Carlos's Debrief

⚡ Quick Summary

DelTA: Discriminative Token Credit Assignment for Reinforcement Learning from Verifiable Reward: Reinforcement learning from verifiable rewards (RLVR) has emerged as a central technique for improving the reasoning capabilities of large language models.
TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation: Public transit route planning traditionally depends on structured map infrastructure and complex routing engines, and no existing dataset supports training models to bypass this dependency.
Perception or Prejudice: Can MLLMs Go Beyond First Impressions of Personality?: Multimodal Large Language Models (MLLMs) are increasingly deployed in human-facing roles where personality perception is critical, yet existing benchmarks evaluate this capability solely on numerical Big Five score prediction, leaving open whether models truly perceive personality through behavioral understanding or merely prejudge through superficial pattern matching.
π-Bench: Evaluating Proactive Personal Assistant Agents in Long-Horizon Workflows: The rise of personal assistant agents, e.
Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training St: Long-context inference in large language models is bottlenecked by the quadratic cost of full attention.

📰 News Headlines

Lobste.rs: Introduction Modern ML workloads depend heavily on custom GPU kernels. Even when a model is expressed as clean tensor operations, the performance almost a...
Lobste.rs: Hello All I am the author of this post. A couple of weeks ago, my previous piece ('The 90 day disclosure policy is dead') ended up here…
Lobste.rs: Friends Matthew Garrett Jonathan McDowell Jo McIntyre Martin Michlmayr Andrew Mobbs Mike Pitt Daniel Silverstone Andy Simpkins Neil Williams
Lobste.rs: Megalodon: Mass GitHub Repo Backdooring via CI Workflows
Lobste.rs: Deleting a Google API key doesn't revoke it immediately. Our testing found successful authentications up to 23 minutes after deletion, and Google has declined to fix it.

📄 ArXiv Papers

🧠 LLMs 11

Forecasting Downstream Performance of LLMs With Proxy Metrics [1]

Arkil Patel, Siva Reddy, Marius Mosbach, Dzmitry Bahdanau

Progress in language model development is often driven by comparative decisions: which architecture to adopt, which pretraining corpus to use, or which training recipe to apply. Making these decisions well requires reliable performance forecasts, yet the two commonly used signals are fundamentally limited.…

arXiv:2605.18607

SEGA: Spectral-Energy Guided Attention for Resolution Extrapolation in Diffusion Transformers [2]

Javad Rajabi, Kimia Shaban, Koorosh Roohi, David B. Lindell, Babak Taati

Diffusion transformers (DiTs) have emerged as a dominant architecture for text-to-image generation, yet their performance drops when generating at resolutions beyond their training range. Existing training-free approaches mitigate this by modifying inference-time attention behavior, often through Rotary Position…

arXiv:2605.22668

DelTA: Discriminative Token Credit Assignment for Reinforcement Learning from Verifiable Rewards [3]

Kaiyi Zhang, Wei Wu, Yankai Lin

Reinforcement learning from verifiable rewards (RLVR) has emerged as a central technique for improving the reasoning capabilities of large language models. Despite its effectiveness, how response-level rewards translate into token-level probability changes remains poorly understood. We introduce a discriminator view…

arXiv:2605.21467

Training Large Language Models to Predict Clinical Events [4]

Benjamin Turtel, Paul Wilczewski, Kris Skotheim

Longitudinal clinical notes contain rich evidence of how patients evolve over time, but converting this signal into training supervision for clinical prediction remains challenging. We extend Foresight Learning to clinical prediction by converting time-ordered MIMIC-III notes into examples consisting of past patient…

arXiv:2605.12817

Perception or Prejudice: Can MLLMs Go Beyond First Impressions of Personality? [5]

Caixin Kang, Tianyu Yan, Sitong Gong, Mingfang Zhang, Liangyang Ouyang

Multimodal Large Language Models (MLLMs) are increasingly deployed in human-facing roles where personality perception is critical, yet existing benchmarks evaluate this capability solely on numerical Big Five score prediction, leaving open whether models truly perceive personality through behavioral understanding or…

arXiv:2605.22109

KVServe: Service-Aware KV Cache Compression for Communication-Efficient Disaggregated LLM Serving [6]

Zedong Liu, Xinyang Ma, Dejun Luo, Hairui Zhao, Bing Lu

LLMs are widely adopted in production, pushing inference systems to their limits. Disaggregated LLM serving (e.g., PD separation and KV state disaggregation) improves scalability and cost efficiency, but it also turns KV into an explicit payload crossing network and storage boundaries, making KV a dominant end-to-end…

arXiv:2605.13734

From Reasoning Chains to Verifiable Subproblems: Curriculum Reinforcement Learning Enables Credit Assignment for LLM Reasoning [7]

Xitai Jiang, Zihan Tang, Wenze Lin, Yang Yue, Shenzhi Wang

Reinforcement learning from verifiable rewards (RLVR) has shown strong promise for LLM reasoning, but outcome-based RLVR remains inefficient on hard problems because correct final-answer rollouts are rare and sample-level credit assignment cannot use partial progress in failed attempts. We introduce SCRL (Subproblem…

arXiv:2605.22074

Spreadsheet-RL: Advancing Large Language Model Agents on Realistic Spreadsheet Tasks via Reinforcement Learning [8]

Banghao Chi, Yining Xie, Mingyuan Wu, Jingcheng Yang, Jize Jiang

Spreadsheet systems (e.g., Microsoft Excel, Google Sheets) play a central role in modern data-centric workflows. As AI agents grow increasingly capable of automating complex tasks, such as controlling computers and generating presentations, building an AI-driven spreadsheet agent has emerged as a promising research…

arXiv:2605.22642

RiT: Vanilla Diffusion Transformers Suffice in Representation Space [9]

Le Zhang, Ning Mang, Aishwarya Agrawal

Flow matching with x-prediction -- regressing the clean data point rather than the ambient velocity -- is known to exploit low-dimensional manifold structure effectively in pixel space li2025back. We ask whether a pretrained representation space, while containing a low-dimensional data manifold of comparable intrinsic…

arXiv:2605.21981

Decoupling the Benefits of Subword Tokenization for Language Model Training via Byte-level Simulation [10]

Théo Gigant, Bowen Peng, Jeffrey Quesnelle

Subword tokenization is an essential part of modern large language models (LLMs), yet its specific contributions to training efficiency and model performance remain poorly understood. In this work, we decouple the effects of subword tokenization by isolating them within a controlled byte-level pretraining pipeline. We…

arXiv:2604.27263

Large language models (LLM) in computational social science: prospects, current state, and challenges [11]

⚙️ Machine Learning 18

Platonic Representations in the Human Brain: Unsupervised Recovery of Universal Geometry [12]

Pablo Marcos-Manchón, Rishi Jha, Lluís Fuentemilla

The Strong Platonic Representation Hypothesis suggests that representational convergence in artificial neural networks can be harnessed constructively: embeddings can be translated across models through a universal latent space without paired data. We ask whether an analogous geometry can be recovered across human…

arXiv:2605.20496

"I didn't Make the Micro Decisions": Measuring, Inducing, and Exposing Goal-Level AI Contributions in Collaboration [13]

Eunsu Kim, Jessica R. Mindel, Kyungjin Kim, Sherry Tongshuang Wu

As large language models (LLMs) increasingly shape how users form, refine, and extend their goals, attributing contributions in human-AI collaboration becomes critical for users calibrating their own reliance and for evaluators assessing AI-assisted work. Yet existing methods focus on final artifacts, missing the…

arXiv:2605.21363

DecQ: Detail-Condensing Queries for Enhanced Reconstruction and Generation in Representation Autoencoders [14]

Tianhang Wang, Yitong Chen, Wei Song, Zuxuan Wu, Min Li

Representation Autoencoders (RAEs) leverage frozen vision foundation models (VFMs) as tokenizer encoders, providing robust high-level representations that facilitate fast convergence and high-quality generation in latent diffusion models. However, freezing the VFM inherently constrains its spatial reconstruction…

arXiv:2605.22777

OmniPro: A Comprehensive Benchmark for Omni-Proactive Streaming Video Understanding [15]

Ruixiang Zhao, Jie Yang, Zijie Xin, Tianyi Wang, Fengyun Rao

Omni-proactive streaming video understanding, i.e., autonomously deciding when to speak and what to say from continuous audio-visual streams, is an emerging capability of omni-modal large language models. Existing benchmarks fall short in three key aspects: they rely primarily on visual signals, adopt polling or…

arXiv:2605.18577

Unsupervised Process Reward Models [16]

Artyom Gadetsky, Maxim Kodryan, Siba Smarak Panigrahi, Hang Guo, Maria Brbic

Process Reward Models (PRMs) are a powerful mechanism for steering large language model reasoning by providing fine-grained, step-level supervision. However, this effectiveness comes at a significant cost: PRMs require expert annotations for every reasoning step, making them costly and difficult to scale. Here, we…

arXiv:2605.10158

Minimalist Visual Inertial Odometry [17]

Francesco Pasti, Jeremy Klotz, Nicola Bellotto, Shree K. Nayar

Visual-Inertial Odometry(VIO), which is critical to mobile robot navigation, uses cameras with a large number of pixels. Capturing and processing camera images requires significant resources. This work presents a minimalist approach to planar odometry, demonstrating that just four visual measurements and an IMU can…

arXiv:2605.19990

TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation [18]

Hanyu Guo, Jiedong Yang, Chao Chen, Longfei Xu, Kaikui Liu

Public transit route planning traditionally depends on structured map infrastructure and complex routing engines, and no existing dataset supports training models to bypass this dependency. We present TransitLM, a large-scale dataset of over 13 million transit route planning records from four Chinese cities covering…

arXiv:2605.22355

Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training Steps [19]

Yanke Zhou, Yiduo Li, Hanlin Tang, Maohua Li, Kan Liu

Long-context inference in large language models is bottlenecked by the quadratic cost of full attention. Existing efficient alternatives often rely either on native sparse training or on heuristic token eviction, creating an undesirable trade-off among efficiency, training cost, and accuracy. In this work, we show…

arXiv:2605.16928

LatentOmni: Rethinking Omni-Modal Understanding via Unified Audio-Visual Latent Reasoning [20]

Yifan Dai, Zhenhua Wu, Bohan Zeng, Daili Hua, Jialing Liu

Joint audio-visual reasoning is essential for omnimodal understanding, yet current multimodal large language models (MLLMs) still struggle when reasoning requires fine-grained evidence from both modalities. A central limitation is that explicit text-based chain-of-thought (CoT) compresses continuous audio-visual…

arXiv:2605.22012

SpaceDG: Benchmarking Spatial Intelligence under Visual Degradation [21]

Xiaolong Zhou, Yifei Liu, Ziyang Gong, Jiarui Li, Qiyue Zhao

Multimodal Large Language Models (MLLMs) have made rapid progress in spatial intelligence, yet existing spatial reasoning benchmarks largely assume pristine visual inputs and overlook the degradations that commonly occur in real-world deployment, such as motion blur, low light, adverse weather, lens distortion, and…

arXiv:2605.22536

Same Architecture, Different Capacity: Optimizer-Induced Spectral Scaling Laws [22]

Nandan Kumar Jha, Brandon Reagen

Scaling laws have made language-model performance predictable from model size, data, and compute, but they typically treat the optimizer as a fixed training detail. We show that this assumption misses a fundamental axis of representation scaling: how effectively the optimizer converts added FFN width into utilized…

arXiv:2605.21803

WorldKV: Efficient World Memory with World Retrieval and Compression [23]

Jung Yi, Minjae Kim, Paul Hyunbin Cho, Wooseok Jang, Sangdoo Yun

Autoregressive video diffusion models have enabled real-time, action-conditioned world generation. However, sustaining a persistent world, where revisiting a previously seen viewpoint yields consistent content, remains an open problem. Full KV-cache attention preserves this consistency but breaks real-time…

arXiv:2605.22718

Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention [24]

Ali Hatamizadeh, Yejin Choi, Jan Kautz

Linear attention replaces the unbounded cache of softmax attention with a fixed-size recurrent state, reducing sequence mixing to linear time and decoding to constant memory. The hard part is not just what to forget, but how to edit this compressed memory without scrambling existing associations. Delta-rule models…

arXiv:2605.22791

Swift Sampling: Selecting Temporal Surprises via Taylor Series [25]

Dahye Kim, Bhuvan Sachdeva, Karan Uppal, Naman Gupta, Vineeth N. Balasubramanian

While most frames in long-form video are redundant, the critical information resides in temporal surprises: moments where the actual visual features deviate from their predicted evolution. Inspired by the human brain's predictive coding, we introduce Swift Sampling, an elegant, training-free frame selection algorithm…

arXiv:2605.22678

Diversed Model Discovery via Structured Table Discovery [26]

Zhengyuan Dong, Renée J. Miller

Model cards describe model behavior through a mixture of textual descriptions and structured artifacts, including performance, configuration, and dataset tables. Existing model search systems rely predominantly on semantic similarity over text, which can produce homogeneous result sets and limit exploration of…

arXiv:2605.22766

Early identification of breakthrough technologies: Insights from science-driven innovations [27]

Catalyst breakthroughs in methane dry reforming: Employing machine learning for future advancements [28]

Leveraging blockchain and smart contracts to combat greenwashing in sustainable development [29]

🔒 Security & Crypto 2

Artificial intelligence and machine learning in cybersecurity: a deep dive into state-of-the-art techniques and future paradigms [30]

Securing the future: exploring post-quantum cryptography for authentication and user privacy in IoT devices [31]

Cluster Computing - With the emergence of quantum computers, traditional cryptographic methods are vulnerable to attacks, emphasizing the need for post-quantum cryptography to secure devices and...

⚛️ Quantum 1

Quantum machine learning: A comprehensive review of integrating AI with quantum computing for computational advancements [32]

🛡️ AI Safety 1

More Context, Larger Models, or Moral Knowledge? A Systematic Study of Schwartz Value Detection in Political Texts [33]

Víctor Yeste, Paolo Rosso

Detecting Schwartz values in political text is difficult because implicit cues often depend on surrounding arguments and fine-grained distinctions between neighboring values. We study when context and explicit moral knowledge help sentence-level value detection. Using the ValuesML/Touch{é} ValueEval format, we compare…

arXiv:2605.22641

🤖 Agents 27

Efficient Agentic Reasoning Through Self-Regulated Simulative Planning [34]

Mingkai Deng, Jinyu Hou, Lara Sá Neves, Varad Pimpalkhute, Taylor W. Killian

How should an agent decide when and how to plan? A dominant approach builds agents as reactive policies with adaptive computation (e.g., chain-of-thought), trained end-to-end expecting planning to emerge implicitly. Without control over the presence, structure, or horizon of planning, these systems dramatically…

arXiv:2605.22138

Lean Refactor: Multi-Objective Controllable Proof Optimization via Agentic Strategy Search [35]

Jialin Lu, Soonho Kong, Rodrigo Stehling, Kaiyu Yang, Zhangyang Wang

We present Lean Refactor, a plug-and-play retrieval-augmented agentic framework for multi-objective, controllable, and version-robust refactoring of Lean proofs. LLM-generated proofs are notoriously correct-but-verbose and brittle across library versions, yet existing refactoring works overlook three practical…

arXiv:2605.20244

GenEvolve: Self-Evolving Image Generation Agents via Tool-Orchestrated Visual Experience Distillation [36]

Sixiang Chen, Zhaohu Xing, Tian Ye, Xinyu Geng, Yunlong Lin

Open-ended image generation is no longer a simple prompt-to-image problem. High-quality generation often requires an agent to combine a model's internal generative ability with external resources. As requests become more diverse and demanding, we aim to develop a general image-generation agent that can self-evolve…

arXiv:2605.21605

ClinSeekAgent: Automating Multimodal Evidence Seeking for Agentic Clinical Reasoning [37]

Juncheng Wu, Letian Zhang, Yuhan Wang, Haoqin Tu, Hardy Chen

Large language models (LLMs) and agentic systems have shown promise for clinical decision support, but existing works largely assume that evidence has already been curated and handed to the model. Real-world clinical workflows instead require agents to actively seek, iteratively plan, and synthesize multimodal…

arXiv:2605.20176

One Sentence, One Drama: Personalized Short-Form Drama Generation via Multi-Agent Systems [38]

Yufei Shi, Weilong Yan, Naixuan Huang, Yucheng Chen, Chenyu Zhang

Existing approaches for digital short-drama production typically rely on one-shot LLM generated scripts and loosely coupled pipelines, which fail to satisfy three key requirements of short-drama generation: (1) narrative pacing, resulting in weak hooks, insufficient escalation, and unattractive endings; (2) spatial…

arXiv:2605.22144

Maestro: Reinforcement Learning to Orchestrate Hierarchical Model-Skill Ensembles [39]

Jinyang Wu, Guocheng Zhai, Ruihan Jin, Yuhao Shen, Zhengxi Lu

The proliferation of large language models (LLMs) and modular skills has endowed autonomous agents with increasingly powerful capabilities. Existing frameworks typically rely on monolithic LLMs and fixed logic to interface with these skills. This gives rise to a critical bottleneck: different LLMs offer distinct…

arXiv:2605.22177

Sensor2Sensor: Cross-Embodiment Sensor Conversion for Autonomous Driving [40]

Jiahao Wang, Bo Sun, Yijing Bai, Vincent Casser, Songyou Peng

Robust training and validation of Autonomous Driving Systems (ADS) require massive, diverse datasets. Proprietary data collected by Autonomous Vehicle (AV) fleets, while high-fidelity, are limited in scale, diversity of sensor configurations, as well as geographic and long-tail-behavioral coverage. In contrast,…

arXiv:2605.22809

π-Bench: Evaluating Proactive Personal Assistant Agents in Long-Horizon Workflows [41]

Haoran Zhang, Luxin Xu, Zhilin Wang, Runquan Gui, Shunkai Zhang

The rise of personal assistant agents, e.g., OpenClaw, highlights the growing potential of large language models to support users across everyday life and work. A core challenge in these settings is proactive assistance, since users often begin with underspecified requests and leave important needs, constraints, or…

arXiv:2605.14678

TerminalWorld: Benchmarking Agents on Real-World Terminal Tasks [42]

Zhaoyang Chu, Jiarui Hu, Xingyu Jiang, Pengyu Zou, Han Li

We introduce TerminalWorld, a scalable data engine that automatically reverse-engineers high-fidelity evaluation tasks from "in-the-wild" terminal recordings. Processing 80,870 terminal recordings, the engine yields a full benchmark of 1,530 validated tasks, spanning 18 real-world categories, ranging from short…

arXiv:2605.22535

ACC: Compiling Agent Trajectories for Long-Context Training [43]

Qisheng Su, Zhen Fang, Shiting Huang, Yu Zeng, Yiming Zhao

Recent development of agents has renewed demand for long-context reasoning capacity of LLMs. However, training LLMs for this capacity requires costly long-document curation or heuristic context synthesis. We observe that agents produce massive trajectories when solving problems, invoking tools and receiving…

arXiv:2605.21850

SaaSBench: Exploring the Boundaries of Coding Agents in Long-Horizon Enterprise SaaS Engineering [44]

Qingnan Ren, Shun Zou, Shiting Huang, Ziao Zhang, Kou Shi

As autonomous coding agents become capable of handling increasingly long-horizon tasks, they have gradually demonstrated the potential to complete end-to-end software development. Although existing benchmarks have recently evolved from localized code editing to from-scratch project generation, they remain confined to…

arXiv:2605.17526

Dissecting ThunderKittens, anatomy of a compact DSL for high-performance AI kernels [45]

Introduction Modern ML workloads depend heavily on custom GPU kernels. Even when a model is expressed as clean tensor operations, the performance almost a...

Lobste.rs

score by collisions, patch by panic [46]

Hello All I am the author of this post. A couple of weeks ago, my previous piece ('The 90 day disclosure policy is dead') ended up here (https://lobste.rs/s/qxkdgl/90_day_disclosure_policy_is_dead). The thread had some good comments specifically the critique that 'saying the model is broken is a complaint, not…

Lobste.rs

Secure Boot and CA Rollover - a heads-up for distributions [47]

Friends Matthew Garrett Jonathan McDowell Jo McIntyre Martin Michlmayr Andrew Mobbs Mike Pitt Daniel Silverstone Andy Simpkins Neil Williams

Lobste.rs

Megalodon: Mass GitHub Repo Backdooring via CI Workflows [48]

Megalodon: Mass GitHub Repo Backdooring via CI Workflows

Lobste.rs

Google API keys keep working after you delete them long enough to be exploited [49]

Deleting a Google API key doesn't revoke it immediately. Our testing found successful authentications up to 23 minutes after deletion, and Google has declined to fix it.

Lobste.rs

FTC to Require Cox Media Group to Pay Nearly $1million to Settle Charges They Deceived Customers About “Active Listening” AI-Powered Marketing Service [50]

FTC to Require Cox Media Group to Pay Nearly $1million to Settle Charges They Deceived Customers About “Active Listening” AI-Powered Marketing Service

Lobste.rs

ChatGPT Won't Let You Type Until Cloudflare Reads Your React State. I Decrypted the Program That Does It [51]

Edit April 2, 2026: I've been getting inbound interest from researchers wanting to run their own queries. The MCP integration I use for my own research lets you analyze live mobile telemetry continuously collected from real devices in the wild, directly from Claude. To access it reach out at buchodi@

Lobste.rs

Launch HN: Superset (YC P26) – IDE for the agents era [52]

Launch HN: Superset (YC P26) – IDE for the agents era

Hacker News

(https://github.com/superset-sh/superset) [53]

Code Editor for the AI Agents Era - Run an army of Claude Code, Codex, etc. on your machine - superset-sh/superset

Hacker News

(https://limitless-osint.com/) [54]

Master open-source intelligence through immersive training, hands-on challenges, and browser-based CTF competitions. Built by Stratir.

Hacker News

Show HN: How to analyze your LLM output – A behavioural health monitor for LLMs [55]

Show HN: How to analyze your LLM output – A behavioural health monitor for LLMs

Hacker News

U.S. to Award Nine Quantum-Computing Firms $2B and Take Equity Stakes [56]

U.S. to Award Nine Quantum-Computing Firms $2B and Take Equity Stakes

Hacker News

(https://www.wsj.com/tech/u-s-to-award-quantum-computing-firms-2-billion-and-take-equity-stakes-7382e6be) [57]

(https://www.wsj.com/tech/u-s-to-award-quantum-computing-firms-2-billion-and-take-equity-stakes-7382e6be)

Hacker News

US Government takes $2B equity stake in nine quantum computing firms [58]

US Government takes $2B equity stake in nine quantum computing firms

Hacker News

(https://arstechnica.com/gadgets/2026/05/us-government-takes-2-billion-equity-stake-in-nine-quantum-computing-firms/) [59]

Beneficiaries include startup backed by firm with links to the Trump family.

Hacker News

Scott Aaronson – The Truth About Quantum Computing [video] [60]

Scott Aaronson – The Truth About Quantum Computing [video]

Hacker News

🖼️ Vision & Imaging 9

AnyMo: Geometry-Aware Setup-Agnostic Modeling of Human Motion in the Wild [61]

Baiyu Chen, Zechen Li, Wilson Wongso, Lihuan Li, Xiachong Lin

As wearable and mobile devices become increasingly embedded in daily life, they offer a practical way to continuously sense human motion in the wild. But inertial signals are highly dependent on the sensing setup, including body location, mounting position, sensor orientation, device hardware, and sampling protocol.…

arXiv:2605.22715

Disentangling Sampling from Training Budget in Class-Imbalanced CT Body Composition Segmentation [62]

Iason Skylitsis, Dimitrios Karkalousos, Ivana Išgum

Class imbalance is a fundamental challenge in medical image segmentation, where frequent classes typically dominate training at the expense of rare classes. Loss-based approaches mitigate imbalance by reweighting the per-pixel loss within the batch, while sampling strategies control which images enter the batch. Yet…

arXiv:2605.20405

FashionLens: Toward Versatile Fashion Image Retrieval via Task-Adaptive Learning [63]

Haokun Wen, Xuemeng Song, Xinghao Xie, Xiaolin Chen, Xiangyu Zhao

Fashion image retrieval is a cornerstone of modern e-commerce systems. A unified framework that supports diverse query formats and search intentions is highly desired in practice. However, existing approaches focus on narrow retrieval tasks and do not fully capture such diversity. Therefore, in this work, we aim to…

arXiv:2605.22552

Q-ARVD: Quantizing Autoregressive Video Diffusion Models [64]

Siao Tang, Xinyin Ma, Gongfan Fang, Xingyi Yang, Xinchao Wang

Autoregressive video diffusion models (ARVDs) have emerged as a promising architecture for streaming video generation, paving the way for real-time interactive video generation and world modeling. Despite their potential, the substantial inference cost of ARVDs remains a major obstacle to practical deployment, making…

arXiv:2605.21072

Segment Anything with Motion, Geometry, and Semantic Adaptation for Complex Nonlinear Visual Object Tracking [65]

Deyi Zhu, Yuji Wang, Yong Liu, Yansong Tang, Bingyao Yu

Traditional visual object tracking (VOT) methods typically rely on task-specific supervised training, limiting their generalization to unseen objects and challenging scenarios with distractors, occlusion, and nonlinear motion. Recent vision foundation models, exemplified by SAM 2, learn strong video understanding…

arXiv:2605.22538

PhysX-Omni: Unified Simulation-Ready Physical 3D Generation for Rigid, Deformable, and Articulated Objects [66]

Ziang Cao, Yinghao Liu, Haitian Li, Runmao Yao, Fangzhou Hong

Simulation-ready physical 3D assets have emerged as a promising direction owing to their broad applicability in downstream tasks. However, most existing 3D generation methods either neglect physical properties or are limited to a single asset category, e.g., rigid, deformable, or articulated objects. To address these…

arXiv:2605.21572

SceneAligner: 3D-Grounded Floorplan Localization in the Wild [67]

Junhyeong Cho, Ruojin Cai, Hadar Averbuch-Elor

Many public buildings provide floorplans with a "you are here" indicator to help visitors orient themselves. Floorplan localization seeks to computationally replicate this capability by determining where visual observations were captured within a floorplan. However, existing methods typically assume controlled…

arXiv:2605.22581

FlowLong: Inference-time Long Video Generation via Manifold-constrained Tweedie Matching [68]

Jangho Park, Geon Yeong Park, Gihyun Kwon, Jong Chul Ye

Extending the generation horizon of video diffusion models to long sequences remains a long-standing and important challenge. Existing training-free approaches fall into two categories: extensions of bidirectional models, which are tightly coupled to specific architectures and suffer from quality degradation over long…

arXiv:2605.20910

Bernini: Latent Semantic Planning for Video Diffusion [69]

Bernini Team, Chenchen Liu, Junyi Chen, Lei Li, Lu Chi

Multimodal large language models (MLLMs) and diffusion models have each reached remarkable maturity: MLLMs excel at reasoning over heterogeneous multimodal inputs with strong semantic grounding, while diffusion models synthesize images and videos with photorealistic fidelity. We argue that these two families can be…

arXiv:2605.22344

⚡ Quick Summary

📰 News Headlines

📄 ArXiv Papers

🧠 LLMs 11

⚙️ Machine Learning 18

🔒 Security &amp; Crypto 2

⚛️ Quantum 1

🛡️ AI Safety 1

🤖 Agents 27

🖼️ Vision &amp; Imaging 9

🔗 All Sources

🔒 Security & Crypto 2

🖼️ Vision & Imaging 9