開発者ツール
B
326 件中 181〜210 件目を表示
-
その.envのAPIキーが、AIエージェントを「内通者」に変える――“人間前提のやり方”は破綻した.envのAPIキーが、AIエージェントを「内通者」に変えるAIエージェントの活用が広がる裏側で、APIキーやトークンなどの認証情報が急増している。人間が使う前提で設計されてきた従来の管理方法は、常時稼働して自律的に外部へアクセスするエージェントの登場によって限界を迎えつつあり、.envに置かれた一本の鍵がエージェントを内通者に変えうると警告する。
-
Stellar Colosseum: A Many-Agent Harness for Long-Horizon Research in Mathematics and Theoretical Computer Science
-
A Chosen Future Can Still Be Rewritten: Causal Writability in Video Models
-
Disentangling Representation Evolution in Transformers through Directional Decomposition
-
Discovery Foundation Models: Toward Open-Ended Discovery Intelligence
-
Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States
-
Verifiable by Construction: Claim-Level Evaluation of Verbatim Citation in Clinical Question Answering
-
Vulnerability Localization Benchmark: Measuring Agentic Security Analysis at Repository Scale
-
HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses
-
Recurrent GraphNeural NetworkswithSet-BasedAggregation
-
Pilot Early, Commit Late: A Real-Options Model of Enterprise AI Adoption under Rapid Technological Progress
-
Bridging Control, Inference, Transport, and Thermodynamics: From Theory to Applications in Learning
-
LLM-Based Schema-Aware Split Learning for Privacy-Preserving Mental Distress Prediction Across Heterogeneous Surveys
-
LongAgent: History-Guided Agentic Search for Longitudinal Outcome Prediction
-
Task-Directed Residual AddUNet:Perfect-Reconstruction Routing for Full-Rate Representations
-
K-Bench: a clinically calibrated benchmark for evaluating large language models in high-risk mental health conversations
-
Learning to Coach for Experiential Learning
-
Proportional-Fair Resource Allocation and Dual-Threshold Early-Exit Inference for Secure Cooperative Multi-Layer Edge Intelligence
-
Accelerating Dropless MoE Training in JAX with NVIDIA Transformer EngineNVIDIA、Transformer Engine で JAX の MoE 学習を 10.4 倍高速化NVIDIA は JAX 向け Transformer Engine の MoE 最適化を解説。dropless MoE の可変長トークンを扱う grouped GEMM、MXFP8 量子化、NCCL EP による dispatch/combine 融合などで、DeepSeek-V3 671B の学習を GB200 上で 103→1,068 TFLOPS/GPU と 10.4 倍に高速化し、1,024 GPU で 97% のスケーリング効率を達成したと報告。NGC MaxText コンテナで再現可能。
-
Per-Matrix Optimality Is Not Enough: Three-Level Optimization for Low-Rank LLM Compression
-
CiteGuard-RAG: A Validation-Centered AI System for Evidence-Grounded Question Answering
-
Sharp Rates and a One-Line Correction for Spectral Representation Learning
-
AlgoEvo: Self-Evolving Agentic Search for Automated Algorithm Discovery
-
Accelerating Transfer-Learning-Based Autotuning with Predictive LLVM IR Performance Ranking
-
Delegating Authorization to Misaligned Agents: Coalitional Alignment and Safe Control
-
When Should a World Model Move? Loss-Conditioned State Execution
-
Navigating Sparse Evidence: Agentic Visual RAG via Explicit Context Selection and Consolidation
-
MoveBench: A Benchmark for Global-Scale Wildlife Movement Forecasting
-
EvoOntology: A Self-Evolving Ontology Layer for Data Agents
-
Transfer Learning for Socioeconomic Estimation in Forced-Displacement Settings