Large Language Models (LLMs) have revolutionized natural language processing, but their varying capabilities and costs pose challenges in practical applications. LLM routing addresses this by dynamically selecting the most suitable LLM for each query/task. Previous approaches treat this as a supervised learning problem, assuming complete knowledge of optimal query-LLM pairings. However, real-world scenarios lack such comprehensive mappings and face evolving user queries. We thus propose to study LLM routing as a contextual bandit problem, enabling adaptive decision-making using bandit feedback without requiring exhaustive inference across all LLMs for all queries (in contrast to supervised routing). To address this problem, we develop a shared embedding space for queries and LLMs, where query and LLM embeddings are aligned to reflect their affinity. This space is initially learned from offline human preference data and refined through online bandit feedback. We instantiate this idea through Preference-prior Informed Linucb fOr adaptive rouTing (PILOT), a novel extension of LinUCB. To handle diverse user budgets for model routing, we introduce an online cost policy modeled as a multi-choice knapsack problem, ensuring resource-efficient routing.

This paper studies bandit convex optimization with constraints, where the learner aims to generate a sequence of decisions under partial information of loss functions such that the cumulative loss is reduced as well as the cumulative constraint violation is simultaneously reduced. We adopt the cumulative \textit{hard} constraint violation as the metric of constraint violation, which is defined by $\sum_{t=1}^{T} \max\{g_t(\boldsymbol{x}_t), 0\}$. Owing to the maximum operator, a strictly feasible solution cannot cancel out the effects of violated constraints compared to the conventional metric known as \textit{long-term} constraints violation. We present a penalty-based proximal gradient descent method that attains a sub-linear growth of both regret and cumulative hard constraint violation, in which the gradient is estimated with a two-point function evaluation. Precisely, our algorithm attains $O(d^2T^{\max\{c,1-c\}})$ regret bounds and $O(d^2T^{1-\frac{c}{2}})$ cumulative hard constraint violation bounds for convex loss functions and time-varying constraints, where $d$ is the dimensionality of the feasible region and $c\in[\frac{1}{2}, 1)$ is a user-determined parameter. We also extend the result for the case where the loss functions are strongly convex and show that both regret and constraint violation bounds can be further reduced.

This paper proposes FieldWorkArena, a benchmark for agentic AI targeting real-world field work. With the recent increase in demand for agentic AI, they are required to monitor and report safety and health incidents, as well as manufacturing-related incidents, that may occur in real-world work environments. Existing agentic AI benchmarks have been limited to evaluating web tasks and are insufficient for evaluating agents in real-world work environments, where complexity increases significantly. In this paper, we define a new action space that agentic AI should possess for real world work environment benchmarks and improve the evaluation function from previous methods to assess the performance of agentic AI in diverse real-world tasks. The dataset consists of videos captured on-site and documents actually used in factories and warehouses, and tasks were created based on interviews with on-site workers and managers. Evaluation results confirmed that performance evaluation considering the characteristics of Multimodal LLM (MLLM) such as GPT-4o is feasible. Additionally, the effectiveness and limitations of the proposed new evaluation method were identified. The complete dataset (HuggingFace) and evaluation program (GitHub) can be downloaded from the following website: this https URL

Neural network potentials (NNPs) offer a powerful alternative to traditional force fields for molecular dynamics (MD) simulations. Accurate and stable MD simulations, crucial for evaluating material properties, require training data encompassing both low-energy stable structures and high-energy structures. Conventional knowledge distillation (KD) methods fine-tune a pre-trained NNP as a teacher model to generate training data for a student model. However, in material-specific models, this fine-tuning process increases energy barriers, making it difficult to create training data containing high-energy structures. To address this, we propose a novel KD framework that leverages a non-fine-tuned, off-the-shelf pre-trained NNP as a teacher. Its gentler energy landscape facilitates the exploration of a wider range of structures, including the high-energy structures crucial for stable MD simulations. Our framework employs a two-stage training process: first, the student NNP is trained with a dataset generated by the off-the-shelf teacher; then, it is fine-tuned with a smaller, high-accuracy density functional theory (DFT) dataset. We demonstrate the effectiveness of our framework by applying it to both organic (polyethylene glycol) and inorganic (L$_{10}$GeP$_{2}$S$_{12}$) materials, achieving comparable or superior accuracy in reproducing physical properties compared to existing methods. Importantly, our method reduces the number of expensive DFT calculations by 10x compared to existing NNP generation methods, without sacrificing accuracy. Furthermore, the resulting student NNP achieves up to 106x speedup in inference compared to the teacher NNP, enabling significantly faster and more efficient MD simulations.

Runtime optimization of the quantum computing within a given computational resource is important to achieve practical quantum advantage. In this paper, we propose a runtime reduction protocol for the lattice surgery, which utilizes the soft information corresponding to the logical measurement error. Our proposal is a simple two-step protocol: operating the lattice surgery with the small number of syndrome measurement cycles, and reexecuting it with full syndrome measurement cycles in cases where the time-like soft information catches logical error symptoms. We firstly discuss basic features of the time-like complementary gap as the concrete example of the time-like soft information based on numerical results. Then, we show that our protocol surpasses the existing runtime reduction protocol called temporally encoded lattice surgery (TELS) for the most cases. In addition, we confirm that the combination of our protocol and the TELS protocol can reduce the runtime further, over 50% in comparison to the naive serial execution of the lattice surgery. The proposed protocol in this paper can be applied to any quantum computing architecture based on the lattice surgery, and we expect that this will be one of the fundamental building blocks of runtime optimization to achieve practical scale quantum computing.

Quantized neural network training optimizes a discrete, non-differentiable objective. The straight-through estimator (STE) enables backpropagation through surrogate gradients and is widely used. While previous studies have primarily focused on the properties of surrogate gradients and their convergence, the influence of quantization hyperparameters, such as bit width and quantization range, on learning dynamics remains largely unexplored. We theoretically show that in the high-dimensional limit, STE dynamics converge to a deterministic ordinary differential equation. This reveals that STE training exhibits a plateau followed by a sharp drop in generalization error, with plateau length depending on the quantization range. A fixed-point analysis quantifies the asymptotic deviation from the unquantized linear model. We also extend analytical techniques for stochastic gradient descent to nonlinear transformations of weights and inputs.

This paper studies a distributionally robust portfolio optimization model with a cardinality constraint for limiting the number of invested assets. We formulate this model as a mixed-integer semidefinite optimization (MISDO) problem by means of the moment-based ambiguity set of probability distributions of asset returns. To exactly solve large-scale problems, we propose a specialized cutting-plane algorithm that is based on bilevel optimization reformulation. We prove the finite convergence of the algorithm. We also apply a matrix completion technique to lower-level SDO problems to make their problem sizes much smaller. Numerical experiments demonstrate that our cutting-plane algorithm is significantly faster than the state-of-the-art MISDO solver SCIP-SDP. We also show that our portfolio optimization model can achieve good investment performance compared with the conventional robust optimization model based on the ellipsoidal uncertainty set.

Proteins are complex biomolecules that perform a variety of crucial functions within living organisms. Designing and generating novel proteins can pave the way for many future synthetic biology applications, including drug discovery. However, it remains a challenging computational task due to the large modeling space of protein structures. In this study, we propose a latent diffusion model that can reduce the complexity of protein modeling while flexibly capturing the distribution of natural protein structures in a condensed latent space. Specifically, we propose an equivariant protein autoencoder that embeds proteins into a latent space and then uses an equivariant diffusion model to learn the distribution of the latent protein representations. Experimental results demonstrate that our method can effectively generate novel protein backbone structures with high designability and efficiency. The code will be made publicly available at this https URL

Simulation of quantum many-body systems is a promising application of quantum computers. However, implementing the time-evolution operator as a quantum circuit efficiently on near-term devices with limited resources is challenging. Standard approaches like Trotterization often require deep circuits, making them impractical. This paper proposes a hybrid quantum-classical algorithm called Local Subspace Variational Quantum Compilation (LSVQC) for compiling the time-evolution operator. The LSVQC uses variational optimization to reproduce the action of the target time-evolution operator within a physically reasonable subspace. Optimization is performed on small local subsystems based on the Lieb-Robinson bound, allowing for cost function evaluation using small-scale quantum devices or classical computers. Numerical simulations on a spin-lattice model and an $\mathit{\text{ab initio}}$ effective model of strongly correlated material Sr$_2$CuO$_3$ demonstrate the algorithm's effectiveness. It is shown that the LSVQC achieves a 95% reduction in circuit depth compared to Trotterization while maintaining accuracy. The subspace restriction also reduces resource requirements and improves accuracy. Furthermore, we estimate the gate count needed to execute the quantum simulations using the LSVQC on near-term quantum computing architectures in the noisy intermediate-scale or early fault-tolerant quantum computing era. Our estimation suggests that the acceptable physical gate error rate for the LSVQC can be significantly larger than for Trotterization.

Training large language models (LLMs) typically involves pre-training on massive corpora, only to restart the process entirely when new data becomes available. A more efficient and resource-conserving approach would be continual pre-training, where models are updated with new data rather than retraining from scratch. However, the introduction of new data often causes distribution shifts, leading to performance degradation on previously learned tasks. In this paper, we take a deeper look at two popular proposals for addressing this distribution shift within the continual learning literature: experience replay and gradient alignment. We consider continual pre-training of models within the Llama family of architectures at a large scale across languages with 100 billion tokens of training data in each language, finding that both replay and gradient alignment lead to more stable learning without forgetting. This conclusion holds both as we vary the model scale and as we vary the number and diversity of tasks. Moreover, we are the first to demonstrate the effectiveness of gradient alignment techniques in the context of LLM pre-training and propose an efficient implementation of meta-experience replay (MER) that imbues experience replay with the benefits of gradient alignment despite negligible compute and memory overhead. Our scaling analysis across model sizes and replay rates indicates that small rates of replaying old examples are definitely a more valuable use of compute than investing in model size, but that it is more compute efficient to scale the size of the model than invest in high rates of replaying old examples.