開発者ツール
B
421 件中 211〜240 件目を表示
-
Claude Code burns ~15–20% of your tokens on a 3-word status lineClaude Codeのステータス行が約15〜20%のトークンを消費との指摘HNで共有されたGist。Claude Codeが3語ほどのステータス行の生成に、トークン全体の約15〜20%を使っていると主張している。ページ題名によれば、agent_summaryを無効化するBunのプリロードをMITライセンスで公開している。数値の根拠や検証方法は未確認。
-
IMRouter: Imbalance-aware Temporal Pattern Routing for LLM-based Time Series Forecasting
-
Grounding Multimodal Reasoning with Evidence-Ablated Negatives
-
AgentHeal: Evaluating Goal-Preserving Resource Recovery of Web Agents under Access Barriers
-
Where Position Matters: Profiling-Guided Layer-Wise RoPE Allocation in Transformers
-
Event2Reward: Learning Robotic Rewards from Progress Relations
-
Sparse Autoencoders as Implicit Diversification Regularisers for Cross-Sectional Cryptocurrency Returns
-
Analyzing Condensation on Patterned Surfaces via Deep Learning
-
Zero-Complement Extrapolation: Parameter Recovery and Classification in Severely Imbalanced Logistic Regression under Rank-Deficient Regimes
-
Reasoning or Pattern Exploitation? Mechanistic Insights from RL-Trained Language Models
-
EmbodiExp: Compositional Experience Transfer for Embodied Agents
-
ACAM-J-MARS: Multi-Atlas Riemannian Spectral Representations for Decoding Advanced Meditation States from fMRI
-
Learning Erasing Equal-Length Relational Pattern Languages in the Limit from Positive Data over Binary Alphabets
-
DICE: Decentralized Incentive-driven Federated Learning for Efficient & Scalable Customization of Pretrained Models in Web3
-
Clean Sheets and Checkpoints: Evidence Complexity for Certifying Sequential Agent Risk
-
Generating Critic-Guided Synthetic Experience in Latent World Models
-
PALATE: Pretrained Autoregressive LATent Embeddings
-
HALO: Human-Grounded Granular Evaluation of Proactive Agents
-
Tactus: Tactile & Pose Sensing Glove for Natural Dexterity Data Collection
-
PhysicsNeMo-Mesh: GPU-Native Meshes and Simulation Cases for Physics AI
-
Credit Saison achieves 954% ROI in a ChatGPT Enterprise pilotCredit Saison、ChatGPT Enterprise試行でROI 954%を達成Credit SaisonのChatGPT Enterprise試行では、投資対効果が954%に達した。リーダーと従業員の双方が関わることで、試行の成果を全社的な導入につなげた経緯が紹介されている。
-
Who Gets to Build the Frontier for AICohere、民主主義圏にはより多くのフロンティアAI開発者が必要と論じるCohereの共同創業者らは、AIが経済や公共サービス、国防を支える基盤になると指摘した。誰がどの法の下で責任を持って作るかが各国の選択肢を左右するため、民主主義圏にはフロンティアAIを担う国や企業がもっと必要だと主張している。
-
Build a Financial Research Assistant with Cohere AI on AWSCohereがAWS上で財務書類検索・回答アシスタントを作る手順を解説Cohere Embed・Rerank・ParseをAmazon SageMaker AI、OpenSearch Serverless、Bedrockと組み合わせ、財務書類をページ単位の引用付きで検索・回答する構成を紹介。索引にEmbed Pro、問い合わせにEmbed Fastを使い分け、Rerank ProとContext Focusで精度・遅延・コストを調整する。
-
Closed Gates Leak: Decomposing Adjacent-Layer Cancellation in Gated Language Models
-
HEARSAYBENCH: Can AI Navigate from Abstract Human Rights to Lived Lives?
-
Tracing the Influence of Retrieved Literature on Scientific Ideation in LLMs
-
When Does Closed-Loop Curriculum Adaptation Help GRPO? An Analysis on Countdown
-
BASILISC: Bayesian Alignment of Siamese-Integrated Latents for Inferring Single-cell Correspondence
-
Reinforcement Learning for the Game of Balatro: A Fast Simulator and Two-Policy Training Approach
-
Distilling the Shape of Uncertainty: What On-Policy Distillation Learns and How to Predict It