Flow matching is a powerful framework for generating images and other data by learning to map noise to structure, but it suffers from a training-inference mismatch: models are trained on clean trajectories but must operate on drifted ones at test time. DEFAR turns this problem on its head, treating the drift itself as a useful signal. It uses the bias to learn corrective directions and to reinforce missing low-frequency information that tends to degrade high-noise generation stages. Experiments on CIFAR-10, CelebA, and ImageNet show consistent gains. Applications include higher-fidelity image synthesis, video generation, and scientific simulations that use diffusion or flow-based generative models.

Authors: Guanbo Huang, Jingjia Mao, Fanding Huang, Fengkai Liu, Xiangyang Luo, Yaoyuan Liang, Jiasheng Lu, Xiaoe Wang, Pei Liu, Ruiliu Fu, Ruqi Huang, Shao-Lun Huang

Paper: https://arxiv.org/abs/2606.28226v1

When AI agents autonomously write and merge code at scale, the usual way of evaluating them — task by task, in isolation — misses something important: the cumulative friction and technical debt that builds up in shared codebases over time. Studying over 930,000 agent-authored pull requests, this paper finds that about half of "integration friction" is a property of the repository ecosystem, not any individual contribution — and agent-authored code concentrates this risk roughly twice as much as human-authored code. Implications are significant for software governance, DevOps policy, AI code review tooling, and organizations that are beginning to integrate autonomous coding agents into production development workflows.

Authors: Daniel Russo

Paper: https://arxiv.org/abs/2606.28235v1

Why do bigger neural networks tend to perform better — and by exactly how much? Scaling laws attempt to answer this, but most existing theory relies on simplified assumptions about infinite width or unlimited data. This work studies how generalization error changes as both model width and dataset size vary simultaneously in a tractable two-layer network, revealing a phase diagram with distinct regimes — including a transition into interpolation — governed by the spectral structure of the target function. While theoretical, these findings have practical implications for deciding how to allocate compute budgets, understanding when more data helps more than more parameters, and designing efficient architectures.

Authors: Julius Girardin, Emanuele Troiani, Yizhou Xu, Vittorio Erba, Florent Krzakala, Lenka Zdeborová

Paper: https://arxiv.org/abs/2606.28242v1

Detecting defects in manufactured goods — a crack in a circuit board, a tear in fabric — requires models that can generalize across wildly different visual conditions. TopoTTA brings an unusual tool to this problem: persistent homology, a mathematical framework that captures the shape and connectivity of structures across scales. Rather than relying on simple pixel-confidence thresholds, it uses topological features derived from anomaly score maps to generate more reliable pseudo-labels at test time. Evaluated across six benchmarks and both 2D and 3D data, it achieves a 15% average F1 improvement. Applications include industrial quality control, medical imaging anomaly detection, and autonomous inspection systems.

Authors: Ali Zia, Usman Ali, Abdul Rehman, Umer Ramzan, Kang Han, Muhammad Faheem, Shahnawaz Qureshi, Wei Xiang

Paper: https://arxiv.org/abs/2606.28268v1

As AI agents gain the ability to use tools, access memory, and coordinate with other agents, they become vulnerable to entirely new classes of attacks — malicious instructions injected through tool outputs, poisoned memory, or compromised peer agents. ANIS proposes a defense architecture modeled on the biological immune system, embedded directly inside the agent's reasoning process rather than bolted on externally. It distinguishes between shallow rule-based defenses and deeper parametric "vaccines," and introduces a self-monitoring layer that adapts to novel threats at runtime. Applications span enterprise AI deployments, autonomous research agents, multi-agent financial systems, and any context where agents act with high autonomy on sensitive tasks.

Authors: Bo Shen, Lifeng Chang, Tianyuan Wei, Yunpeng Li, Feng Shi, Yichen Han, Peijie Gao, Shiyi Kuang, Xin Chang, Dehui Li

Paper: https://arxiv.org/abs/2606.28270v1

Cellular networks in cities are under constant, unpredictable stress — traffic jams, concerts, and commutes all reshape how and where data flows. Predicting this demand accurately is essential for carriers to allocate bandwidth intelligently. PEHT introduces a transformer-based model that separates core network traffic signals from external urban mobility and congestion data, then fuses them efficiently using Low-Rank Adaptation (LoRA) to keep the model lightweight. Tested on real Milan telecom data, it outperforms existing approaches. Practical uses include real-time dynamic spectrum allocation, 5G network planning, smart city infrastructure management, and adaptive resource scheduling in densely populated urban environments.

Authors: Abdolazim Rezaei, Mehdi Sookhak, Mahboobeh Haghparast

Paper: https://arxiv.org/abs/2606.28274v1

The volume of scientific papers being published is growing faster than human reviewers can keep pace with — a crisis accelerated by AI-assisted research generation. This paper proposes a taxonomy of AI-human collaboration levels in peer review, then introduces PAT (Paper Assistant Tool), an agentic system that reads full manuscripts and produces structured evaluations, including checks of mathematical proofs and experimental validity. In pilot deployments at STOC and ICML, PAT caught real errors before submission. Broader applications include pre-submission screening tools for journals, institutional research quality checks, and helping junior researchers self-audit their work before seeking expert review.

Authors: Rajesh Jayaram, Drew Tyler, David Woodruff, Corinna Cortes, Yossi Matias, Vahab Mirrokni, Vincent Cohen-Addad

Paper: https://arxiv.org/abs/2606.28277v1

Designing computer chips is extraordinarily complex, requiring expertise across logic, timing, and physical layout — making it a compelling frontier for AI automation. HORIZON treats hardware design the same way modern AI treats software: as an evolving codebase that an agent can iteratively improve. By wrapping design tasks in a structured "project pack" with executable evaluators, the agent refines hardware descriptions through a self-correcting git-based loop, achieving full completion on several established benchmarks. Potential applications include dramatically accelerating chip design cycles, democratizing access to hardware engineering, and enabling autonomous exploration of hardware architectures for custom AI accelerators and embedded systems.

Authors: Cunxi Yu, Chenhui Deng, Nathaniel Pinckney, Brucek Khailany

Paper: https://arxiv.org/abs/2606.28279v1

In competitive games — from poker to cybersecurity — there isn't always a single optimal strategy, but rather a whole family of equally valid equilibria. Which one an AI solver picks can quietly determine how it behaves against opponents who don't play perfectly. This work reveals that the choice of algorithm, not random chance, systematically drives which equilibrium gets selected. Regularized methods like R-NaD tend toward maximum-entropy (most unpredictable) strategies, while regret-averaging methods like CFR drift toward more exploitable ones. This has direct implications for AI agents in auctions, negotiations, multi-agent games, and any setting where strategic robustness against imperfect opponents matters.

Authors: Luis Leal

Paper: https://arxiv.org/abs/2606.28308v1

Robotic hands capable of dexterous manipulation have made impressive strides, but teaching a single hand to perform multiple tasks simultaneously — without one task undoing another — remains a hard unsolved problem. Imagine a robot hand that already knows how to hold an object securely; asking it to also open a latch might cause it to loosen its grip. DexCompose addresses this by assigning explicit "ownership" of individual fingers to different tasks, letting each sub-policy operate in its own movement subspace. Applications range from assembly-line robotics to prosthetics, where a single robotic hand must seamlessly handle complex, overlapping manipulation demands in real time.

Authors: Dihong Huang, Zhenyu Wei, Zhuxiu Xu, Yunchao Yao, Sikai Li, Mingyu Ding

Paper: https://arxiv.org/abs/2606.28323v1