Unmanned aerial vehicles require reliable autonomous positioning beyond the limitations of GNSS, motivating the development of a vision-based, end-to-end Finding Point in Map algorithm. This study introduces MuRDE-FPN, an enhanced Feature Pyramid Network (FPN) designed for precise UAV localization, building upon a lightweight one-stream OS-PCPVT transformer backbone. MuRDE-FPN integrates Efficient Channel Attention for adaptive channel recalibration and features two novel components: a MultiReceptive Deformable Enhancement block that utilizes DCNv2 with varying kernel sizes to refine the semantically rich […]

Ver mais

This blog post outlines techniques for improving the training performance of your PyTorch model without compromising its accuracy. To do so, we will wrap a PyTorch model in a LightningModule and use the Trainer class to enable various training optimizations. By changing only a few lines of code, we can reduce the training time on a single GPU from 22.53 minutes to 2.75 minutes while maintaining the model’s prediction accuracy. Yes, that’s a 8x performance boost!

Ver mais

Reinforcement Learning from Verifiable Rewards (RLVR) trains large language models (LLMs) from sampled trajectories, making decoding strategy a core component of learning rather than a purely inference-time choice. Sampling temperature directly controls the exploration–exploitation trade-off by modulating policy entropy, yet existing methods rely on static values or heuristic adaptations that are decoupled from task-level rewards. We propose Introspective LLM, a hierarchical reinforcement learning framework that learns to control sampling temperature during generation. At each decoding step, the model […]

Ver mais

arXiv:2601.00254v1 Announce Type: new Abstract: The rapid advancement of Large Language Models (LLMs) presents new opportunities for automated software vulnerability detection, a crucial task in securing modern codebases. This paper presents a comparative study on the effectiveness of LLM-based techniques for detecting software vulnerabilities. The study evaluates three approaches, Retrieval-Augmented Generation (RAG), Supervised Fine-Tuning (SFT), and a Dual-Agent LLM framework, against a baseline LLM model. A curated dataset was compiled from Big-Vul and real-world code repositories from GitHub, […]

Ver mais

arXiv:2601.20986v1 Announce Type: new Abstract: Social movements supporting the UN’s Sustainable Development Goals (SDGs) play a vital role in improving human lives. If journalists were aware of the relationship between social movements and external events, they could provide more precise, time-sensitive reporting about movement issues and SDGs. Our SMART system achieves this goal by collecting data from multiple sources, extracting emotions on various themes, and then using a transformer-based forecasting engine (DEEP) to predict quantity and intensity of […]

Ver mais

I watch a lot of nature documentaries. I’m not very choosy about the animals covered, whether whales, moles, lions, ants, chameleons, blowfish, or mosquitoes. I’m even fascinated by footage of bacteria under a microscope. I’m usually immersed as I sit in front of my large-screen television, so long as I learn something about the intricacies of the species filmed in vibrant colors. What do they eat and how do they avoid being eaten? What are their life expectancies, […]

Ver mais

In early September, EFF submitted an amicus brief to Ecuador’s Constitutional Court supporting a constitutional challenge filed by Ecuadorian NGOs, including INREDH and LaLibre. The case challenges the constitutionality of the Ley Orgánica de Inteligencia (LOI) and its implementing regulation, the General Regulation of the LOI. EFF’s amicus brief argues that the LOI enables disproportionate surveillance and secrecy that undermine constitutional and Inter-American human rights standards. EFF urges the Constitutional Court to declare the LOI and its regulation […]

Ver mais

arXiv:2601.12518v1 Announce Type: cross Abstract: Cooperative MARL often assumes frequent access to global information in a data buffer, such as team rewards or other agents’ actions, which is typically unrealistic in decentralized MARL systems due to high communication costs. When communication is limited, agents must rely on outdated information to estimate gradients and update their policies. A common approach to handle missing data is called importance sampling, in which we reweigh old data from a base policy to […]

Ver mais