Next Location Prediction is a fundamental task in the study of human mobility, with wide-ranging applications in transportation planning, urban governance, and epidemic forecasting. In practice, when humans attempt to predict the next location in a trajectory, they often visualize the trajectory on a map and reason based on road connectivity and movement trends. However, the vast majority of existing next-location prediction models do not reason over maps \textbf{in the way that humans do}. Fortunately, the recent development of Vision-Language Models (VLMs) has demonstrated strong capabilities in visual perception and even visual reasoning. This opens up a new possibility: by rendering both the road network and trajectory onto an image and leveraging the reasoning abilities of VLMs, we can enable models to perform trajectory inference in a human-like manner. To explore this idea, we first propose a method called Vision-Guided Location Search (VGLS), which evaluates whether a general-purpose VLM is capable of trajectory-based reasoning without modifying any of its internal parameters. Based on insights from the VGLS results, we further propose our main approach: VLMLocPredictor, which is composed of two stages: In the first stage, we design two Supervised Fine-Tuning (SFT) tasks that help the VLM understand road network and trajectory structures and acquire basic reasoning ability on such visual inputs. In the second stage, we introduce Reinforcement Learning from Visual Map Feedback, enabling the model to self-improve its next-location prediction ability through interaction with the environment. Experiments conducted on datasets from four different cities show that our method achieves state-of-the-art (SOTA) performance and exhibits superior cross-city generalization compared to other LLM-based approaches.

The absence of a signal of new particles at the Large Hadron Collider (LHC) may suggest that new physics is realized in a scenario that is hard to detect due to the absence or very large mass of new colored particles. Hence, this study focuses on setups with dominant electroweak production of color-neutral new particles and multi-lepton signals from their decays. The conventional approach to searches for new physics, also known as "cut-and-count analysis", is to apply a set of constraints on different kinematic variables (called "cuts" or "selection") that improve the signalto-background ratio. However, the scenarios we consider can be challenging for this standard approach due to the small production cross section and the similarity of signal and background features. For such problems, machine learning (ML) offers a promising alternative [1-6]. We investigate how much ML can increase the discovery reach, and whether machine learning models can be trained in such a way that they work in a large region of parameter space and not just for a single point. This is an important issue, in particular in new physics scenarios with many free parameters, as signal kinematics vary from point to point. As a concrete example, we consider a supersymmetry (SUSY) scenario with a gravitino lightest supersymmetric particle (LSP) whose mass is in the GeV range.