Monthly collection

April 2026

The papers that drew the most attention during this month of AI research.

Papers: 77
Citations: 953
Research labs: 4

Explore the year

View all months →

Aug2025 Sep2025 Oct2025 Nov2025 Dec2025 Jan2026 Feb2026 Mar2026 Apr2026 May2026 Jun2026 Jul2026 Aug2026

1,016papers
curated and indexed

77 papers published in April 2026

01

arXiv.org

GrandCode: Achieving Grandmaster Level in Competitive Programming via Agentic Reinforcement Learning

GrandCode is a multi-agent reinforcement learning (RL) system for competitive programming, introduced by the DeepReinforce Team. It orchestrates modules for hypothesis proposal, solution generation, summarization, and test-case generation, jointly optimized via post-training and online test-time RL. A key innovation is Agentic GRPO, which combines…

DeepReinforce Team, Xiaoya Li, Guoyin Wang, Songqiao Su, et al.

Published: Apr 2026
Upvotes: 639
Citations: 3

Read summary Original

02

arXiv.org

Adam's Law: Textual Frequency Law on Large Language Models

The paper proposes the Textual Frequency Law (TFL), which states that for LLMs, higher-frequency textual data should be preferred for both prompting and fine-tuning when meaning is kept constant. The framework includes Textual Frequency Distillation (TFD) to enhance frequency estimation via story completion, and Curriculum Textual Frequency Training (CTFT)…

Hongyuan Adam Lu, Z. L., Victor Wei, Zefan Zhang, et al.

Published: Apr 2026
Upvotes: 510
Citations: 3

Read summary Original

03

arXiv.org

Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability

This preprint challenges the narrative that supervised fine-tuning (SFT) memorizes while reinforcement learning generalizes, showing that cross-domain generalization in reasoning SFT is conditional on optimization dynamics, training data, and base-model capability. Using math-only long chain-of-thought (CoT) SFT on pretrained base models, the authors find…

Qihan Ren, Peng Wang, Ruikun Cai, Shuai Shao, et al.

Published: Apr 2026
Upvotes: 330
Citations: 3

Read summary Original

04

arXiv.org

SkillClaw: Let Skills Evolve Collectively with Agentic Evolver

SkillClaw is a framework for collective skill evolution in multi-user LLM agent ecosystems, addressing the problem of static skills that do not improve with experience. It aggregates interaction trajectories from multiple users, groups them by referenced skills, and uses an agentic evolver to refine existing skills, create new ones, or skip updates based…

Ziyu Ma, Shidong Yang, Yuxiang Ji, Xucong Wang, et al.

Published: Apr 2026
Upvotes: 295
Citations: 48

Read summary Original

05

arXiv.org

Recursive Multi-Agent Systems

RecursiveMAS is a recursive multi-agent framework that scales agent collaboration by treating the entire system as a unified latent-space recursive computation. It connects heterogeneous agents via lightweight RecursiveLink modules: inner links enable latent thought generation within each agent, and outer links transfer latent states across agents, forming…

Jiaru Zou, Rui Pan, Ruizhong Qiu, Pan Lu, et al.

Published: Apr 2026
Upvotes: 287
Citations: 3

Read summary Original

06

arXiv.org

ClawBench: Can AI Agents Complete Everyday Online Tasks?

ClawBench is a benchmark of 153 everyday online tasks across 144 live websites and 15 categories, focusing on write-heavy, state-changing workflows like purchases, reservations, and applications. Unlike sandboxed benchmarks, it evaluates agents on production sites with dynamic content and anti-bot defenses, using final-request interception to block…

Yuxuan Zhang, Yubo Wang, Yipeng Zhu, Penghui Du, et al.

Published: Apr 2026
Upvotes: 265
Citations: 19

Read summary Original

07

arXiv.org

Tstars-Tryon 1.0: Robust and Realistic Virtual Try-On for Diverse Fashion Items

Tstars-Tryon 1.0, developed by the Pailitao Team at Alibaba Group, is a commercial-scale virtual try-on system designed for robustness, realism, versatility, and efficiency. It handles challenging real-world cases like extreme poses, lighting variations, and motion blur, while preserving garment details and avoiding synthetic artifacts. The system supports…

Mengting Chen, Zhengrui Chen, Yongchao Du, Zuan Gao, et al.

Published: Apr 2026
Upvotes: 253
Citations: 2

Read summary Original

08

arXiv.org

WildDet3D: Scaling Promptable 3D Detection in the Wild

WildDet3D introduces a unified, geometry-aware architecture for open-vocabulary monocular 3D object detection that accepts text, point, and box prompts, and optionally incorporates depth cues at inference time. The model uses dual-vision encoders (image and RGBD) with a depth fusion module, a promptable detector, and a 3D detection head with unambiguous…

Weikai Huang, Jieyu Zhang, Sijun Li, Taoyang Jia, et al.

Published: Apr 2026
Upvotes: 248
Citations: 7

Read summary Original

09

arXiv.org

LLaDA2.0-Uni: Unifying Multimodal Understanding and Generation with Diffusion Large Language Model

LLaDA2.0-Uni is a unified discrete diffusion large language model (dLLM) that integrates multimodal understanding and generation in a single framework. Its architecture combines a SigLIP-VQ semantic tokenizer, a 16B MoE dLLM backbone, and a diffusion decoder. The tokenizer converts images into discrete semantic tokens, enabling block-level masked diffusion…

Inclusion AI, Tiwei Bie, Haoxing Chen, Tieyuan Chen, et al.

Published: Apr 2026
Upvotes: 243
Citations: 5

Read summary Original

10

arXiv.org

InCoder-32B-Thinking: Industrial Code World Model for Thinking

The paper introduces InCoder-32B-Thinking, a 32B-parameter code model trained to reason about industrial software development (chip design, GPU optimization, embedded systems, 3D modeling) by integrating an Error-driven Chain-of-Thought (ECoT) synthesis framework and an Industrial Code World Model (ICWM). ECoT generates reasoning traces by modeling…

Jian Yang, Wei Zhang, Jiajun Wu, Junhang Cheng, et al.

Published: Apr 2026
Upvotes: 239
Citations: 3

Read summary Original

11

arXiv.org

Video-MME-v2: Towards the Next Stage in Benchmarks for Comprehensive Video Understanding

Video-MME-v2 is a new benchmark for evaluating video multimodal large language models (MLLMs), designed to address the gap between inflated leaderboard scores and real-world capabilities. It introduces a progressive three-level hierarchy (visual information aggregation, temporal dynamics modeling, and complex reasoning) and a group-based evaluation…

Chaoyou Fu, Haozhi Yuan, Yuhao Dong, Yi-Fan Zhang, et al.

Published: Apr 2026
Upvotes: 236
Citations: 23

Read summary Original

12

arXiv.org

Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond

This preprint introduces a taxonomy for world models in agentic AI, organized along two axes: capability level (L1 Predictor, L2 Simulator, L3 Evolver) and governing-law regime (physical, digital, social, scientific). L1 models learn one-step transition operators; L2 composes them into multi-step, action-conditioned rollouts satisfying long-horizon…

Meng Chu, Xuan Billy Zhang, Kevin Qinghong Lin, Lingdong Kong, et al.

Published: Apr 2026
Upvotes: 232
Citations: 10

Read summary Original

13

arXiv.org

Heterogeneous Scientific Foundation Model Collaboration

The paper introduces Eywa, a heterogeneous agentic framework that integrates domain-specific foundation models (FMs) with large language model (LLM) agents to solve scientific tasks involving non-linguistic data like time series and tabular data. Eywa uses an FM-LLM 'Tsaheylu' interface, implemented via the Model Context Protocol, allowing LLMs to…

Zihao Li, Jiaru Zou, Feihao Fang, Xuying Ning, et al.

Published: Apr 2026
Upvotes: 222
Citations: 1

Read summary Original

14

arXiv.org

OpenWorldLib: A Unified Codebase and Definition of Advanced World Models

OpenWorldLib is a standardized inference framework for advanced world models, addressing the lack of a unified definition in the field. The paper proposes that a world model is a model or framework centered on perception, equipped with interaction and long-term memory capabilities, for understanding and predicting the complex world. It categorizes…

DataFlow Team, Bohan Zeng, Daili Hua, Kaixin Zhu, et al.

Published: Apr 2026
Upvotes: 203
Citations: 3

Read summary Original

15

arXiv.org

Self-Distilled RLVR

The paper analyzes on-policy self-distillation (OPSD) for LLM post-training, where a model serves as both teacher and student, with the teacher using privileged information. It proves that OPSD's information asymmetry creates an irreducible mutual information gap, causing privileged information leakage and performance degradation after early gains. To…

Chenxu Yang, Chuanyu Qin, Qingyi Si, Minghui Chen, et al.

Published: Apr 2026
Upvotes: 181
Citations: 56

Read summary Original

16

arXiv.org

HY-Embodied-0.5: Embodied Foundation Models for Real-World Agents

Tencent's HY-Embodied-0.5 is a family of vision-language foundation models designed for real-world embodied agents, bridging the gap between general VLMs and physical-world tasks. The suite includes an efficient 2B-activated-parameter model (MoT-2B) for edge deployment and a powerful 32B-activated-parameter model (MoE-A32B) for complex reasoning. Key…

Tencent Robotics X, HY Vision Team, :, Xumin Yu, et al.

Published: Apr 2026
Upvotes: 181
Citations: 12

Read summary Original

17

arXiv.org

From Context to Skills: Can Language Models Learn from Context Skillfully?

The paper introduces Ctx2Skill, a self-evolving framework that autonomously discovers, refines, and selects context-specific skills from complex contexts to improve language models' context learning, without human annotation or external feedback. It uses a multi-agent self-play loop with a Challenger generating tasks and rubrics, a Reasoner solving them,…

Shuzheng Si, Haozhe Zhao, Yu Lei, Qingyi Wang, et al.

Published: Apr 2026
Upvotes: 171
Citations: 15

Read summary Original

18

arXiv.org

AgentSPEX: An Agent SPecification and EXecution Language

AgentSPEX is a YAML-based specification and execution language for LLM-agent workflows, designed to overcome the limitations of reactive prompting and Python-coupled orchestration frameworks. It provides explicit control flow (typed steps, branching, loops, parallel execution), reusable submodules, and explicit context management. Workflows run in a…

Pengcheng Wang, Jerry Huang, Jiarui Yao, Rui Pan, et al.

Published: Apr 2026
Upvotes: 168
Citations: 6

Read summary Original

19

arXiv.org

Seedance 2.0: Advancing Video Generation for World Complexity

Seedance 2.0, released by ByteDance in early February 2026, is a native multimodal audio-video generation model that supports text, image, audio, and video inputs. It generates 4-15 second clips at 480p/720p, with a Fast version for low latency. The model excels in real-world complexity, multimodal reference and editing, high-fidelity binaural audio, and…

Team Seedance, De Chen, Liyang Chen, Xin Chen, et al.

Published: Apr 2026
Upvotes: 168
Citations: 78

Read summary Original

20

arXiv.org

The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook

This survey provides a unified overview of latent space in language-based models, arguing that it is emerging as a native computational substrate beyond explicit token-level processing. It addresses the fragmentation in the field by organizing research into five sequential perspectives: Foundation, Evolution, Mechanism, Ability, and Outlook. The survey…

Xinlei Yu, Zhangquan Chen, Yongbo He, Tianyu Fu, et al.

Published: Apr 2026
Upvotes: 152
Citations: 30

Read summary Original

21

arXiv.org

The Past Is Not Past: Memory-Enhanced Dynamic Reward Shaping

The paper introduces MEDS, a framework that enhances reinforcement learning for large language models by dynamically penalizing recurrent error patterns. It addresses the issue of reduced sampling diversity, where policies repeatedly generate similar erroneous behaviors. MEDS stores historical behavioral signals by reusing layer-wise logits from the final…

Yang Liu, Enxi Wang, Yufei Gao, Weixin Zhang, et al.

Published: Apr 2026
Upvotes: 144
Citations: 1

Read summary Original

22

arXiv.org

ClawGUI: A Unified Framework for Training, Evaluating, and Deploying GUI Agents

ClawGUI is an open-source framework that integrates online reinforcement learning (RL) training, standardized evaluation, and real-device deployment for GUI agents. It addresses three gaps: closed training pipelines, misaligned evaluation protocols, and a broken deployment loop. ClawGUI-RL provides the first open-source RL infrastructure supporting…

Fei Tang, Zhiqiong Lu, Boxuan Zhang, Weiming Lu, et al.

Published: Apr 2026
Upvotes: 143
Citations: 10

Read summary Original

23

arXiv.org

HY-World 2.0: A Multi-Modal World Model for Reconstructing, Generating, and Simulating 3D Worlds

HY-World 2.0 is an open-source multi-modal world model framework that unifies 3D world generation and reconstruction. For sparse inputs (text or single-view images), it generates navigable 3D Gaussian Splatting (3DGS) scenes via a four-stage pipeline: panorama generation (HY-Pano 2.0), trajectory planning (WorldNav), world expansion (WorldStereo 2.0), and…

Team HY-World, Chenjie Cao, Xuhui Zuo, Zhenwei Wang, et al.

Published: Apr 2026
Upvotes: 127
Citations: 19

Read summary Original

24

arXiv.org

MinerU2.5-Pro: Pushing the Limits of Data-Centric Document Parsing at Scale

MinerU2.5-Pro improves document parsing purely through data engineering and training strategy, keeping the 1.2B-parameter architecture of MinerU2.5 unchanged. The authors identify that state-of-the-art models share failure patterns on hard samples, indicating a data bottleneck rather than an architectural one. They build a Data Engine with three…

Bin Wang, Tianyao He, Linke Ouyang, Fan Wu, et al.

Published: Apr 2026
Upvotes: 125
Citations: 14

Read summary Original

25

arXiv.org

From Skills to Talent: Organising Heterogeneous Agents as a Real-World Company

The paper introduces OneManCompany (OMC), a framework that elevates multi-agent systems to an organizational level by decoupling agent capabilities from organizational structure. OMC encapsulates skills, tools, and configurations into portable 'Talents' that run on heterogeneous backends via 'Containers', with six typed organizational interfaces. A…

Zhengxu Yu, Yu Fu, Zhiyuan He, Yuxuan Huang, et al.

Published: Apr 2026
Upvotes: 125
Citations: 0

Read summary Original

26

arXiv.org

GameWorld: Towards Standardized and Verifiable Evaluation of Multimodal Game Agents

GameWorld is a benchmark for evaluating multimodal large language models (MLLMs) as game agents in browser environments. It includes 34 diverse games across five genres (Runner, Arcade, Platformer, Puzzle, Simulation) and 170 tasks, each with outcome-based, state-verifiable metrics computed from serialized gameAPI state. The benchmark supports two agent…

Mingyu Ouyang, Siyuan Hu, Kevin Qinghong Lin, Hwee Tou Ng, et al.

Published: Apr 2026
Upvotes: 123
Citations: 10

Read summary Original

27

Research paper

Claw-Eval: Towards Trustworthy Evaluation of Autonomous Agents

Claw-Eval is an end-to-end evaluation suite for LLM-based autonomous agents, addressing gaps in trajectory-opaque grading, underspecified safety/robustness, and narrow task coverage. It comprises 300 human-verified tasks across 9 categories (General, Multimodal, Multi-turn Dialogue) with 2,159 rubric items. Each run is audited via three evidence channels:…

Bowen Ye, Rang Li, Qibin Yang, Yuanxin Liu, et al.

Published: Apr 2026
Upvotes: 122
Citations: 38

Read summary Original

28

arXiv.org

World-R1: Reinforcing 3D Constraints for Text-to-Video Generation

World-R1 is a framework that improves 3D consistency in text-to-video generation by using reinforcement learning (RL) instead of architectural changes. It builds on the Wan 2.1 model and uses Flow-GRPO to optimize the model with rewards from 3D foundation models (Depth Anything 3) and vision-language models (Qwen3-VL). The reward system includes meta-view…

Weijie Wang, Xiaoxuan He, Youping Gu, Yifan Yang, et al.

Published: Apr 2026
Upvotes: 119
Citations: 7

Read summary Original

29

NVIDIA

TriAttention: Efficient Long Reasoning with Trigonometric KV Compression

TriAttention is a KV cache compression method for long-context LLM reasoning. It exploits the observation that pre-RoPE Q/K vectors are highly concentrated around fixed non-zero centers, which remain stable across positions and contexts. This concentration causes attention to follow predictable distance preferences, which can be modeled as a trigonometric…

Weian Mao, Xi Lin, Wei Huang, Yuxin Xie, et al.

Published: Apr 2026
Upvotes: 117
Citations: 14

Read summary Original

30

arXiv.org

When Numbers Speak: Aligning Textual Numerals and Visual Instances in Text-to-Video Diffusion Models

The paper introduces NUMINA, a training-free framework for improving numerical alignment in text-to-video diffusion models, which often fail to generate the correct number of objects specified in prompts. NUMINA uses an identify-then-guide paradigm: first, it identifies prompt-layout inconsistencies by selecting discriminative self- and cross-attention…

Zhengyang Sun, Yu Chen, Xin Zhou, Xiaofan Li, et al.

Published: Apr 2026
Upvotes: 116
Citations: 0

Read summary Original

31

arXiv.org

Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe

This paper systematically investigates on-policy distillation (OPD) for large language models, identifying conditions for success, the token-level mechanism, and practical remedies. The authors find that OPD succeeds only when the student and teacher share compatible thinking patterns and the teacher offers genuinely new knowledge beyond the student's…

Yaxuan Li, Yuxin Zuo, Bingxiang He, Jinqian Zhang, et al.

Published: Apr 2026
Upvotes: 114
Citations: 145

Read summary Original

32

Z.ai / GLM

GLM-5V-Turbo: Toward a Native Foundation Model for Multimodal Agents

GLM-5V-Turbo is a native multimodal foundation model for agentic tasks, integrating perception, reasoning, planning, and execution. It introduces CogViT, a vision encoder trained via distillation and contrastive learning, and Multimodal Multi-Token Prediction (MMTP) using a shared <|image|> token for efficiency. The model undergoes joint RL over 30+ task…

GLM-V Team, :, Wenyi Hong, Xiaotao Gu, et al.

Published: Apr 2026
Upvotes: 113
Citations: 11

Read summary Original

33

arXiv.org

SKILL0: In-Context Agentic Reinforcement Learning for Skill Internalization

SKILL0 is a reinforcement learning framework that internalizes agent skills into model parameters, eliminating the need for inference-time skill retrieval. It uses in-context RL during training, where skills are provided as visual context and progressively withdrawn via a Dynamic Curriculum that evaluates each skill's on-policy helpfulness. Experiments on…

Zhengxi Lu, Zhiyuan Yao, Jinyang Wu, Chengcheng Han, et al.

Published: Apr 2026
Upvotes: 103
Citations: 47

Read summary Original

34

arXiv.org

Generative World Renderer

The paper introduces a large-scale dataset for generative world rendering, curated from two AAA games (Cyberpunk 2077 and Black Myth: Wukong) to address the domain gap in inverse and forward rendering. The dataset includes over 4 million frames at 720p/30 FPS, with synchronized RGB and five G-buffer channels (depth, normals, albedo, metallic, roughness),…

Zheng-Hui Huang, Zhixiang Wang, Jiaming Tan, Ruihan Yu, et al.

Published: Apr 2026
Upvotes: 103
Citations: 1

Read summary Original

35

arXiv.org

MegaStyle: Constructing Diverse and Scalable Style Dataset via Consistent Text-to-Image Style Mapping

The paper introduces MegaStyle, a scalable data curation pipeline for constructing a large-scale, high-quality style dataset (MegaStyle-1.4M) with intra-style consistency and inter-style diversity. It leverages the consistent text-to-image style mapping of large generative models like Qwen-Image to generate style pairs from curated prompts. The pipeline…

Junyao Gao, Sibo Liu, Jiaxing Li, Yanan Sun, et al.

Published: Apr 2026
Upvotes: 103
Citations: 2

Read summary Original

36

arXiv.org

RationalRewards: Reasoning Rewards Scale Visual Generation Both Training and Test Time

The paper introduces RationalRewards, a reasoning-based reward model for visual generation that produces structured, multi-dimensional critiques before assigning scores, unlike traditional scalar reward models. It is trained using Preference-Anchored Rationalization (PARROT), a variational framework that recovers rationales from preference data via…

Haozhe Wang, Cong Wei, Weiming Ren, Jiaming Liu, et al.

Published: Apr 2026
Upvotes: 103
Citations: 7

Read summary Original

37

arXiv.org

KnowRL: Boosting LLM Reasoning via Reinforcement Learning with Minimal-Sufficient Knowledge Guidance

KnowRL is a reinforcement learning (RL) training framework that addresses reward sparsity in large language model (LLM) reasoning by treating hint design as a minimal-sufficient guidance problem. Instead of injecting long solution prefixes or abstract templates, KnowRL decomposes guidance into atomic knowledge points (KPs) and selects compact,…

Linhao Yu, Tianmeng Yang, Siyu Ding, Renren Jin, et al.

Published: Apr 2026
Upvotes: 102
Citations: 8

Read summary Original

38

arXiv.org

FORGE: Fine-grained Multimodal Evaluation for Manufacturing Scenarios

FORGE is a benchmark for evaluating Multimodal Large Language Models (MLLMs) in manufacturing scenarios. It introduces a dataset combining real-world 2D images and 3D point clouds with fine-grained annotations like exact model numbers. The benchmark includes three tasks: Workpiece Verification (WORKVERI), Structural Surface Inspection (SURFINSP), and…

Xiangru Jian, Hao Xu, Wei Pang, Xinjian Zhao, et al.

Published: Apr 2026
Upvotes: 98
Citations: 2

Read summary Original

39

arXiv.org

Extending One-Step Image Generation from Class Labels to Text via Discriminative Text Representation

This paper extends MeanFlow, a one-step generation framework originally designed for class-label conditioning, to flexible text-to-image (T2I) generation. The authors find that directly integrating LLM-based text encoders into MeanFlow with standard training fails, and they identify that high-quality text representations must possess strong semantic…

Chenxi Zhao, Chen Zhu, Xiaokun Feng, Aiming Hao, et al.

Published: Apr 2026
Upvotes: 96
Citations: 2

Read summary Original

40

Research paper

Xiaomi OneVL: One-Step Latent Reasoning and Planning with Vision-Language Explanation

OneVL is a unified Vision-Language-Action (VLA) and World Model framework for autonomous driving that performs one-step latent reasoning and planning with vision-language explanations. It addresses the latency of explicit Chain-of-Thought (CoT) reasoning by compressing reasoning into compact latent tokens supervised by dual auxiliary decoders: a language…

Jinghui Lu, Jiayi Guan, Zhijian Huang, Jinlong Li, et al.

Published: Apr 2026
Upvotes: 96
Citations: 3

Read summary Original

41

arXiv.org

Visual Generation in the New Era: An Evolution from Atomic Mapping to Agentic World Modeling

This roadmap paper argues that visual generation must evolve from appearance synthesis to intelligent visual generation, grounded in structure, dynamics, and causal relations. It proposes a five-level taxonomy—Atomic, Conditional, In-Context, Agentic, and World-Modeling Generation—to organize progress from passive rendering to interactive, world-aware…

Keming Wu, Zuhao Yang, Kaichen Zhang, Shizun Wang, et al.

Published: Apr 2026
Upvotes: 92
Citations: 4

Read summary Original

42

arXiv.org

Programming with Data: Test-Driven Data Engineering for Self-Improving LLMs from Raw Corpora

The paper introduces Programming with Data (ProDa), a paradigm that maps the LLM data-engineering lifecycle onto software development by using a shared three-level knowledge structure (L1 concepts, L2 relations, L3 reasoning chains) extracted from raw corpora. This structure serves as the common foundation for both training data synthesis and benchmark…

Chenkai Pan, Xinglong Xu, Yuhang Xu, Yujun Wu, et al.

Published: Apr 2026
Upvotes: 91
Citations: 1

Read summary Original

43

arXiv.org

Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligence

Agent-World is a self-evolving training arena for general-purpose AI agents, developed by Renmin University of China and ByteDance Seed. It addresses two key bottlenecks in agent training: scalable realism in environment synthesis and continuous self-improvement. The system has two main components: (1) Agentic Environment-Task Discovery, which autonomously…

Guanting Dong, Junting Lu, Junjie Huang, Wanjun Zhong, et al.

Published: Apr 2026
Upvotes: 89
Citations: 14

Read summary Original

44

arXiv.org

CoInteract: Physically-Consistent Human-Object Interaction Video Synthesis via Spatially-Structured Co-Generation

CoInteract is an end-to-end framework for speech-driven human-object interaction (HOI) video synthesis, conditioned on a person reference image, a product reference image, text prompts, and speech audio. It addresses two common failures in diffusion models: structural instability in hands and faces, and physically implausible contact (e.g., hand-object…

Xiangyang Luo, Xiaozhe Xin, Tao Feng, Xu Guo, et al.

Published: Apr 2026
Upvotes: 88
Citations: 0

Read summary Original

45

arXiv.org

OpenGame: Open Agentic Coding for Games

OpenGame is an open-source agentic framework for end-to-end web game creation from natural-language specifications. It introduces Game Skill, comprising a Template Skill that evolves a library of project skeletons and a Debug Skill that maintains a living protocol of verified fixes, to address cross-file inconsistencies and integration errors. The…

Yilei Jiang, Jinyuan Hu, Qianyin Xiao, Yaozhi Zheng, et al.

Published: Apr 2026
Upvotes: 85
Citations: 5

Read summary Original

46

Annual Meeting of the Association for Computational Linguistics

LLaTiSA: Towards Difficulty-Stratified Time Series Reasoning from Visual Perception to Semantics

The paper introduces LLATISA, a Vision-Language Model (VLM) for time series reasoning (TSR), and HITSR, a hierarchical dataset of 83k samples. The authors formalize a four-level TSR taxonomy (L1: Numerical Read-out, L2: Pattern Perception, L3: Semantic Reasoning, L4: Predictive Inference) to address fragmented task definitions and unreliable benchmarks.…

Yueyang Ding, HaoPeng Zhang, Rui Dai, Yi Wang, et al.

Published: Apr 2026
Upvotes: 84
Citations: 1

Read summary Original

47

arXiv.org

MiniCPM-o 4.5: Towards Real-Time Full-Duplex Omni-Modal Interaction

MiniCPM-o 4.5 is a 9B-parameter open-source multimodal large language model (MLLM) designed for real-time full-duplex omni-modal interaction, enabling simultaneous perception and response. It introduces Omni-Flow, a unified streaming framework that aligns multimodal inputs and outputs along a shared temporal axis, converting turn-based interaction into a…

Junbo Cui, Bokai Xu, Chongyi Wang, Tianyu Yu, et al.

Published: Apr 2026
Upvotes: 83
Citations: 16

Read summary Original

48

arXiv.org

LPM 1.0: Video-based Character Performance Model

LPM 1.0 is a video-based character performance model that generates identity-consistent conversational videos in real time. It addresses the 'performance trilemma'—the challenge of achieving high expressiveness, real-time inference, and long-horizon identity stability simultaneously. The system comprises a 17B-parameter Diffusion Transformer (Base LPM)…

Ailing Zeng, Casper Yang, Chauncey Ge, Eddie Zhang, et al.

Published: Apr 2026
Upvotes: 82
Citations: 5

Read summary Original

49

arXiv.org

Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation

This survey is the first comprehensive review of Attention Sink (AS) in Transformers, a phenomenon where disproportionate attention is focused on a small set of uninformative tokens. The authors synthesize over 210 studies, organizing the field into three key dimensions: Fundamental Utilization (e.g., Sink Token Preservation, Attention Redistribution,…

Zunhai Su, Hengyuan Zhang, Wei Wu, Yifan Zhang, et al.

Published: Apr 2026
Upvotes: 82
Citations: 3

Read summary Original

50

arXiv.org

Near-Future Policy Optimization

The paper introduces Near-Future Policy Optimization (NPO), a mixed-policy reinforcement learning with verifiable rewards (RLVR) method that improves training by using trajectories from a near-future checkpoint of the same training run. The authors formalize a quality-variance trade-off, defining effective learning signal S = Q/V, where Q is the fraction…

Chuanyu Qin, Chenxu Yang, Qingyi Si, Naibin Gu, et al.

Published: Apr 2026
Upvotes: 77
Citations: 2

Read summary Original

51

Research paper

EXAONE 4.5 Technical Report

EXAONE 4.5 is LG AI Research's first open-weight vision-language model, integrating a custom 1.2B-parameter vision encoder with the EXAONE 4.0 32B language backbone. It supports six languages and a 256K token context, achieved via context extension during SFT. The model uses hybrid attention, GQA, 2D RoPE, and MTP for efficiency. Pre-training includes two…

Eunbi Choi, Kibong Choi, Sehyun Chun, Seokhee Hong, et al.

Published: Apr 2026
Upvotes: 75
Citations: 0

Read summary Original

52

arXiv.org

RADIO-ViPE: Online Tightly Coupled Multi-Modal Fusion for Open-Vocabulary Semantic SLAM in Dynamic Environments

RADIO-ViPE is an online, calibration-free semantic SLAM system that processes raw monocular RGB video to produce geometry-aware, open-vocabulary 3D grounding. It tightly couples multi-modal embeddings from agglomerative foundation models (RADIO/RADSeg) with geometric information within a dense bundle adjustment framework, using a factor graph with…

Zaid Nasser, Mikhail Iumanov, Tianhao Li, Maxim Popov, et al.

Published: Apr 2026
Upvotes: 75
Citations: 0

Read summary Original

53

arXiv.org

A Simple Baseline for Streaming Video Understanding

The paper introduces SIMPLESTREAM, a minimal baseline for streaming video understanding that feeds only the most recent N frames to an off-the-shelf VLM, without any memory, retrieval, or compression mechanisms. Evaluated on OVO-Bench and StreamingBench, SIMPLESTREAM with Qwen3-VL-8B and 4 frames achieves 67.7% average accuracy on OVO-Bench and 80.59% on…

Yujiao Shen, Shulin Tian, Jingkang Yang, Ziwei Liu

Published: Apr 2026
Upvotes: 74
Citations: 13

Read summary Original

54

arXiv.org

Large Language Models Explore by Latent Distilling

The paper introduces Exploratory Sampling (ESamp), a decoding method for large language models (LLMs) that promotes semantic diversity during generation. ESamp trains a lightweight Latent Distiller (LD) at test time to predict the LLM's deep-layer hidden representations from its shallow-layer ones. The prediction error serves as a novelty signal,…

Yuanhao Zeng, Ao Lu, Lufei Li, Zheng Zhang, et al.

Published: Apr 2026
Upvotes: 74
Citations: 0

Read summary Original

55

arXiv.org

HERMES++: Toward a Unified Driving World Model for 3D Scene Understanding and Generation

HERMES++ is a unified driving world model that integrates 3D scene understanding and future geometry prediction within a single framework. It uses a Bird's-Eye View (BEV) representation to consolidate multi-view spatial information, making it compatible with Large Language Models (LLMs). The model introduces LLM-enhanced world queries for knowledge…

Xin Zhou, Dingkang Liang, Xiwu Chen, Feiyang Tan, et al.

Published: Apr 2026
Upvotes: 74
Citations: 3

Read summary Original

56

Meta AI

Think in Strokes, Not Pixels: Process-Driven Image Generation via Interleaved Reasoning

This paper introduces process-driven image generation, a multi-step paradigm that decomposes text-to-image synthesis into an interleaved reasoning trajectory of textual planning and visual generation. The method uses a recurring four-stage cycle: Plan, Sketch, Inspect, and Refine, where the model generates incremental instructions and scene descriptions,…

Lei Zhang, Junjiao Tian, Zhipeng Fan, Kunpeng Li, et al.

Published: Apr 2026
Upvotes: 73
Citations: 3

Read summary Original

57

arXiv.org

OmniShow: Unifying Multimodal Conditions for Human-Object Interaction Video Generation

OmniShow is an end-to-end framework for Human-Object Interaction Video Generation (HOIVG), synthesizing videos conditioned on text, reference images, audio, and pose. It introduces Unified Channel-wise Conditioning to inject reference images and pose via channel concatenation with pseudo-frame tokens, and Gated Local-Context Attention for precise…

Donghao Zhou, Guisheng Liu, Hao Yang, Jiatong Li, et al.

Published: Apr 2026
Upvotes: 73
Citations: 3

Read summary Original

58

arXiv.org

Elucidating the SNR-t Bias of Diffusion Probabilistic Models

This paper identifies a Signal-to-Noise Ratio-timestep (SNR-t) bias in Diffusion Probabilistic Models (DPMs), where during inference the SNR of denoised samples becomes misaligned with their timestep due to accumulated prediction and discretization errors. The authors provide empirical evidence and theoretical proof showing that reverse-process samples…

Meng Yu, Lei Sun, Jianhao Zeng, Xiangxiang Chu, et al.

Published: Apr 2026
Upvotes: 73
Citations: 2

Read summary Original

59

Meta AI

Tuna-2: Pixel Embeddings Beat Vision Encoders for Multimodal Understanding and Generation

Tuna-2 is a native unified multimodal model that performs visual understanding and generation directly from raw pixel embeddings, eliminating pretrained vision encoders such as VAEs and representation encoders. It uses simple patch embedding layers to encode images and a single transformer decoder for joint processing, with pixel-space flow matching for…

Zhiheng Liu, Weiming Ren, Xiaoke Huang, Shoufa Chen, et al.

Published: Apr 2026
Upvotes: 71
Citations: 10

Read summary Original

60

arXiv.org

All Roads Lead to Rome: Incentivizing Divergent Thinking in Vision-Language Models

This paper investigates the behavioral differences between Reinforcement Learning (RL) and base Vision-Language Models (VLMs). The authors find that RL models, such as those trained with Group Relative Policy Optimization (GRPO), exhibit deeper but narrower reasoning, while base models show broader and more diverse thinking patterns. They identify a…

Xinyu Tian, Shu Zou, Zhaoyuan Yang, Mengqi He, et al.

Published: Apr 2026
Upvotes: 70
Citations: 0

Read summary Original

61

arXiv.org

RAGEN-2: Reasoning Collapse in Agentic RL

RAGEN-2 identifies a failure mode in multi-turn LLM agent reinforcement learning called template collapse, where reasoning appears diverse within inputs but becomes input-agnostic across inputs, invisible to entropy-based metrics. The authors decompose reasoning quality into within-input diversity (conditional entropy) and cross-input distinguishability…

Zihan Wang, Chi Gui, Xing Jin, Qineng Wang, et al.

Published: Apr 2026
Upvotes: 69
Citations: 10

Read summary Original

62

Qwen

OccuBench: Evaluating AI Agents on Real-World Professional Tasks via Language Environment Simulation

OCCUBENCH is a benchmark for evaluating AI agents on real-world professional tasks across 100 scenarios, 65 domains, and 10 industry categories. It uses Language Environment Simulators (LESs), where an LLM simulates domain-specific tool responses, enabling evaluation in domains without public environments. A multi-agent pipeline synthesizes 382 solvable,…

Xiaomeng Hu, Yinger Zhang, Fei Huang, Jianhong Tu, et al.

Published: Apr 2026
Upvotes: 69
Citations: 1

Read summary Original

63

arXiv.org

ReVSI: Rebuilding Visual Spatial Intelligence Evaluation for Accurate Assessment of VLM 3D Reasoning

The paper introduces ReVSI, a benchmark for evaluating vision-language models' (VLMs) 3D spatial reasoning, addressing validity issues in the existing VSI-Bench. Two key pitfalls are identified: annotation-to-video ground-truth drift (errors from point-cloud-based annotations) and scene-observability mismatch (questions unanswerable under sparse frame…

Yiming Zhang, Jiacheng Chen, Jiaqi Tan, Yongsen Mao, et al.

Published: Apr 2026
Upvotes: 69
Citations: 7

Read summary Original

64

arXiv.org

Co-Evolving Policy Distillation

The paper proposes Co-Evolving Policy Distillation (CoPD) to consolidate multiple expert capabilities into a single model. It analyzes two standard paradigms: mixed RLVR, which suffers from capability divergence cost, and the static pipeline of training experts then performing on-policy distillation (OPD), which fails to fully absorb teacher capabilities…

Naibin Gu, Chenxu Yang, Qingyi Si, Chuanyu Qin, et al.

Published: Apr 2026
Upvotes: 68
Citations: 3

Read summary Original

65

Proceedings of the Special Interest Group on Computer Graphics and Interactive Techniques Conference Conference Papers

Video Analysis and Generation via a Semantic Progress Function

This paper introduces the Semantic Progress Function (SPF), a one-dimensional representation that captures how semantic meaning evolves over time in video sequences. The SPF is computed by measuring pairwise semantic distances between frames using a pretrained embedder (SigLIP) and fitting a smooth curve via regularized least squares. Departures from a…

Gal Metzer, Sagi Polaczek, Ali Mahdavi-Amiri, Raja Giryes, et al.

Published: Apr 2026
Upvotes: 64
Citations: 1

Read summary Original

66

arXiv.org

SpatialEvo: Self-Evolving Spatial Intelligence via Deterministic Geometric Environments

SpatialEvo introduces a self-evolving framework for 3D spatial reasoning that replaces model consensus with deterministic geometric feedback. The core contribution is the Deterministic Geometric Environment (DGE), which defines validation rules for 16 spatial reasoning task categories and computes exact ground truth from point clouds and camera poses,…

Dinging Li, Yingxiu Zhao, Xinrui Cheng, Kangheng Lin, et al.

Published: Apr 2026
Upvotes: 63
Citations: 3

Read summary Original

67

arXiv.org

DiPO: Disentangled Perplexity Policy Optimization for Fine-grained Exploration-Exploitation Trade-Off

The paper introduces DiPO (Disentangled Perplexity Policy Optimization), a method for fine-grained exploration-exploitation trade-off in Reinforcement Learning with Verifiable Rewards (RLVR) for LLMs. It identifies two dilemmas in GRPO-based methods: extreme sample groups (all-correct or all-error) yield zero advantage, and perplexity (PPL) distribution…

Xiaofan Li, Ming Yang, Zhiyuan Ma, Shichao Ma, et al.

Published: Apr 2026
Upvotes: 62
Citations: 0

Read summary Original

68

arXiv.org

Qwen3.5-Omni Technical Report

Qwen3.5-Omni is a fully omnimodal large language model that scales to hundreds of billions of parameters and supports a 256k context length. It is pretrained on a massive dataset including over 100 million hours of audio-visual content. The model uses a Thinker-Talker architecture with Hybrid-Attention Mixture-of-Experts (MoE) for both components, enabling…

Qwen Team

Published: Apr 2026
Upvotes: 60
Citations: 87

Read summary Original

69

arXiv.org

Leveraging Verifier-Based Reinforcement Learning in Image Editing

The paper introduces Edit-R1, a framework for image editing that uses a verifier-based Reasoning Reward Model (RRM) trained with reinforcement learning. The RRM decomposes editing instructions into principles, evaluates edited images against each, and produces interpretable scores. Training involves a cold-start SFT phase with filtered chain-of-thought…

Hanzhong Guo, Jie Wu, Jie Liu, Yu Gao, et al.

Published: Apr 2026
Upvotes: 59
Citations: 5

Read summary Original

70

arXiv.org

Memory Intelligence Agent

The paper introduces the Memory Intelligence Agent (MIA), a framework for deep research agents that combines a Manager-Planner-Executor architecture with dual memory systems. The Memory Manager stores compressed historical trajectories as non-parametric memory, while the Planner, trained via alternating reinforcement learning, serves as parametric memory.…

Jingyang Qiao, Weicheng Meng, Yu Cheng, Zhihang Lin, et al.

Published: Apr 2026
Upvotes: 58
Citations: 4

Read summary Original

71

arXiv.org

VOID: Video Object and Interaction Deletion

VOID is a video object removal framework that generates physically plausible counterfactual videos when an object is removed, addressing limitations of existing methods that only handle photometric effects like shadows. It uses a two-pass approach: first, a video diffusion model (CogVideoX) synthesizes a counterfactual trajectory guided by a quadmask,…

Saman Motamed, William Harvey, Benjamin Klein, Luc Van Gool, et al.

Published: Apr 2026
Upvotes: 57
Citations: 4

Read summary Original

72

arXiv.org

Steerable Visual Representations

The paper introduces Steerable Visual Representations (SteerViT), a method to make pretrained Vision Transformers (ViTs) steerable by natural language. SteerViT injects text into the visual encoder via lightweight gated cross-attention layers (early fusion), unlike late-fusion models like CLIP. It is trained on a referential segmentation pretext task using…

Jona Ruthardt, Manu Gaur, Deva Ramanan, Makarand Tapaswi, et al.

Published: Apr 2026
Upvotes: 57
Citations: 2

Read summary Original

73

arXiv.org

Embarrassingly Simple Self-Distillation Improves Code Generation

The paper introduces Simple Self-Distillation (SSD), a method that improves LLM code generation by sampling solutions from the model itself with specific temperature and truncation settings, then fine-tuning on those raw, unverified outputs via standard supervised fine-tuning. SSD requires no verifier, teacher model, or reinforcement learning. It improves…

Ruixiang Zhang, Richard He Bai, Huangjie Zheng, Navdeep Jaitly, et al.

Published: Apr 2026
Upvotes: 56
Citations: 28

Read summary Original

74

ACM Transactions on Graphics

Strips as Tokens: Artist Mesh Generation with Native UV Segmentation

The paper introduces Strips as Tokens (SATO), a framework for generating artist-quality 3D meshes with native UV segmentation using autoregressive transformers. SATO uses a strip-based tokenization strategy that serializes meshes as connected chains of faces, preserving edge flow and structural regularity. This unified representation allows the same token…

Rui Xu, Dafei Qin, Kaichun Qiao, Qiujie Dong, et al.

Published: Apr 2026
Upvotes: 56
Citations: 2

Read summary Original

75

arXiv.org

CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discovery

CORAL is a framework for autonomous multi-agent evolution on open-ended problems, replacing fixed evolutionary search heuristics with long-running LLM agents that control retrieval, proposal, evaluation, and knowledge accumulation. It uses shared persistent memory (attempts, notes, skills), asynchronous multi-agent execution, and heartbeat-based…

Ao Qu, Han Zheng, Zijian Zhou, Yihao Yan, et al.

Published: Apr 2026
Upvotes: 55
Citations: 29

Read summary Original

76

arXiv.org

DMax: Aggressive Parallel Decoding for dLLMs

DMax is a new paradigm for diffusion language models (dLLMs) that enables aggressive parallel decoding while preserving generation quality by mitigating error accumulation. It reformulates decoding as a progressive self-refinement from mask embeddings to token embeddings, using two key components: On-Policy Uniform Training (OPUT) and Soft Parallel…

Zigeng Chen, Gongfan Fang, Xinyin Ma, Ruonan Yu, et al.

Published: Apr 2026
Upvotes: 54
Citations: 8

Read summary Original

77

arXiv.org

ClawGym: A Scalable Framework for Building Effective Claw Agents

ClawGym is a scalable framework for developing Claw-style personal agents, addressing the lack of systematic data synthesis, training, and evaluation in such environments. It introduces ClawGym-SynData, a dataset of 13.5K filtered tasks generated via persona-driven top-down and skill-grounded bottom-up pipelines, with realistic mock workspaces and hybrid…

Fei Bai, Huatong Song, Shuang Sun, Daixuan Cheng, et al.

Published: Apr 2026
Upvotes: 54
Citations: 3

Read summary Original