Rising on arXiv - 2025-07-11

A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (yes, I do a lot of filtering to keep away most of the genAI stuff; explainer).

1. Multi-target tracking: Turning raw sensor data into usable information about the world is very tricky computationally (just take a look at any animal's brain); historically speaking a lot of computing — even before we had computers — was driven by one variation or another of this problem, and for good and for ill that still seems to be the case.

Some recent articles:

• Adaptive Resource Management in Cognitive Radar via Deep Deterministic Policy Gradient

• Belief Propagation-based Target Handover in Distributed Integrated Sensing and Communication

• Transformer Based Multi-Target Bernoulli Tracking for Maritime Radar

• Frequency Range 3 for ISAC in 6G: Potentials and Challenges

• Cross-Channel Unlabeled Sensing over a Union of Signal Subspaces

2. Spatial audio: You can look at this as the dual of the sensor data integration problem: how do you tweak inputs to the human auditory system to fool it (for, say, entertainment purposes)?

Some recent articles:

• SonicMotion: Dynamic Spatial Audio Soundscapes with Latent Diffusion Models

• Machine Learning in Acoustics: A Review and Open-Source Repository

• SEE-2-SOUND: Zero-Shot Spatial Environment-to-Spatial Sound

• Sonic4D: Spatial Audio Generation for Immersive 4D Scene Exploration

• Learning Magnitude Distribution of Sound Fields via Conditioned Autoencoder

3. Autoregressive modeling: A necessarily mixed bag of papers given how basic the concept is, but you might find something interesting in there.

Some recent articles:

• Supply Chain Optimization via Generative Simulation and Iterative Decision Policies

• UniCode$^2$: Cascaded Large-scale Codebooks for Unified Multimodal Understanding and Generation

• JanusDNA: A Powerful Bi-directional Hybrid DNA Foundation Model

• Object-centric Denoising Diffusion Models for Physical Reasoning

• Token Communication in the Era of Large Models: An Information Bottleneck-Based Approach

Thank you to arXiv for use of its open access interoperability

(Originally posted on my blog.)