学習・ファインチューニング A
114 件中 61〜90 件目を表示
-
SpecFirst: Behavioral Specification Elicitation as a First-Class Step in Agent-Based Program Synthesis from Scratch
-
Anatomy Contextualized Adaption of CT Foundation Models
-
MindForge: Teaching Small Language Models Whole-Life-Cycle Software Engineering via Source-Free Program Synthesis
-
InferScale: GPU-Native KV Injection for Personalized LLM Serving
-
On-Policy Distillation for LLM Safety: A Routing Approach to Template-Robust Realignment
-
ScratchSim: A Procedural Synthetic Data Pipeline for Surface Scratch Detection
-
Lottery Tickets Are Not Deployment Tickets
-
KubernetesはAIを動かすプラットフォームに。横浜でKubeCon＋CloudNativeCon Japan 2026が開幕KubeCon Japan 2026 が横浜で開幕、K8s が AI 基盤にクラウドネイティブ技術の国内最大級イベント『KubeCon＋CloudNativeCon Japan 2026』が2026年7月29日、パシフィコ横浜で開幕した。KubernetesをAIワークロードを動かす基盤と位置づける潮流が主題として掲げられている。※抜粋は冒頭で途切れており、基調講演やセッションの具体的内容・登壇者は確認できない。
-
Foundation Models for Face Presentation Attack Detection: A Unified Linear-Probing Benchmark
-
Latent-IM: Latent Interaction Management for Speech LLMs
-
Temporally Centered SIGReg Improves Multi-Task LeWorldModel Learning: From Analysis to Method
-
BioVLN: A Simulation Platform for Visual Language Navigation in Biomedical Laboratories
-
DIRECT: Direct Decoding for Efficient and Aligned Sequence Labeling with Large Language Models
-
SERPO: Self-Evolving Rubric Policy Optimization for Open-Ended Test-Time Reinforcement Learning
-
Amortized Moment Matching for Visual Generation
-
Budget-Aware LLM Discovery via Cost-Calibrated Frontier Utility
-
When Does Span-Guided Detoxification Help? Human Preferences and Evaluator Diagnostics in a Controlled Comparison
-
Enhancing Generative Information Extraction with Two-step Validation: A Product Attribute Use Case
-
FARI: Robust One-Step Inversion for Watermarking in Diffusion Models
-
Constitutional Midtraining: Content Presence Drives Alignment Gains
-
Filesystem-Based Memory for LLM Agents: Organization, Evolution, and Sustainability
-
FedWeave: Rethinking the Unit of Specialization in Heterogeneous Federated MoE-LoRA
-
Prosody-driven Jailbreaks in Audio LLMs: A Controlled Study and Mechanistic Analysis
-
Misalignment Has a Personality: A Big Five Account of Emergent Misalignment
-
Diagnosing Fine-Grained Inconsistency Classification in Financial Disclosure Text
-
Dissecting Sensitivity to Training Language in Self-Supervised Speech Learning Using Neural Audio Codec Tokens
-
Quoting Akshat BubnaModal CTO、顧客の無認証エンドポイント悪用を証言――基盤は無侵害と強調Simon Willison氏がModalのCTO、Akshat Bubna氏のReutersでの発言を引用。あるModal顧客が無認証のエンドポイントを公開しており、誰でもそのサンドボックスでコード実行できる状態だったため、これが「rogue agent（暴走エージェント）」に悪用されたと説明。ただしModalのプラットフォームや分離機構自体が侵害されたわけではないと強調している。事案の詳細な背景はexcerpt外。
-
Spend Experts Where You Are Unsure: Confidence-Adaptive Routing for Mixture-of-Experts LoRA
-
Falling Behind Drives Unsafe Development in an Idealised AI Race Experiment
-
CHARM: A Multimodal Graph Foundation Model with Hierarchical Context Modeling for Zero-Shot Transfer