrid help framework that integrates AI-generated hints with an escalation mechanism, allowing students to request feedback from instructors when AI support falls short. This design leverages the strengths of AI for scale and responsiveness while reserving instructor effort for moments of greatest need. We deployed this tool in a data science programming course with 82 students. We observe that out of the total 673 AI-generated hints, students rated 146 (22%) as unhelpful. Among those, only 16 (11%) of the cases were escalated to the instructors. A qualitative investigation of instructor responses showed that those feedback instances were incorrect or insufficient roughly half of the time. This finding suggests that when AI support fails, even instructors with expertise may need to pay greater attention to avoid making mistakes. We will publicly release the tool for broader adoption and enable further studies in other classrooms. Our work contributes a practical approach to scaling high-quality support and informs future efforts to effectively integrate AI and humans in education.

cs.CY↗

arXiv2025

ConDABench: Interactive Evaluation of Language Models for Data Analysis

Real-world data analysis tasks often come with under-specified goals and unclean data. User interaction is necessary to understand and disambiguate a user's intent, and hence, essential to solving these complex tasks. Existing benchmarks for evaluating LLMs on data analysis tasks do not capture these complexities or provide first-class support for interactivity. We introduce ConDABench, a framework for generating conversational data analysis (ConDA) benchmarks and evaluating external tools on the generated benchmarks. \bench consists of (a) a multi-agent workflow for generating realistic benchmarks from articles describing insights gained from public datasets, (b) 1,420 ConDA problems generated using this workflow, and (c) an evaluation harness that, for the first time, makes it possible to systematically evaluate conversational data analysis tools on the generated ConDA problems. Evaluation of state-of-the-art LLMs on the benchmarks reveals that while the new generation of models are better at solving more instances, they are not necessarily better at solving tasks that require sustained, long-form engagement. ConDABench is an avenue for model builders to measure progress towards truly collaborative models that can complete complex interactive tasks.

cs.CL↗

arXiv2025

Do Code Models Suffer from the Dunning-Kruger Effect?

As artificial intelligence systems increasingly collaborate with humans in creative and technical domains, questions arise about the cognitive boundaries and biases that shape our shared agency. This paper investigates the Dunning-Kruger Effect (DKE), the tendency for those with limited competence to overestimate their abilities in state-of-the-art LLMs in coding tasks. By analyzing model confidence and performance across a diverse set of programming languages, we reveal that AI models mirror human patterns of overconfidence, especially in unfamiliar or low-resource domains. Our experiments demonstrate that less competent models and those operating in rare programming languages exhibit stronger DKE-like bias, suggesting that the strength of the bias is proportionate to the competence of the models.

cs.AI↗

arXiv2025

The Invisible Mentor: Inferring User Actions from Screen Recordings to Recommend Better Workflows

Many users struggle to notice when a more efficient workflow exists in feature-rich tools like Excel. Existing AI assistants offer help only after users describe their goals or problems, which can be effortful and imprecise. We present InvisibleMentor, a system that turns screen recordings of task completion into vision-grounded reflections on tasks. It detects issues such as repetitive edits and recommends more efficient alternatives based on observed behavior. Unlike prior systems that rely on logs, APIs, or user prompts, InvisibleMentor operates directly on screen recordings. It uses a two-stage pipeline: a vision-language model reconstructs actions and context, and a language model generates structured, high-fidelity suggestions. In evaluation, InvisibleMentor accurately identified inefficient workflows, and participants found its suggestions more actionable, tailored, and more helpful for learning and improvement compared to a prompt-based spreadsheet assistant.

cs.HC↗

arXiv2025

Collaboration and Conflict between Humans and Language Models through the Lens of Game Theory

Language models are increasingly deployed in interactive online environments, from personal chat assistants to domain-specific agents, raising questions about their cooperative and competitive behavior in multi-party settings. While prior work has examined language model decision-making in isolated or short-term game-theoretic contexts, these studies often neglect long-horizon interactions, human-model collaboration, and the evolution of behavioral patterns over time. In this paper, we investigate the dynamics of language model behavior in the iterated prisoner's dilemma (IPD), a classical framework for studying cooperation and conflict. We pit model-based agents against a suite of 240 well-established classical strategies in an Axelrod-style tournament and find that language models achieve performance on par with, and in some cases exceeding, the best-known classical strategies. Behavioral analysis reveals that language models exhibit key properties associated with strong cooperative strategies - niceness, provocability, and generosity while also demonstrating rapid adaptability to changes in opponent strategy mid-game. In controlled "strategy switch" experiments, language models detect and respond to shifts within only a few rounds, rivaling or surpassing human adaptability. These results provide the first systematic characterization of long-term cooperative behaviors in language model agents, offering a foundation for future research into their role in more complex, mixed human-AI social environments.

cs.AI↗