0x4861·Oct 26Understanding Luong Attention in Neural Machine TranslationWhy Global and Local Attention and Input-feeding work, what the English–German experiments show, and where the methods fall shortA response icon1A response icon1
0x4861·Oct 26Luong Attention 논문으로 이해하는 번역 모델의 어텐션Global·Local Attention과 Input-feeding이 해결하려던 문제부터 수식, 양방향 실험, 장단점까지 살펴봅니다
Andrey Lukyanenko·Oct 26Jeva: Rethinking System One Models from First PrinciplesA review of Jeva, a decision-native architecture for system one models that encodes the request and each option independently, refines only…
Isanrfauzi·Oct 23Seberapa Andal Forensik Digital Android? Review Kritis atas Sebuah Systematic Literature ReviewReview jurnal Procedia Computer Science 2024 tentang investigasi forensik pada perangkat Android
Andrey Lukyanenko·Sep 13DeepSeek-V4.1-Flash: Frontier Agents on a Smaller Memory BudgetDeepSeek-V4.1-Flash is a 552B multimodal MoE model that attacks KV cache cost along all three of its multiplicative dimensions at once…
Andrey Lukyanenko·Sep 8Qwen-Drive 1.0: Turning a General VLM into a Driving Foundation ModelQwen-Drive-1.0 turns Qwen3.5–4B into a unified autonomous-driving model for 3D perception, scene understanding, reasoning, and trajectory…A response icon2A response icon2
Andrey Lukyanenko·Aug 3Beyond Bigger MoE: How Kimi K3 Scales Context, Depth, and AgentsKimi K3 is a 2.8-trillion-parameter open-weight multimodal MoE model that combines hybrid linear–global attention, selective residual…
Jang·Jul 29Can We Really Trust Every View in Multi-View Classification?Reducing Unreliable Evidence with Testing-Time Evidence Filtering
Dong-Keon Kim·Jul 20Correlated-Error Forecasting: Calibrating Multivariate Predictions with Residual DependenciesA plug-and-play covariance model for sequential model based probabilistic forecastingA response icon1A response icon1
Andrey Lukyanenko·Jul 20Harness Handbook: The Missing Layer for Editing AI AgentsHarness Handbook proposes a new way to navigate large AI agent codebases by organizing them around behaviors instead of files. The result…
