We consider learning a predictive model to be subsequently used for a given downstream task (described by an algorithm) that requires access to the model evaluation. This task need not be prediction, and this situation is frequently encountered in machine-learning-augmented scientific computing. We show that this setting differs from classical supervised learning, and in general it cannot be solved by minimizing the mean square error of the model predictions as is frequently performed in the literature. Instead, we find that the maximum prediction error on the support of the downstream task algorithm can serve as an effective estimate for the subsequent task performance. With this insight, we formulate a task-specific supervised learning problem based on the given sampling measure, whose solution serves as a reliable surrogate model for the downstream task. Then, we discretize the empirical risk based on training data, and develop an iterative algorithm to solve the task-specific supervised learning problem. Three illustrative numerical examples on trajectory prediction, optimal control and minimum energy path computation demonstrate the effectiveness of the approach.

Annotating data is a time-consuming and costly task, but it is inherently required for supervised machine learning. Active Learning (AL) is an established method that minimizes human labeling effort by iteratively selecting the most informative unlabeled samples for expert annotation, thereby improving the overall classification performance. Even though AL has been known for decades, AL is still rarely used in real-world applications. As indicated in the two community web surveys among the NLP community about AL, two main reasons continue to hold practitioners back from using AL: first, the complexity of setting AL up, and second, a lack of trust in its effectiveness. We hypothesize that both reasons share the same culprit: the large hyperparameter space of AL. This mostly unexplored hyperparameter space often leads to misleading and irreproducible AL experiment results. In this study, we first compiled a large hyperparameter grid of over 4.6 million hyperparameter combinations, second, recorded the performance of all combinations in the so-far biggest conducted AL study, and third, analyzed the impact of each hyperparameter in the experiment results. In the end, we give recommendations about the influence of each hyperparameter, demonstrate the surprising influence of the concrete AL strategy implementation, and outline an experimental study design for reproducible AL experiments with minimal computational effort, thus contributing to more reproducible and trustworthy AL research in the future.

Visuals are valuable tools for teaching math word problems (MWPs), helping young learners interpret textual descriptions into mathematical expressions before solving them. However, creating such visuals is labor-intensive and there is a lack of automated methods to support this process. In this paper, we present Math2Visual, an automatic framework for generating pedagogically meaningful visuals from MWP text descriptions. Math2Visual leverages a pre-defined visual language and a design space grounded in interviews with math teachers, to illustrate the core mathematical relationships in MWPs. Using Math2Visual, we construct an annotated dataset of 1,903 visuals and evaluate Text-to-Image (TTI) models for their ability to generate visuals that align with our design. We further fine-tune several TTI models with our dataset, demonstrating improvements in educational visual generation. Our work establishes a new benchmark for automated generation of pedagogically meaningful visuals and offers insights into key challenges in producing multimodal educational content, such as the misrepresentation of mathematical relationships and the omission of essential visual elements.

Visually impaired people could benefit from Visual Question Answering (VQA) systems to interpret text in their surroundings. However, current models often struggle with recognizing text in the photos taken by this population. Through in-depth interviews with visually impaired individuals, we identified common framing conventions that frequently result in misaligned text. Existing VQA benchmarks primarily feature well-oriented text captured by sighted users, under-representing these challenges. To address this gap, we introduce ROtated SAm-pling ( ROSA), a decoding strategy that enhances VQA performance in text-rich images with incorrectly oriented text. ROSA outperforms Greedy decoding by 11.7 absolute points in the best-performing model.

Recent advances in knowledge representation learning (KRL) highlight the urgent necessity to unify symbolic knowledge graphs (KGs) with language models (LMs) for richer semantic understanding. However, existing approaches typically prioritize either graph structure or textual semantics, leaving a gap: a unified framework that simultaneously captures global KG connectivity, nuanced linguistic context, and discriminative reasoning semantics. To bridge this gap, we introduce KG-BiLM, a bidirectional LM framework that fuses structural cues from KGs with the semantic expressiveness of generative transformers. KG-BiLM incorporates three key components: (i) Bidirectional Knowledge Attention, which removes the causal mask to enable full interaction among all tokens and entities; (ii) Knowledge-Masked Prediction, which encourages the model to leverage both local semantic contexts and global graph connectivity; and (iii) Contrastive Graph Semantic Aggregation, which preserves KG structure via contrastive alignment of sampled sub-graph representations. Extensive experiments on standard benchmarks demonstrate that KG-BiLM outperforms strong baselines in link prediction, especially on large-scale graphs with complex multi-hop relations - validating its effectiveness in unifying structural information and textual semantics.

Tokenization serves as a foundational step for Large Language Models (LLMs) to process text. In new domains or languages, the inefficiency of the tokenizer will slow down the training and generation of LLM. The mismatch in vocabulary also hinders deep knowledge transfer between LLMs like token-level distillation. To mitigate this gap, we propose an efficient method named TokAlign to replace the vocabulary of LLM from the token co-occurrences view, and further transfer the token-level knowledge between models. It first aligns the source vocabulary to the target one by learning a one-to-one mapping matrix for token IDs. Model parameters, including embeddings, are rearranged and progressively fine-tuned for the new vocabulary. Our method significantly improves multilingual text compression rates and vocabulary initialization for LLMs, decreasing the perplexity from 3.4$\text{e}^2$ of strong baseline methods to 1.2$\text{e}^2$ after initialization. Experimental results on models across multiple parameter scales demonstrate the effectiveness and generalization of TokAlign, which costs as few as 5k steps to restore the performance of the vanilla model. After unifying vocabularies between LLMs, token-level distillation can remarkably boost (+4.4% than sentence-level distillation) the base model, costing only 235M tokens.

The quantification of social science remains a longstanding challenge, largely due to the philosophical nature of its foundational theories. Although quantum computing has advanced rapidly in recent years, its relevance to social theory remains underexplored. Most existing research focuses on micro-cognitive models or philosophical analogies, leaving a gap in system-level applications of quantum principles to the analysis of social systems. This study addresses that gap by proposing a theoretical and computational framework that combines quantum mechanics with Generative AI to simulate the emergence and evolution of social norms. Drawing on core quantum concepts--such as superposition, entanglement, and probabilistic measurement--this research models society as a dynamic, uncertain system and sets up five ideal-type experiments. These scenarios are simulated using 25 generative agents, each assigned evolving roles as compliers, resistors, or enforcers. Within a simulated environment monitored by a central observer (the Watcher), agents interact, respond to surveillance, and adapt to periodic normative disruptions. These interactions allow the system to self-organize under external stress and reveal emergent patterns. Key findings show that quantum principles, when integrated with generative AI, enable the modeling of uncertainty, emergence, and interdependence in complex social systems. Simulations reveal patterns including convergence toward normative order, the spread of resistance, and the spontaneous emergence of new equilibria in social rules. In conclusion, this study introduces a novel computational lens that lays the groundwork for a quantum-informed social theory. It offers interdisciplinary insights into how society can be understood not just as a structure to observe but as a dynamic system to simulate and redesign through quantum technologies.

Content creation has dramatically progressed with the rapid advancement of large language models like ChatGPT and Claude. While this progress has greatly enhanced various aspects of life and work, it has also negatively affected certain areas of society. A recent survey revealed that nearly 30% of college students use generative AI to help write academic papers and reports. Most countermeasures treat the detection of AI-generated text as a binary classification task and thus lack robustness. This approach overlooks human involvement in the generation of content even though human-machine collaboration is becoming mainstream. Besides generating entire texts, people may use machines to complete or revise texts. Such human involvement varies case by case, which makes binary classification a less than satisfactory approach. We refer to this situation as participation detection obfuscation. We propose using BERTScore as a metric to measure human involvement in the generation process and a multi-task RoBERTa-based regressor trained on a token classification task to address this problem. To evaluate the effectiveness of this approach, we simulated academic-based scenarios and created a continuous dataset reflecting various levels of human involvement. All of the existing detectors we examined failed to detect the level of human involvement on this dataset. Our method, however, succeeded (F1 score of 0.9423 and a regressor mean squared error of 0.004). Moreover, it demonstrated some generalizability across generative models. Our code is available at https://github.com/gyc-nii/CAS-CS-and-dual-head-detector

Large Language Models (LLMs) often require domain-specific fine-tuning to address targeted tasks, which risks degrading their general capabilities. Maintaining a balance between domain-specific enhancements and general model utility is a key challenge. This paper proposes a novel approach named APT (Weakness Case Acquisition and Iterative Preference Training) to enhance domain-specific performance with self-generated dis-preferred weakness data (bad cases and similar cases). APT uniquely focuses on training the model using only those samples where errors occur, alongside a small, similar set of samples retrieved for this purpose. This targeted training minimizes interference with the model's existing knowledge base, effectively retaining generic capabilities. Experimental results on the LLama-2 and Mistral-V0.3 models across various benchmarks demonstrate that APT ensures no reduction in generic capacity and achieves superior performance on downstream tasks compared to various existing methods. This validates our method as an effective strategy for enhancing domain-specific capabilities without sacrificing the model's broader applicability.

Agentic workflows -- where multiple large language model (LLM) instances interact to solve tasks -- are increasingly built on feedback mechanisms, where one model evaluates and critiques another. Despite the promise of feedback-driven improvement, the stability of agentic workflows rests on the reliability of the judge. However, judges may hallucinate information, exhibit bias, or act adversarially -- introducing critical vulnerabilities into the workflow. In this work, we present a systematic analysis of agentic workflows under deceptive or misleading feedback. We introduce a two-dimensional framework for analyzing judge behavior, along axes of intent (from constructive to malicious) and knowledge (from parametric-only to retrieval-augmented systems). Using this taxonomy, we construct a suite of judge behaviors and develop WAFER-QA, a new benchmark with critiques grounded in retrieved web evidence to evaluate robustness of agentic workflows against factually supported adversarial feedback. We reveal that even strongest agents are vulnerable to persuasive yet flawed critiques -- often switching correct answers after a single round of misleading feedback. Taking a step further, we study how model predictions evolve over multiple rounds of interaction, revealing distinct behavioral patterns between reasoning and non-reasoning models. Our findings highlight fundamental vulnerabilities in feedback-based workflows and offer guidance for building more robust agentic systems.