Kangqi Ni

Jinbiao Wei, Qianran Ma, Yilun Zhao et al.

We present OpenComputer, a verifier-grounded framework for constructing verifiable software worlds for computer-use agents. OpenComputer integrates four components: (1) app-specific state verifiers that expose structured inspection endpoints over real applications, (2) a self-evolving verification layer that improves verifier reliability using execution-grounded feedback, (3) a task-generation pipeline that synthesizes realistic and machine-checkable desktop tasks, and (4) an evaluation harness that records full trajectories and computes auditable partial-credit rewards. In its current form, OpenComputer covers 33 desktop applications and 1,000 finalized tasks spanning browsers, office tools, creative software, development environments, file managers, and communication applications. Experiments show that OpenComputer's hard-coded verifiers align more closely with human adjudication than LLM-as-judge evaluation, especially when success depends on fine-grained application state. Frontier agents struggle with end-to-end completion despite partial progress, and open-source models exhibit sharp drops from their OSWorld-Verified scores, exposing a persistent gap in robust computer automation.

Jinbiao Wei, Yilun Zhao, Kangqi Ni et al.

End-to-end GUI agents for real desktop environments require large amounts of high-quality interaction data, yet collecting human demonstrations is expensive and existing synthetic pipelines often suffer from limited task diversity or noisy, goal-drifting trajectories. We present a trajectory expansion framework Anchor that bootstraps scalable desktop supervision from a small set of verified seed demonstrations. Starting from each seed, we identify branch points that correspond to meaningful state changes and propose new, state-grounded task variants conditioned on the current GUI context. An executing agent then follows the proposed instructions to generate new trajectories, while a verifier enforces task completion via state-aware checks and trajectory-level consistency. To improve supervision quality, we further apply task-conditioned step-level filtering to remove ungrounded actions and denoise post-branch segments to maintain coherent intent. Experiments on standard desktop benchmarks, OSWorld and WindowsAgentArena, show that models fine-tuned on our expanded corpus achieve consistent improvements over zero-shot agents and representative synthesis baselines, and generalize across applications and operating systems.

Shashank Sonkar, Kangqi Ni, Sapana Chaudhary et al.

Large Language Models (LLMs), when used in educational settings without pedagogical fine-tuning, often provide immediate answers rather than guiding students through the problem-solving process. This approach falls short of pedagogically best practices and limits their effectiveness as educational tools. We term the objective of training LLMs to emulate effective teaching strategies as `pedagogical alignment.' In this paper, we investigate Learning from Human Preferences (LHP) algorithms to achieve this alignment objective. A key challenge in this process is the scarcity of high-quality preference datasets to guide the alignment. To address this, we propose a novel approach for constructing a large-scale dataset using synthetic data generation techniques, eliminating the need for time-consuming and costly manual annotation. Leveraging this dataset, our experiments with Llama and Mistral models demonstrate that LHP methods outperform standard supervised fine-tuning (SFT), improving pedagogical alignment accuracy by 13.1% and 8.7% respectively. Existing evaluation methods also lack quantitative metrics to adequately measure the pedagogical alignment of LLMs. To address this gap, we propose novel perplexity-based metrics that quantify LLMs' tendency to provide scaffolded guidance versus direct answers, offering a robust measure of pedagogical alignment. Our analysis provides compelling evidence for the superiority of LHP methods over SFT in optimizing LLMs' behavior, underscoring the potential of LHP methods in better aligning LLMs with educational objectives and fostering effective learning experiences. Code and models are available \href{https://github.com/luffycodes/Tutorbot-Spock}{here}.

Shashank Sonkar, Kangqi Ni, Lesa Tran Lu et al.

We introduce a new area of study in the field of educational Natural Language Processing: Automated Long Answer Grading (ALAG). Distinguishing itself from Automated Short Answer Grading (ASAG) and Automated Essay Grading (AEG), ALAG presents unique challenges due to the complexity and multifaceted nature of fact-based long answers. To study ALAG, we introduce RiceChem, a dataset derived from a college chemistry course, featuring real student responses to long-answer questions with an average word count notably higher than typical ASAG datasets. We propose a novel approach to ALAG by formulating it as a rubric entailment problem, employing natural language inference models to verify whether each criterion, represented by a rubric item, is addressed in the student's response. This formulation enables the effective use of MNLI for transfer learning, significantly improving the performance of models on the RiceChem dataset. We demonstrate the importance of rubric-based formulation in ALAG, showcasing its superiority over traditional score-based approaches in capturing the nuances of student responses. We also investigate the performance of models in cold start scenarios, providing valuable insights into the practical deployment considerations in educational settings. Lastly, we benchmark state-of-the-art open-sourced Large Language Models (LLMs) on RiceChem and compare their results to GPT models, highlighting the increased complexity of ALAG compared to ASAG. Despite leveraging the benefits of a rubric-based approach and transfer learning from MNLI, the lower performance of LLMs on RiceChem underscores the significant difficulty posed by the ALAG task. With this work, we offer a fresh perspective on grading long, fact-based answers and introduce a new dataset to stimulate further research in this important area. Code: \url{https://github.com/luffycodes/Automated-Long-Answer-Grading}.

Kangqi Ni, Wenyue Hua, Xiaoxiang Shi et al.

Multi-agent applications often execute complex tasks as multi-stage workflows, where each stage is an LLM call whose output becomes part of context for subsequent steps. Existing LLM serving systems largely assume homogeneous clusters with identical model replicas. This design overlooks the potential of heterogeneous deployments, where models of different sizes and capabilities enable finer trade-offs between latency and performance. However, heterogeneity introduces new challenges in scheduling across models with diverse throughput and performance. We present Chimera, a predictive scheduling system for multi-agent workflow serving on heterogeneous LLM clusters that jointly improves end-to-end latency and task performance. Chimera applies semantic routing to estimate per-model confidence scores for each request, predicts the total remaining output length of the workflow, and estimates per-model congestion using in-flight predicted token volumes for load balancing. We evaluate Chimera on representative agentic workflows for code generation and math reasoning using multiple heterogeneous LLM configurations. Across comparable settings, Chimera traces the best latency-performance frontier, reducing end-to-end latency by 1.2--2.4$\times$ and improving task performance by 8.0-9.5 percentage points on average over competitive baselines including vLLM.

Jinbiao Wei, Kangqi Ni, Yilun Zhao et al.

Computer-use agents provide a promising path toward general software automation because they can interact directly with arbitrary graphical user interfaces instead of relying on brittle, application-specific integrations. Despite recent advances in benchmark performance, strong computer-use agents remain expensive and slow in practice, since most systems invoke large multimodal models at nearly every interaction step. We argue that this uniform allocation of compute is fundamentally inefficient for long-horizon GUI tasks. Such trajectories are highly heterogeneous: many steps are routine and can be handled reliably by smaller, cheaper policies, while errors tend to concentrate at a relatively small number of high-risk moments. Across computer-use benchmarks, these failures repeatedly take two forms: progress stalls, where the agent loops, repeats ineffective actions, or fails to make meaningful progress, and silent semantic drift, where the agent continues taking locally plausible actions after already deviating from the user's true goal. To address this inefficiency, we propose an event-driven, step-level cascade for computer-use agents that runs a small policy by default and escalates to a stronger model only when lightweight learned monitors detect elevated risk. Our framework combines two complementary signals: a Stuck Monitor that detects degraded progress from recent reasoning-action history and triggers recovery, and a Milestone Monitor that identifies semantically meaningful checkpoints where sparse verification is most informative for catching drift. This design turns always-on frontier-model inference into adaptive, on-demand compute allocation over the course of an evolving interaction. The framework is modular and deployment-oriented: it can be layered on top of existing computer-use agents without changing the underlying agent architecture or retraining the large model.

Kangqi Ni, Zhen Tan, Zijie Liu et al.

Reinforcement Learning with Verifiable Rewards (RLVR), primarily driven by the Group Relative Policy Optimization (GRPO) algorithm, is a leading approach for enhancing the reasoning abilities of Large Language Models (LLMs). Despite its wide adoption, GRPO's gains are often inconsistent; for instance, a model may show significant improvement in one reasoning domain, like mathematics, yet remain stagnant in another, such as medicine. This inconsistency raises a critical question: under what conditions does GRPO improve reasoning and generalize out-of-distribution (OOD)? We investigate this from a data distribution perspective. We first prove theoretically that GRPO is a conservative reweighting scheme, bounded by the base model's distribution and thus unable to discover completely novel solutions. We further validate this in carefully designed controlled studies by training transformers from scratch, evaluating generalization across reasoning depth, input length, token representation, and compositionality. Our results provide a principled explanation for GRPO's boundaries: OOD improvement emerges only when the target task aligns with the model's pretrained biases, while gains on in-distribution (ID) tasks diminish as performance saturates. This reframes GRPO not as a universal reasoning enhancer but as a tool that sharpens pretraining biases. Our findings motivate future development of algorithms that can expand a model's capabilities beyond its pretraining origin.