전체 글394 [2026-1] 백승우 - AutoWebWorld: Synthesizing Infinite Verifiable Web Environments via Finite State Machines https://arxiv.org/abs/2602.14296 AutoWebWorld: Synthesizing Infinite Verifiable Web Environments via Finite State MachinesThe performance of autonomous Web GUI agents heavily relies on the quality and quantity of their training data. However, a fundamental bottleneck persists: collecting interaction trajectories from real-world websites is expensive and difficult to verify. Tarxiv.org 2026. 3. 10. [2026-1] 김지은 - Image Super-Resolution via Iterative Refinement 본 글에서는 기존 GAN 기반의 Super-Resolution이 불안정하고 artifact가 생긴다는 문제를 해결하기 위해 DDPM을 활용해 저해상도 이미지를 Super-Resolution하는 새로운 방법을 제안한 SR3를 살펴본다.1. IntroductionSingle-image super-resolution은 하나의 저해상도 이미지에 대해 여러 개의 가능한 고해상도 결과가 존재하는 multi-modal inverse problem이다. 단순 regression 기반 방법은 평균을 예측하기 때문에 고배율에서 blur가 발생하며, GAN, VAE, AR 모델은 각각 학습 불안정성, 샘플 품질 한계, 높은 계산 비용 등의 문제를 가진다. SR3는 이러한 한계를 해결하기 위해 Diffusion Model 기.. 2026. 2. 28. [2026-1] 김효민 - GQA: Training Generalized Multi-Query Transformer Models fromMulti-Head Checkpoints [Paper]GQA : https://arxiv.org/abs/2305.13245 GQA: Training Generalized Multi-Query Transformer Models from Multi-Head CheckpointsMulti-query attention (MQA), which only uses a single key-value head, drastically speeds up decoder inference. However, MQA can lead to quality degradation, and moreover it may not be desirable to train a separate model just for faster inference. We (1) prarxiv.org MQ.. 2026. 2. 28. [2026-1] 박승원 - RoFormer: Enhanced Transformer with Rotary Position Embedding 논문 정보게재 학술지: Neurocomputing, Volume 568, 1 February 2024, 127063저자: (Zhuiyi Technology)Jainlin Su, Murtadha Ahmed, Yu Lu, Shengfeng Pan, Wen Bo, Yunfeng Liuhttps://arxiv.org/abs/2104.09864 RoFormer: Enhanced Transformer with Rotary Position EmbeddingPosition encoding recently has shown effective in the transformer architecture. It enables valuable supervision for dependency modeling between elem.. 2026. 2. 27. 이전 1 2 3 4 ··· 99 다음
