Zihan Dong

Jiaqi Liu, Shi Qiu, Mairui Li et al.

Automating scientific discovery requires more than generating papers from ideas. Real research is iterative: hypotheses are challenged from multiple perspectives, experiments fail and inform the next attempt, and lessons accumulate across cycles. Existing autonomous research systems often model this process as a linear pipeline: they rely on single-agent reasoning, stop when execution fails, and do not carry experience across runs. We present AutoResearchClaw, a multi-agent autonomous research pipeline built on five mechanisms: structured multi-agent debate for hypothesis generation and result analysis, a self-healing executor with a \textsc{Pivot}/\textsc{Refine} decision loop that transforms failures into information, verifiable result reporting that prevents fabricated numbers and hallucinated citations, human-in-the-loop collaboration with seven intervention modes spanning full autonomy to step-by-step oversight, and cross-run evolution that converts past mistakes into future safeguards. On ARC-Bench, a 25-topic experiment-stage benchmark, AutoResearchClaw outperforms AI Scientist v2 by 54.7%. A human-in-the-loop ablation across seven intervention modes reveals that precise, targeted collaboration at high-leverage decision points consistently outperforms both full autonomy and exhaustive step-by-step oversight. We position AutoResearchClaw as a research amplifier that augments rather than replaces human scientific judgment. Code is available at https://github.com/aiming-lab/AutoResearchClaw.

Haoxiang Jiang, Zihan Dong, Tianci Liu et al.

Pointwise reward modeling offers critical signals for LLM post-training, yet struggles with absolute scoring in subjective, non-verifiable settings. Rubric-based methods address this by decomposing evaluation into explicit criteria, but existing approaches typically depend on frontier LLMs and suffer from ties caused by hard Boolean aggregation. We present RUBRIC-ARROW, an alternating framework that jointly trains a rubric generator and a rubric-conditioned judge, with its RL stage using only pairwise preference data. Our method couples a probability-based scoring rule that reduces ties with phase-specific preference-based rewards and an alternating GRPO scheme that together train the pointwise evaluator. Extensive experiments show that RUBRIC-ARROW achieves competitive reward-modeling accuracy and yields consistent gains for downstream policy post-training.

Yang Zhou, Can Jin, Zihan Dong et al.

Reinforcement learning improves the reasoning ability of large language models but remains costly and sample-inefficient, as many rollouts provide weak learning signals. Difficulty-aware data selection methods attempt to address this by prioritizing moderately difficult prompts, yet our analysis reveals three limitations: difficulty estimates become inaccurate under policy drift, data selection alone yields limited final-performance gains, and inference efficiency remains largely unchanged. These findings suggest that efficient and effective RL requires more than filtering by difficulty: the policy should learn to solve hard tasks while producing concise responses for easy ones. To this end, we propose **Dare**, a unified framework that co-evolves difficulty estimation with the policy via self-normalized importance sampling, maintains diverse difficulty coverage through a symmetric Beta sampling distribution, and applies tailored training strategies across difficulty tiers with adaptive compute allocation. Extensive experiments across multiple models and domains demonstrate that **Dare** consistently outperforms existing methods in training efficiency, final effectiveness, and inference efficiency, producing more concise responses on easy tasks while improving correctness on hard ones. Code is available at https://github.com/EtaYang10th/DARE.

Yang Zhou, Zihan Dong, Zhenting Wang et al.

Agent skills can remarkably improve task success rates by using human-written procedural documents, but their quality is difficult to assess without environment-grounded verification. Existing skill generation methods heavily rely on preference logs rather than direct environment interaction, often yielding negligible or even degraded gains. We identify that it is a fundamental timing bottleneck: robust skills should be posterior-based, distilled from empirical environment interaction rather than prior plans. In this study, we introduce the Posterior Distillation Index (PDI), a trajectory-level metric that quantifies how well a distilled skill is grounded in the task-environment evidence. To operationalize PDI, we present SPARK (Structured Pipelines for Autonomous Runnable tasKs and sKill generation) for preserving task execution evidence towards full trajectory-level analysis. SPARK generates environment-verified trajectories used to compute PDI, and it applies PDI as an online diagnostic and intervention signal to ensure posterior skill formation. Across 86 runnable tasks, SPARK-generated skills consistently surpass no-skill baselines and outperform human-written skills on student models (inference cost up to 1,000x cheaper than teacher models). These findings show that PDI-guided distillation produces efficient and transferable skills grounded in the task-environment interaction. We release our code at https://github.com/EtaYang10th/spark-skills .

Zihan Dong, ZhengDong Zhang

Bloodstain pattern analysis plays a crucial role in crime scene investigations by providing valuable information through the study of unique blood patterns. Conventional image analysis methods, like Thresholding and Contrast, impose stringent requirements on the image background and is labor-intensive in the context of droplet image segmentation. The Segment Anything Model (SAM), a recently proposed method for extensive image recognition, is yet to be adequately assessed for its accuracy and efficiency on bloodstain image segmentation. This paper explores the application of pre-trained SAM and fine-tuned SAM on bloodstain image segmentation with diverse image backgrounds. Experiment results indicate that both pre-trained and fine-tuned SAM perform the bloodstain image segmentation task with satisfactory accuracy and efficiency, while fine-tuned SAM achieves an overall 2.2\% accuracy improvement than pre-trained SAM and 4.70\% acceleration in terms of speed for image recognition. Analysis of factors that influence bloodstain recognition is carried out. This research demonstrates the potential application of SAM on bloodstain image segmentation, showcasing the effectiveness of Artificial Intelligence application in criminology research. We release all code and demos at \url{https://github.com/Zdong104/Bloodstain_Analysis_Ai_Tool}

Ran Xu, Tianci Liu, Zihan Dong et al.

Standard reward models typically predict scalar scores that fail to capture the multifaceted nature of response quality in non-verifiable domains, such as creative writing or open-ended instruction following. To address this limitation, we propose Rubric-ARM, a framework that jointly optimizes a rubric generator and a judge using reinforcement learning from preference feedback. Unlike existing methods that rely on static rubrics or disjoint training pipelines, our approach treats rubric generation as a latent action learned to maximize judgment accuracy. We introduce an alternating optimization strategy to mitigate the non-stationarity of simultaneous updates, providing theoretical analysis that demonstrates how this schedule reduces gradient variance during training. Extensive experiments show that Rubric-ARM achieves state-of-the-art performance among baselines on multiple benchmarks and significantly improves downstream policy alignment in both offline and online reinforcement learning settings.

Ran Xu, Yuchen Zhuang, Zihan Dong et al. · gatech

Search-augmented LLMs often struggle with complex reasoning tasks due to ineffective multi-hop retrieval and limited reasoning ability. We propose AceSearcher, a cooperative self-play framework that trains a single large language model (LLM) to alternate between two roles: a decomposer that breaks down complex queries and a solver that integrates retrieved contexts for answer generation. AceSearcher couples supervised fine-tuning on a diverse mixture of search, reasoning, and decomposition tasks with reinforcement fine-tuning optimized for final answer accuracy, eliminating the need for intermediate annotations. Extensive experiments on three reasoning-intensive tasks across 10 datasets show that AceSearcher outperforms state-of-the-art baselines, achieving an average exact match improvement of 7.6%. Remarkably, on document-level finance reasoning tasks, AceSearcher-32B matches the performance of the DeepSeek-V3 model using less than 5% of its parameters. Even at smaller scales (1.5B and 8B), AceSearcher often surpasses existing search-augmented LLMs with up to 9x more parameters, highlighting its exceptional efficiency and effectiveness in tackling complex reasoning tasks. Our code will be published at https://github.com/ritaranx/AceSearcher and https://huggingface.co/AceSearcher.

Shuhang Lin, Chuhao Zhou, Xiao Lin et al.

Knowledge Graph Question Answering (KGQA) has shown promise for grounded and interpretable reasoning, yet existing approaches often fail to provide reliable coverage guarantees over retrieved answers. While Conformal Prediction (CP) offers a principled framework for producing prediction sets with statistical guarantees, prior methods suffer from critical limitations in both calibration validity and score discriminability, resulting in violated coverage guarantees and excessively large prediction sets. To address these pitfalls, we propose Conformal Path Reasoning (CPR), a trustworthy KGQA framework with two key innovations. First, we perform query-level conformal calibration over path-level scores, preserving the exchangeability while generating path prediction sets. Second, we introduce the Residual Conformal Value Network (RCVNet), a lightweight module trained via PUCT-guided exploration to learn discriminative path-level nonconformity scores. Experiments on benchmarks show that CPR significantly improves the Empirical Coverage Rate by 34% while reducing average prediction set size by 40% compared to conformal baselines. These results validate the efficacy of CPR in satisfying coverage guarantees with substantially more compact answer sets.

Zekun Wang, Anant Gupta, Zihan Dong et al.

Large language models (LLMs) increasingly receive information as streams of passages, conversations, and long-context workflows. While longer context windows expose more evidence, they do not ensure that useful information is preserved and reused. We study continual context consolidation: writing current context into model weights while limiting interference with previously consolidated information. We propose \textbf{S}elf-\textbf{Co}nsolidating \textbf{L}anguage Models (SCoL), a post-training framework in which, given current context, an LLM learns to generate textual update instructions specifying which of its own Transformer layers should be updated. Because committed updates change the model that later generates future selections, we train SCoL with meta-reinforcement learning over an evolving model state. We instantiate SCoL with supervised QA rewards on SQuAD knowledge incorporation and intrinsic likelihood-based rewards for LongBench v2 long-context consolidation. Across both settings, SCoL improves acquisition and retention over prompting, summarization, batch test-time training, and sequential finetuning baselines. Analysis of learned selection patterns shows that SCoL encourages the LLM to generate sparse update locations that align with layers of high Fisher information, suggesting that the model learns to route plasticity toward loss-sensitive regions while limiting interference. Moreover, SCoL transfers from shorter meta-training streams to longer LongBench v2 streams at evaluation, suggesting that our framework supports scalable streaming consolidation.

Zihan Dong, Zhixian Zhang, Yang Zhou et al.

Evaluating mathematical reasoning in LLMs is constrained by limited benchmark sizes and inherent model stochasticity, yielding high-variance accuracy estimates and unstable rankings across platforms. On difficult problems, an LLM may fail to produce a correct final answer, yet still provide reliable pairwise comparison signals indicating which of two candidate solutions is better. We leverage this observation to design a statistically efficient evaluation framework that combines standard labeled outcomes with pairwise comparison signals obtained by having models judge auxiliary reasoning chains. Treating these comparison signals as control variates, we develop a semiparametric estimator based on the efficient influence function (EIF) for the setting where auxiliary reasoning chains are observed. This yields a one-step estimator that achieves the semiparametric efficiency bound, guarantees strict variance reduction over naive sample averaging, and admits asymptotic normality for principled uncertainty quantification. Across simulations, our one-step estimator substantially improves ranking accuracy, with gains increasing as model output noise grows. Experiments on GPQA Diamond, AIME 2025, and GSM8K further demonstrate more precise performance estimation and more reliable model rankings, especially in small-sample regimes where conventional evaluation is pretty unstable.

Zhengdong Zhang, Zihan Dong, Yang Shi et al.

The rapid evolution of artificial intelligence (AI), specifically large language models (LLMs), has opened opportunities for various educational applications. This paper explored the feasibility of utilizing ChatGPT, one of the most popular LLMs, for automating feedback for Java programming assignments in an introductory computer science (CS1) class. Specifically, this study focused on three questions: 1) To what extent do students view LLM-generated feedback as formative? 2) How do students see the comparative affordances of feedback prompts that include their code, vs. those that exclude it? 3) What enhancements do students suggest for improving AI-generated feedback? To address these questions, we generated automated feedback using the ChatGPT API for four lab assignments in the CS1 class. The survey results revealed that students perceived the feedback as aligning well with formative feedback guidelines established by Shute. Additionally, students showed a clear preference for feedback generated by including the students' code as part of the LLM prompt, and our thematic study indicated that the preference was mainly attributed to the specificity, clarity, and corrective nature of the feedback. Moreover, this study found that students generally expected specific and corrective feedback with sufficient code examples, but had diverged opinions on the tone of the feedback. This study demonstrated that ChatGPT could generate Java programming assignment feedback that students perceived as formative. It also offered insights into the specific improvements that would make the ChatGPT-generated feedback useful for students.

Tianci Liu, Ruirui Li, Zihan Dong et al.

Large language models (LLMs) have achieved remarkable performance on various natural language tasks. However, they are trained on static corpora and their knowledge can become outdated quickly in the fast-changing world. This motivates the development of knowledge editing (KE) to update specific knowledge in LLMs without changing unrelated others or compromising their pre-trained capabilities. Previous efforts sought to update a small amount of parameters of a LLM and proved effective for making selective updates. Nonetheless, the edited LLM often exhibits degraded ability to reason about the new knowledge. In this work, we identify a key issue: heterogeneous token overfitting (HTO), where the LLM overfits different tokens in the provided knowledge at varying rates. To tackle this, we propose OVERTONE, a token-level smoothing method that mitigates HTO by adaptively refining the target distribution. Theoretically, OVERTONE offers better parameter updates with negligible computation overhead. It also induces an implicit DPO but does not require preference data pairs. Extensive experiments across four editing methods, two LLMs, and diverse scenarios demonstrate the effectiveness and versatility of our method.

Zihan Dong, Ruijia Wu, Linjun Zhang

The increasing reliance on human preference feedback to judge AI-generated pseudo labels has created a pressing need for principled, budget-conscious data acquisition strategies. We address the crucial question of how to optimally allocate a fixed annotation budget between ground-truth labels and pairwise preferences in AI. Our solution, grounded in semi-parametric inference, casts the budget allocation problem as a monotone missing data framework. Building on this formulation, we introduce Preference-Calibrated Active Learning (PCAL), a novel method that learns the optimal data acquisition strategy and develops a statistically efficient estimator for functionals of the data distribution. Theoretically, we prove the asymptotic optimality of our PCAL estimator and establish a key robustness guarantee that ensures robust performance even with poorly estimated nuisance models. Our flexible framework applies to a general class of problems, by directly optimizing the estimator's variance instead of requiring a closed-form solution. This work provides a principled and statistically efficient approach for budget-constrained learning in modern AI. Simulations and real-data analysis demonstrate the practical benefits and superior performance of our proposed method.

Zekun Wang, Anant Gupta, Zihan Dong et al.

Catastrophic forgetting remains a central obstacle for continual learning in neural models. Popular approaches -- replay and elastic weight consolidation (EWC) -- have limitations: replay requires a strong generator and is prone to distributional drift, while EWC implicitly assumes a shared optimum across tasks and typically uses a diagonal Fisher approximation. In this work, we study the gradient geometry of diffusion models, which can already produce high-quality replay data. We provide theoretical and empirical evidence that, in the low signal-to-noise ratio (SNR) regime, per-sample gradients become strongly collinear, yielding an empirical Fisher that is effectively rank-1 and aligned with the mean gradient. Leveraging this structure, we propose a rank-1 variant of EWC that is as cheap as the diagonal approximation yet captures the dominant curvature direction. We pair this penalty with a replay-based approach to encourage parameter sharing across tasks while mitigating drift. On class-incremental image generation datasets (MNIST, FashionMNIST, CIFAR-10, ImageNet-1k), our method consistently improves average FID and reduces forgetting relative to replay-only and diagonal-EWC baselines. In particular, forgetting is nearly eliminated on MNIST and FashionMNIST and is roughly halved on ImageNet-1k. These results suggest that diffusion models admit an approximately rank-1 Fisher. With a better Fisher estimate, EWC becomes a strong complement to replay: replay encourages parameter sharing across tasks, while EWC effectively constrains replay-induced drift.

Zihan Dong, Xinyu Fan, Zixiang Tang et al.

Controlling desktop applications via software remains a fundamental yet under-served problem. Existing multi-modal large language models (MLLMs) ingest screenshots and task instructions to generate keystrokes and mouse events, but they suffer from prohibitive inference latency, poor sample efficiency on long-horizon sparse-reward tasks, and infeasible on-device deployment. We introduce a lightweight hierarchical reinforcement learning framework, ComputerAgent, that formulates OS control as a two-level option process (manager and subpolicy), employs a triple-modal state encoder (screenshot, task ID, numeric state) to handle visual and contextual diversity, integrates meta-actions with an early-stop mechanism to reduce wasted interactions, and uses a compact vision backbone plus small policy networks for on-device inference (15M parameters). On a suite of 135 real-world desktop tasks, ComputerAgent attains 92.1% success on simple tasks (<8 steps) and 58.8% on hard tasks (>=8 steps), matching or exceeding 200B-parameter MLLM baselines on simple scenarios while reducing model size by over four orders of magnitude and halving inference time. These results demonstrate that hierarchical RL offers a practical, scalable alternative to monolithic MLLM-based automation for computer control.

Zihan Dong, Xin Zhou, Ryumei Nakada et al.

Network representation learning seeks to embed networks into a low-dimensional space while preserving the structural and semantic properties, thereby facilitating downstream tasks such as classification, trait prediction, edge identification, and community detection. Motivated by challenges in brain connectivity data analysis that is characterized by subject-specific, high-dimensional, and sparse networks that lack node or edge covariates, we propose a novel contrastive learning-based statistical approach for network edge embedding, which we name as Adaptive Contrastive Edge Representation Learning (ACERL). It builds on two key components: contrastive learning of augmented network pairs, and a data-driven adaptive random masking mechanism. We establish the non-asymptotic error bounds, and show that our method achieves the minimax optimal convergence rate for edge representation learning. We further demonstrate the applicability of the learned representation in multiple downstream tasks, including network classification, important edge detection, and community detection, and establish the corresponding theoretical guarantees. We validate our method through both synthetic data and real brain connectivities studies, and show its competitive performance compared to the baseline method of sparse principal components analysis.