Contributions de Pitpitt


Rechercher des contributionsaffichermasquer
⧼contribs-top⧽
⧼contribs-date⧽
(les plus récentes | les plus anciennes) Voir ( | ) (20 | 50 | 100 | 250 | 500)

6 novembre 2025

27 octobre 2025

  • 20:0727 octobre 2025 à 20:07 diff hist +906‎ N LightMem ‎ Page créée avec « == EN CONSTRUCTION == == Définition == xxxxx == Français == '''LightMem''' == Anglais == '''LightMem''' Lightweight and Efficient Memory-Augmented Generation A memory system for Large Language Models that addresses the significant computational overhead of existing memory architectures. The system draws inspiration from human memory processes to create a more efficient approach to storing and retrieving information during extended conversations. By impl... » actuelle
  • 20:0627 octobre 2025 à 20:06 diff hist +976‎ N Seedream ‎ Page créée avec « == EN CONSTRUCTION == == Définition == xxxxx == Français == '''Seedream''' == Anglais == '''Seedream''' Toward Next-generation Multimodal Image Generation A multimodal image generation system that unifies text-to-image synthesis, image editing, and multi-image composition in a single framework. The system achieves state-of-the-art performance while maintaining ultra-fast inference speeds, generating high-resolution images up to 4K resolution. The model... » actuelle
  • 20:0227 octobre 2025 à 20:02 diff hist +997‎ N Paper2Video ‎ Page créée avec « == EN CONSTRUCTION == == Définition == xxxxx == Français == '''xxxxx ''' == Anglais == '''Paper2Video''' Automatic Video Generation from Scientific Papers A system that automatically generates academic presentation videos from research papers. The work addresses the time-consuming process of creating presentation videos, which typically requires hours of slide design, recording, and editing for just a few minutes of content. The paper presents both a be... » actuelle
  • 20:0127 octobre 2025 à 20:01 diff hist +39‎ VideoCanvas ‎Aucun résumé des modifications actuelle
  • 20:0027 octobre 2025 à 20:00 diff hist +990‎ N VideoCanvas ‎ Page créée avec « == EN CONSTRUCTION == == Définition == xxxxx == Français == '''VideoCanvas''' == Anglais == '''VideoCanvas''' ==Sources== [XXXX Sources : XXXXX ] Unified Video Completion from Arbitrary Spatiotemporal Patches via In-Context Conditioning A unified framework for arbitrary spatio-temporal video completion that allows users to place content patches at any location and timestamp in a video, with the model filling in the remaining regions. This approach... »
  • 19:5927 octobre 2025 à 19:59 diff hist +1‎ UniVideo ‎Aucun résumé des modifications actuelle
  • 19:5827 octobre 2025 à 19:58 diff hist +940‎ N UniVideo ‎ Page créée avec « == EN CONSTRUCTION == == Définition == xxxxx == Français == '''UniVideo''' == Anglais == '''xxxUniVideoxx ''' A unified framework that combines video understanding, generation, and editing capabilities within a single model. Unlike existing approaches that handle these tasks separately, UniVideo can interpret complex multimodal instructions and perform diverse video operations through a dual-stream architecture. The system demonstrates strong performance a... »
  • 19:5727 octobre 2025 à 19:57 diff hist +640‎ N Representation Autoencoders ‎ Page créée avec « == EN CONSTRUCTION == == Définition == xxxxx == Français == '''xxxxx ''' == Anglais == '''Representation Autoencoders''' representation encoders like DINO or SigLIP. The method challenges the common assumption that semantic encoders are unsuitable for reconstruction tasks and demonstrates that they can actually provide superior performance for image generation. Replacing VAEs with pretrained representation encoders in Diffusion Transformers enhances gen... » actuelle
  • 19:5627 octobre 2025 à 19:56 diff hist +782‎ N OmniVideoBench ‎ Page créée avec « == EN CONSTRUCTION == == Définition == xxxxx == Français == '''OmniVideoBench ''' == Anglais == '''OmniVideoBench''' A comprehensive benchmark designed to evaluate how well multimodal large language models (MLLMs) can understand and reason across both audio and visual information in videos. The benchmark addresses a critical gap in current evaluation methods, which often focus on single modalities or fail to properly integrate audio-visual reasoning in a l... » actuelle
  • 19:5427 octobre 2025 à 19:54 diff hist +1 357‎ N Mode Collapse ‎ Page créée avec « == EN CONSTRUCTION == == Définition == xxxxx == Français == ''' xxxxx ''' == Anglais == ''' Mode Collapse ''' While deep learning has expanded the possibilities for highly expressive variational families, the practical benefits of these tools for variational inference (VI) are often limited by the minimization of the traditional Kullback-Leibler objective, which can yield suboptimal solutions. A major challenge in this context is \emph{mode collapse}: the... » actuelle
  • 19:5227 octobre 2025 à 19:52 diff hist +614‎ N Verbalized Sampling ‎ Page créée avec « == EN CONSTRUCTION == == Définition == xxxxx == Français == ''' xxxxx ''' == Anglais == '''Verbalized Sampling''' How to Mitigate Mode Collapse and Unlock LLM Diversity https://arxiv.org/abs/2510.01171 A training-free prompting strategy to mitigate mode collapse in LLMs by requesting responses with probabilities. Achieves 2-3x diversity improvement while maintaining quality. Model-agnostic framework with CLI/API for creative writing, synthetic data gen... » actuelle

21 octobre 2025

17 octobre 2025

15 octobre 2025

13 octobre 2025

10 octobre 2025

7 octobre 2025

(les plus récentes | les plus anciennes) Voir ( | ) (20 | 50 | 100 | 250 | 500)
Récupérée de « https://datafranca.org/wiki/Spécial:Contributions/Pitpitt »