Current large language models excel at broad, general-purpose tasks, but consistently underperform when exposed to highly specialized domains that require deep cultural, linguistic, and subject-matter expertise. In particular, traditional medical systems such as Ayurveda embody centuries of nuanced textual and clinical knowledge that mainstream LLMs fail to accurately interpret or apply. We introduce AyurParam-2.9B, a domain-specialized, bilingual language model fine-tuned from Param-1-2.9B using an extensive, expertly curated Ayurveda dataset spanning classical texts and clinical guidance. AyurParam's dataset incorporates context-aware, reasoning, and objective-style Q&A in both English and Hindi, with rigorous annotation protocols for factual precision and instructional clarity. Benchmarked on BhashaBench-Ayur, AyurParam not only surpasses all open-source instruction-tuned models in its size class (1.5--3B parameters), but also demonstrates competitive or superior performance compared to much larger models. The results from AyurParam highlight the necessity for authentic domain adaptation and high-quality supervision in delivering reliable, culturally congruent AI for specialized medical knowledge.

We study the problem of learning multi-task, multi-agent policies for cooperative, temporal objectives, under centralized training, decentralized execution. In this setting, using automata to represent tasks enables the decomposition of complex tasks into simpler sub-tasks that can be assigned to agents. However, existing approaches remain sample-inefficient and are limited to the single-task case. In this work, we present Automata-Conditioned Cooperative Multi-Agent Reinforcement Learning (ACC-MARL), a framework for learning task-conditioned, decentralized team policies. We identify the main challenges to ACC-MARL's feasibility in practice, propose solutions, and prove the correctness of our approach. We further show that the value functions of learned policies can be used to assign tasks optimally at test time. Experiments show emergent task-aware, multi-step coordination among agents, e.g., pressing a button to unlock a door, holding the door, and short-circuiting tasks.

Smoothness is known to be crucial for acceleration in offline optimization, and for gradient-variation regret minimization in online learning. Interestingly, these two problems are actually closely connected -- accelerated optimization can be understood through the lens of gradient-variation online learning. In this paper, we investigate online learning with Hölder smooth functions, a general class encompassing both smooth and non-smooth (Lipschitz) functions, and explore its implications for offline optimization. For (strongly) convex online functions, we design the corresponding gradient-variation online learning algorithm whose regret smoothly interpolates between the optimal guarantees in smooth and non-smooth regimes. Notably, our algorithms do not require prior knowledge of the Hölder smoothness parameter, exhibiting strong adaptivity over existing methods. Through online-to-batch conversion, this gradient-variation online adaptivity yields an optimal universal method for stochastic convex optimization under Hölder smoothness. However, achieving universality in offline strongly convex optimization is more challenging. We address this by integrating online adaptivity with a detection-based guess-and-check procedure, which, for the first time, yields a universal offline method that achieves accelerated convergence in the smooth regime while maintaining near-optimal convergence in the non-smooth one.

While existing work focuses primarily on task success, we argue that effective real-world agents require optimizing three dimensions: productivity (task completion), proactivity (asking essential questions), and personalization (adapting to diverse user preferences). We introduce UserVille, an interactive environment with LLM-based user simulators enabling diverse, configurable user preferences. Leveraging UserVille, we introduce PPP, a multi-objective reinforcement learning approach that jointly optimizes all three dimensions: Productivity, Proactivity, and Personalization. Experiments on software engineering and deep research tasks show that agents trained with PPP achieve substantial improvements over strong baselines such as GPT-5 (+21.6 on average), demonstrating the ability to ask strategic clarifying questions, adapt to unseen user preferences, and improve task success through better interaction. This work demonstrates that explicitly optimizing for user-centered interaction is critical for building practical and effective AI agents.

Large language models (LLMs) are increasingly used to simulate humans, with applications ranging from survey prediction to decision-making. However, are LLMs strictly necessary, or can smaller, domain-grounded models suffice? We identify a large class of simulation problems in which individuals make choices among discrete options, where a graph neural network (GNN) can match or surpass strong LLM baselines despite being three orders of magnitude smaller. We introduce Graph-basEd Models for human Simulation (GEMS), which casts discrete choice simulation tasks as a link prediction problem on graphs, leveraging relational knowledge while incorporating language representations only when needed. Evaluations across three key settings on three simulation datasets show that GEMS achieves comparable or better accuracy than LLMs, with far greater efficiency, interpretability, and transparency, highlighting the promise of graph-based modeling as a lightweight alternative to LLMs for human simulation. Our code is available at https://github.com/schang-lab/gems.

Flood insurance is an effective strategy for individuals to mitigate disaster-related losses. However, participation rates among at-risk populations in the United States remain strikingly low. This gap underscores the need to understand and model the behavioral mechanisms underlying insurance decisions. Large language models (LLMs) have recently exhibited human-like intelligence across wide-ranging tasks, offering promising tools for simulating human decision-making. This study constructs a benchmark dataset to capture insurance purchase probabilities across factors. Using this dataset, the capacity of LLMs is evaluated: while LLMs exhibit a qualitative understanding of factors, they fall short in estimating quantitative probabilities. To address this limitation, InsurAgent, an LLM-empowered agent comprising five modules including perception, retrieval, reasoning, action, and memory, is proposed. The retrieval module leverages retrieval-augmented generation (RAG) to ground decisions in empirical survey data, achieving accurate estimation of marginal and bivariate probabilities. The reasoning module leverages LLM common sense to extrapolate beyond survey data, capturing contextual information that is intractable for traditional models. The memory module supports the simulation of temporal decision evolutions, illustrated through a roller coaster life trajectory. Overall, InsurAgent provides a valuable tool for behavioral modeling and policy analysis.

Catastrophic forgetting is one of the fundamental issues of continual learning because neural networks forget the tasks learned previously when trained on new tasks. The proposed framework is a new path-coordinated framework of continual learning that unites the Neural Tangent Kernel (NTK) theory of principled plasticity bounds, statistical validation by Wilson confidence intervals, and evaluation of path quality by the use of multiple metrics. Experimental evaluation shows an average accuracy of 66.7% at the cost of 23.4% catastrophic forgetting on Split-CIFAR10, a huge improvement over the baseline and competitive performance achieved, which is very close to state-of-the-art results. Further, it is found out that NTK condition numbers are predictive indicators of learning capacity limits, showing the existence of a critical threshold at condition number $>10^{11}$. It is interesting to note that the proposed strategy shows a tendency of lowering forgetting as the sequence of tasks progresses (27% to 18%), which is a system stabilization. The framework validates 80% of discovered paths with a rigorous statistical guarantee and maintains 90-97% retention on intermediate tasks. The core capacity limits of the continual learning environment are determined in the analysis, and actionable insights to enhance the adaptive regularization are offered.

Speculative decoding accelerates LLMs by using a lightweight draft model to generate tokens autoregressively before verifying them in parallel with a larger target model. However, determining the optimal number of tokens to draft remains a key challenge limiting the approach's effectiveness. Dynamic speculative decoding aims to intelligently decide how many tokens to draft to achieve maximum speedups. Existing methods often rely on hand-tuned, sensitive thresholds (e.g., token entropy), which are costly to set and generalize poorly across models and domains. We propose TapOut, an online, training-free, plug-and-play algorithm for dynamic speculation policy selection using multi-armed bandits. Our approach employs a meta-algorithm that selects among multiple parameter-free dynamic speculation strategies based on past reward and exploration. We conduct extensive experiments across diverse model pairs and datasets, showing that TapOut achieves competitive or superior speedups compared to well-established dynamic speculation baselines without any hyperparameter tuning.

The debate over whether "thinking machines" could replace human intellectual labor has existed in both public and expert discussions since the mid-twentieth century, when the concept and terminology of Artificial Intelligence (AI) first emerged. For decades, this idea remained largely theoretical. However, with the recent advent of Generative AI - particularly Large Language Models (LLMs) - and the widespread adoption of tools such as ChatGPT, the issue has become a practical reality. Many fields that rely on human intellectual effort are now being reshaped by AI tools that both expand human capabilities and challenge the necessity of certain forms of work once deemed uniquely human but now easily automated. Education, somewhat unexpectedly, faces a pivotal responsibility: to devise long-term strategies for cultivating human skills that will remain relevant in an era of pervasive AI in the intellectual domain. In this context, we identify the limitations of current AI systems - especially those rooted in LLM technology - argue that the fundamental causes of these weaknesses cannot be resolved through existing methods, and propose directions within the constructivist paradigm for transforming education to preserve the long-term advantages of human intelligence over AI tools.

Low-resource countries represent over 90% of maternal deaths, with Pakistan among the top four countries contributing nearly half in 2023. Since these deaths are mostly preventable, large language models (LLMs) can help address this crisis by automating health communication and risk assessment. However, sexual and reproductive health (SRH) communication in conservative contexts often relies on indirect language that obscures meaning, complicating LLM-based interventions. We conduct a two-stage study in Pakistan: (1) analyzing data from clinical observations, interviews, and focus groups with clinicians and patients, and (2) evaluating the interpretive capabilities of five popular LLMs on this data. Our analysis identifies two axes of communication (referential domain and expression approach) and shows LLMs struggle with semantic drift, myths, and polysemy in clinical interactions. We contribute: (1) empirical themes in SRH communication, (2) a categorization framework for indirect communication, (3) evaluation of LLM performance, and (4) design recommendations for culturally-situated SRH communication.