「SHAP」の検索結果
69 件Chain-of-thought reasoning provides a structured computation between a model's input and final answer. Yet it
A growing body of work establishes that large language models are not mere statistical memorizers, but are cap
Industrial fraud detection often relies on costly expert-crafted features that overlook graph-structured relat
This paper proposes a level-set-based physics-driven neural network solver (LSPDNN) for 3-D electromagnetic in
Industrial process monitoring is fundamental to the safety and economic performance of modern process plants.
On-Policy Distillation (OPD) facilitates the transfer of knowledge from domain expert to student in the post-t
Generative diffusion models have emerged as a class of powerful techniques for various imaging applications, i
Agent harnesses (the system prompt, tool set, execution hooks, and context-management scaffolding around a mod
Long horizon Large Language Model (LLM) agents rely on external memory systems to preserve user preferences an
Mid-training, the stage between pre-training and alignment, is where a model's per-domain data composition is
Analysts in emerging equity markets keep answering the same questions. Did fundamentals match the market's res
Benchmark scores are a central currency in model releases: they inform purchasing decisions, shape public trus
Dynamic layer routing reduces the inference cost of Large Language Models (LLMs) by learning to skip layers fo
We study scheming in LLM agents, in which agents covertly pursue misaligned goals. Our focus is to understand
Recursive self-improvement (RSI) requires a concrete mechanism through which an AI system observes its capabil
NumPy is a widely used Python library for numerical scientific computing, known for its declarative APIs and i
Vision transformers (ViTs) have achieved remarkable generalization across visual domains, yet little is known
Three-dimensional femoral reconstruction from radiographs supports surgical planning, implant sizing, and post
Pancreatic tumor segmentation in 3D CT volumes is challenged by extreme scale variability across both the panc
Reconstructing a three-dimensional left-ventricular (LV) endocardial surface from cardiac magnetic resonance (
The learning-rate schedule is a consequential choice in training deep networks, yet the policies in common use
Input devices for robotic microsurgery are frequently described as preserving the surgeon's trained technique,
Understanding generalization remains a central challenge in machine learning because it requires jointly consi
We convert black-box clinical prediction models for tabular data into standalone nomograms that can be audited
Multi-agent debate, in which several LLMs exchange arguments before answering, is widely assumed to improve an
Generative and agentic AI are reshaping both the production and evaluation of scientific research. These devel
Large language models (LLMs) increasingly shape communication, learning, work, creativity, and decision-making
We introduce FramingQA, a benchmark that measures the model sensitivity to question framing across law, medici
Prompt-based interventions: system prompts, personas, role instructions, reliably reshape what a language mode
Standard Convolutional Neural Networks (CNNs) exhibit severe performance degradation due to a strong inductive
Transformers are a dominant architecture in modern machine learning, powering applications across vision, lang
Graphic designs, such as posters, advertisements, and infographics, are an important medium for communicating
How robot body configuration shapes human intervention during approach remains underexplored. We conducted a w
Quadruped robot locomotion policies are often trained using reinforcement learning, which in turn relies heavi
Generalizable and robust dexterous in-hand manipulation requires a policy to infer object pose, geometry, cont
Large language models act as strategic agents and models of human choice, yet choosing like a strategic agent
Training strategy, namely whether to retrain from scratch or fine-tune from the previous checkpoint, is an ove
Research teams and organizations often explore unfamiliar free-text collections, from survey comments and revi
Vision-language-action (VLA) models adapted through supervised fine-tuning (SFT) inherit a structural asymmetr
Multi-fingered dexterous manipulation remains a frontier for real-world reinforcement learning (RL) due to the
Visuomotor imitation policies can achieve high performance under in-distribution visual conditions yet fail wh
World-action models (WAM) predict future states to guide robot actions, enabling learning from both action-fre
Mixture-of-experts (MoE) architectures increase model capacity by combining a collection of expert predictors
Recognizing specific objects onboarded without a labeled training set recurs across manufacturing and service
Verifying that manufactured batches of milling tools or carbide rotary burrs conform to production order sheet
While traditional stable matching algorithms, such as the Gale-Shapley algorithm, prioritize stability, they m
We consider semi-supervised classification from a partially classified sample arising from a two-component Wei
We develop a framework for mechanism design with AI agents whose alignment (preferences) and capabilities (fea
We study a class of product-reference diffusion algorithms for sampling from a discrete distribution. We show
We study kernel ridge regression under anisotropic Gaussian data, where the input covariance decays as a power
As large language models and increasingly capable AI agents are deployed in high-risk settings, aligning them
Nearest neighbor classification relies fundamentally on how locality is defined, yet conventional $k$-NN impos
Bayesian inference in compound loss models must often be repeated across policies, market scenarios, and prior
In neuroevolution, indirect encoding generates neural network connectivity from a compact genome rather than s
Object classification in event-based computer vision is a task that is attracting considerable research attent
People's trust in AI advice diverges as they use it, deepening for some and eroding for others. We study this
Driver behavior is heterogeneous, context-dependent, and changes over time, and these properties shape the tra
The entropy production rate (EPR) quantifies irreversibility of a nonequilibrium steady state, yet standard fo
In a deliberative poll, once submissions outnumber what anyone will read, some mechanism chooses which argumen
Spatial aliasing occurs when two or more distinct locations produce highly similar place-cell representations,
In over-the-counter corporate bond markets, dealers compete for client trades by quoting bid and ask prices. T
In Shapley-Scarf housing markets, Ma (1994) shows that top trading cycles (TTC) is the unique mechanism satisf
An invariant behavioral profile is the defining vulnerability of traditional honeypot installations: a skilled
We study constrained coalition formation in games induced by friends, enemies, and neutrals, under the two sta
Spiking neural networks (SNNs) offer an energy-efficient alternative to conventional deep neural networks by e
This note aims to serve as an entry point to the literature on learning in games, a topic with significant the
Spiking point cloud networks usually scan space in a fixed, input-agnostic order, which leaves the most distin
この研究では、記憶を維持し更新する能力を強化するため、繰り返し神経ネットワークを用いて記憶の特性を分析しました。記憶を維持するための神経計算の一種として、ダイショビュールノーマリゼーション（Divisive Normal
シミュレーション駆動設計では、高精度なシミュレーションを少なくすることで設計を実現しています。既存の手法では、その問題に取り組むために最適化アルゴリズムが改善されてきましたが、問題の定義自体は検討されていません。この論文