推論・効率化 A
158 件中 121〜150 件目を表示
-
Knowledge before Reasoning: EC-Reason-Bench, a Training-Free Diagnostic Benchmark for LLM Enzyme Classification
-
How GPT-5.6 fuses frontier intelligence with frontier efficiencyOpenAI、GPT-5.6がフロンティア級の知能と効率を融合と説明OpenAIは、GPT-5.6がモデル・推論・エージェント型ワークフローの各面で効率を高め、最先端の知能と効率を両立させると説明した。より有用な知能をより低コストで提供することを狙う。
-
Pass the Baton: Trajectory-Relayed On-Policy Distillation
-
$π\mathbf{R}^2$: Reactive Real-time Flow Policies
-
Desktop-Delta Bench: Do Computer-Use Models Understand Desktop GUI Transitions?
-
UniMem: Complementary Episodic-to-Parametric Memory for Boundary-Agnostic Task Streams
-
MDTransformer: A Hardware-Software Co-Design of Mode-Division Photonic Transformer Accelerator with Inverse-Designed Coherent Crossbar
-
Parallel Decoding Distillation for Fast Image and Video Generation
-
MemLens: A Value-Aware Memory Management System with Interactive Analytics for LLM-based Agents
-
Quasi-SVD: Learning a Lie-constrained matrix factorisation for real-time imaging
-
A Cost-Effective Multimodal LLM Reasoning Framework for Question Answering over Irregular Clinical Time Series
-
The OlmoEarth Platform: Geospatial inference at planetary scaleAllen AI、地球規模の地理空間推論基盤「OlmoEarth」を公開Allen Institute for AI (Ai2) が Hugging Face 上で、地球規模の地理空間推論を担う基盤プラットフォーム「OlmoEarth」を公開した内容。衛星・地球観測データを惑星スケールで推論するインフラを扱う。※本文未取得のため、モデル構成・対応タスク・具体的なスケール指標・既存手法との比較は確認不可。タグに moe を含む点から Mixture-of-Experts 構成の可能性がうかがえるが断定はできない。
-
Penelope: Localized Latent Recurrence for Efficient Structured Reasoning
-
SAM3D-Guided Object-Centric Representation Alignment for Vision-Language-Action Models
-
Minimizing Targeted Activations: Input-Only Suppression of Evaluation-Awareness Latents in Large Language Models
-
Stemma: Induced Decision Regions Reveal LLM Provenance
-
VAD to the Bone: Ultra-Tiny Speech Activity Detection for Edge Deployment
-
HiSkill: Empowering LLM Agents with Hierarchical Skill Graphs
-
AngelSpec: Towards Real-World High Performance Inference with Speculative Decoding
-
Prototype Adaptation for Zero-Shot sEMG Movement Classification
-
LFM2.5-Encoders for Fast Long-Context Inference on CPULiquid AI、CPUで高速な長文脈推論を狙う「LFM2.5-Encoders」を公開Liquid AI が Hugging Face 上で、CPU 上での高速な長文脈推論に最適化したエンコーダーモデル群「LFM2.5-Encoders」を公開した。長いコンテキストを扱う埋め込み・検索といった用途で、GPU に頼らない CPU 推論の効率化を狙うとされる。※本文 (excerpt) 未取得のため、モデルサイズ・具体的なベンチマーク値・対応タスク・既存エンコーダーとの比較の詳細は確認不可。タイトルおよび配信元 (Hugging Face / LiquidAI) からの中立要約。
-
Speculate While You Reason: Teaching Agents to Predict Their Next Tool Call via Joint Agent-Speculator RL
-
Detecting CSAM Text-to-Image LoRAs From Weights
-
IRIS: Reusable Identity Representations from Frozen LLMs for Entity Alignment
-
OrthKD: Extracting Generalized Clinical Knowledge from Heterogeneous Teachers for Lightweight Deployment
-
Mind the Missing Split: Resolving Feature Heterogeneity in Swarm Learning with Random Forests
-
Bits and Memories: Measuring Verbatim Extraction Across LLM Quantization
-
Every Time I Hire a Linguist, Inference Costs Go Down: On Linguistic Rules as Effective Prompt Compressors
-
CAST: Game Solvers as Turn-Level Teachers for LLM Agents
-
CoSA: Accelerating Long-Context Inference via Proxy-Kernel Co-Designed Sparse Attention