arXiv:2602.05010v1 Announce Type: new Abstract: In the real world, robots frequently make errors, yet little is known about people’s social responses to errors outside of lab settings. Prior work has shown that social signals are reliable and useful for error management in constrained interactions, but it is unclear if this holds in the real world – especially with a non-social robot in repeated and group interactions with successive or propagated errors. To explore this, we built a coffee […]

Ver mais

The more a thing tends to be permanent, the more it tends to be lifeless. Allan Watts Hey, listen. You’re not lazy. Let’s get that out of the way right now. Lazy people don’t lie awake at night, hearts pounding, staring at the ceiling while mentally screaming at themselves for not sending that email, not starting that proposal, not having that hard conversation. You do. I do. This is the special, advanced-class procrastination reserved for people who are […]

Ver mais

arXiv:2602.17696v1 Announce Type: new Abstract: Large language models (LLMs) are often assumed to contain “safety regions” — parameter subsets whose modification directly influences safety behaviors. We conduct a systematic evaluation of four safety region identification methods spanning different parameter granularities, from individual weights to entire Transformer layers, across four families of backbone LLMs with varying sizes. Using ten safety identification datasets, we find that the identified safety regions exhibit only low to moderate overlap, as measured by IoU. […]

Ver mais

arXiv:2602.02531v1 Announce Type: new Abstract: The hypersonic unstart phenomenon poses a major challenge to reliable air-breathing propulsion at Mach 5 and above, where strong shock-boundary-layer interactions and rapid pressure fluctuations can destabilize inlet operation. Here, we demonstrate a deep reinforcement learning (DRL)- based active flow control strategy to control unstart in a canonical two-dimensional hypersonic inlet at Mach 5 and Reynolds number $5times 10^6$. The in-house CFD solver enables high-fidelity simulations with adaptive mesh refinement, resolving key flow […]

Ver mais

arXiv:2601.07965v1 Announce Type: new Abstract: When a model knows when it does not know, many possibilities emerge. The first question is how to enable a model to recognize that it does not know. A promising approach is to use confidence, computed from the model’s internal signals, to reflect its ignorance. Prior work in specific domains has shown that calibration can provide reliable confidence estimates. In this work, we propose a simple, effective, and universal training-free method that applies […]

Ver mais

Here’s my take on teaching AI to play a video game, with the fun twist that this time nobody ever heard of it. DDNet (aka Teeworlds) is an open source retro multiplayer platformer with different game modes like pvp and race modes. Players can walk, jump, use grappling hook and various weapons. In this project, I focused on solo race mode. For the algorithm I chose PPO, but tried various reward shaping methods that I found interesting/promising, such […]

Ver mais

How can a trillion-parameter Large Language Model achieve state-of-the-art enterprise performance while simultaneously cutting its total parameter count by 33.3% and boosting pre-training efficiency by 49%? Yuan Lab AI releases Yuan3.0 Ultra, an open-source Mixture-of-Experts (MoE) large language model featuring 1T total parameters and 68.8B activated parameters. The model architecture is designed to optimize performance in enterprise-specific tasks while maintaining competitive general-purpose capabilities. Unlike traditional dense models, Yuan3.0 Ultra utilizes sparsity to scale capacity without a linear increase […]

Ver mais

arXiv:2603.10030v1 Announce Type: new Abstract: AI transport libraries move bytes efficiently, but they commonly assume that buffers are already correctly allocated, placed, shared, registered, and safe under completion and teardown pressure. This paper presents dmaplane, a Linux kernel module that makes this missing layer explicit as buffer orchestration. dmaplane exposes a stable kernel UAPI via /dev/dmaplane and composes ring-based command channels, DMA buffer lifecycle management, dma-buf export for cross-device sharing, a kernel-space RDMA engine, NUMA-aware allocation and verification, […]

Ver mais