「generation」の検索結果
20 件While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, relia
In proton therapy, plans are typically optimized on a single planning CT, making robustness evaluation essenti
Reliable video generation requires more than high-quality frames to form a coherent story: a model must mainta
Diffusion models have shown remarkable performance on diverse generation tasks. Recent work finds that imposin
Token-level text anomaly detection, as an emerging trend of text anomaly detection, moves beyond coarse-graine
Generative diffusion models have emerged as a class of powerful techniques for various imaging applications, i
Recent multi-shot audio-video generators can produce increasingly coherent and cinematic outputs, but coherenc
We study scheming in LLM agents, in which agents covertly pursue misaligned goals. Our focus is to understand
Remote sensing multimodal large language models (RS-MLLMs) have advanced scene understanding and visual questi
Implicit neural representation (INR) has achieved remarkable progress in novel view synthesis and image/video
Organoids are three-dimensional tissue models whose morphology provides important insights into tumor developm
While multimodal large language models (MLLMs) achieve remarkable performance on generic image captioning, the
Large language models (LLMs) have recently emerged as a promising tool for automating robot design from high-l
Retrieval-augmented generation (RAG) enables large language models (LLMs) to answer questions by accessing ext
World-action models guide action generation with predicted future observations, but vision-centric predictions
Constant optimization refines the numerical coefficients of candidate expressions in tree-based genetic progra
Generative modeling directly on geometric manifolds can avoid errors introduced by flattening non-Euclidean da
In this paper, we introduce ES-AHD, a novel framework that fundamentally integrates Evolution Strategy (ES) in
Population optimizers such as CMA-ES, DE, and multi-objective evolutionary algorithms drive search mainly thro
Most LLM-based automated algorithm design methods optimize a designated component within a human-specified sca