Prépublication
· 2026 arXiv
EN
We propose BrickBench, a benchmark for agentic text-conditioned LEGO-set design. Given a prompt, an agent is tasked with producing an assembly that not only satisfies semantic and design criteria, but that can also be physically built. To do so, it must select...
Peter Kulits, Yiqing Xu et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
Bifurcations are ubiquitous in physical systems, from structural buckling to fluid and climate dynamics, yet they remain largely unexplored in deep learning. At a symmetry-breaking bifurcation, a single input admits multiple equally valid solutions, violating...
Anna Zimmel, Fleur Hendriks et al.
#cs.LG
Prépublication
· 2026 arXiv
EN
Recent incidents have highlighted the challenge of monitoring LLM agents and the danger of models deceiving people. We show that white-box deception detection via probes can be scaled up to frontier monitoring settings by collecting the largest deception datas...
Oskar J. Hollinsworth, Alex F. Spies et al.
#cs.LG
Prépublication
· 2026 arXiv
EN
AI agents can now conduct real-world cyberattacks, scale up capabilities with the number of agents, and collectively pursue misaligned goals to obtain rewards. Together, these factors raise the risk of a population explosion of misaligned agents: agents could...
Erin Crawley, Hidenori Tanaka
#cs.AI
Prépublication
· 2026 arXiv
EN
Streaming Video Large Language Models (VLMs) enable continuous video understanding, yet existing benchmarks focus on low-dynamic scenarios. Under bounded context budgets, models must balance temporal history, spatial resolution, and temporal granularity; spars...
Yuxuan Hu, Weikang Shi et al.
#cs.CV
Prépublication
· 2026 arXiv
EN
Multimodal large language models (MLLMs) struggle with spatial, embodied, physical, and temporal reasoning. We hypothesize that these failures reflect a shared deficit in visual transition reasoning, and test whether this capability can serve as a shared train...
Zheyu Fan, Yue Zhang et al.
#cs.CV
Prépublication
· 2026 arXiv
EN
Generating realistic human-object interactions (HOI) in complex 3D scenes requires two complementary capabilities: reasoning about interaction feasibility in the environment and synthesizing realistic human-object motion. However, supervision for these capabil...
Mingyuan Lei, Yoonchang Sung et al.
#cs.CV
Prépublication
· 2026 arXiv
EN
LLM developers post-train their models to exhibit prosocial values and behavioral traits, which are enumerated in an alignment target. However, while recent post-training developments have yielded models that score highly on alignment evaluations, training mod...
Andy Liu, Mehar Bhatia et al.
#cs.CL
Prépublication
· 2026 arXiv
EN
Safety benchmarks typically report one overall score for a suite of datasets, each of which may target one or more safety-related attributes, so models with similar overall scores can have very different attribute profiles. Comparing models is more tractable a...
Christopher M. Stewart, Preston Botter et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
Skill-augmented agents improve sample efficiency by distilling successful trajectories into reusable strategies. Yet most existing approaches remain text-centric, linearizing spatial layouts and action-state correspondences into language that loses critical ge...
Hongxing Li, Dingming Li et al.
#cs.CV
Prépublication
· 2026 arXiv
EN
Existing spatial reasoning benchmarks mainly test spatial perception: reading off relations already visible in the input. Yet real-world spatial intelligence demands predictive spatial reasoning: constructing a scene from observations, anticipating how an inte...
Hongxing Li, Jinyue Su et al.
#cs.CV
Prépublication
· 2026 arXiv
EN
Gravitational waves from vacuum first-order phase transitions probe the nonlinear dynamics of relativistic bubble collisions. Predicting their spectrum requires resolving Lorentz-contracted scalar bubble walls throughout a volume containing many bubbles, makin...
Malte Buschmann, Toby Opferkuch
#gr-qc