AI Papers · Daily Picks

Language-based world models enable agentic environment simulation across multiple domains and enhance general agent performance through scalable simulation and improved downstream task performance.

2606.24597 · ▲ 137 · Code

Large language model agents' memory systems have evolved into complex data management frameworks requiring systematic evaluation across multiple modules and workloads to understand their performance characteristics and trade-offs.

2606.24775 · ▲ 105 · Code

PlanBench-XL evaluates large language model agents' ability to plan and adapt in complex tool-rich environments with limited visibility and dynamic disruptions.

2606.22388 · ▲ 95 · Code

Wan-Streamer is a unified, end-to-end multimodal model that enables real-time audio-visual interaction through causal attention mechanisms and integrated processing of visual, audio, and text modalities.

2606.25041 · ▲ 95

EnterpriseClawBench presents a benchmark for enterprise agents based on real-world sessions with 852 reproducible tasks, emphasizing comprehensive evaluation metrics beyond single performance scores.

2606.23654 · ▲ 79 · Code

OpenRath introduces a PyTorch-like programming model for multi-agent systems using Session as a central runtime abstraction that enables explicit fork, merge, and replay operations while recording comprehensive execution state.

2606.19409 · ▲ 76 · Code

Grouped Query Experts (GQE) improves Transformer efficiency by selectively activating query heads based on token content while maintaining key-value cache benefits of grouped-query attention.

2606.20945 · ▲ 75

Agentic Data Tailoring paradigm uses learnable data processing to structure high-entropy multimodal streams, with DataClaw_0-9B model achieving robust alignment through SFT and GRPO on a novel benchmark.

2606.21337 · ▲ 73 · Code

A novel on-policy generative field distillation framework called DanceOPD is proposed to unify text-to-image generation, local editing, and global editing capabilities in flow-matching models through capability-specific routing and velocity-based training.

2606.27377 · ▲ 71

DomainShuttle enables open domain subject-driven text-to-video generation with high fidelity and flexibility across in-domain and cross-domain scenarios through domain-aware modeling and dual RoPE schemes.

2606.26058 · ▲ 64 · Code

NatureBench presents a cross-disciplinary benchmark of 90 scientific tasks derived from Nature publications to assess AI coding agents' ability to achieve discovery rather than just reproduction, revealing that current agents primarily rely on methodological translation rather than genuine scientific innovation.

2606.24530 · ▲ 60 · Code

World Action Models are predictive-action systems that generate future states for decision-making, with designs balancing representational richness against computational constraints.

2606.20781 · ▲ 54 · Code

EDV is a three-stage framework that uses multiple heterogeneous agents to collaboratively construct reliable experiences for LLM agents, preventing self-confirmatory errors through execute-distill-verify processes.

2606.24428 · ▲ 51

ICWM enables robot policies to infer system variables from self-generated interactions, allowing adaptation to novel configurations without parameter updates by treating system identification as an in-context adaptation problem.

2606.26025 · ▲ 50

KaLM-Reranker-V1 is a fast reranker that decouples query and passage computation using encoder-decoder architecture with Matryoshka embedding pooling and cross-attention for efficient relevance modeling.

2606.22807 · ▲ 47

On-policy skill distillation framework extracts dense hindsight supervision from completed trajectories to improve language agent training efficiency and performance.

2606.26790 · ▲ 46 · Code

Researchers developed a new benchmark and dataset for photography assistance, along with a unified multimodal model that provides both composition guidance and pose recommendations during image capture.

2606.25763 · ▲ 45 · Code

An open-source data curation pipeline for training agentic language models is presented, demonstrating superior performance through systematic experimentation and scalable training data.

2606.24855 · ▲ 45

MemGUI-Agent addresses long-horizon mobile GUI task limitations through proactive context management using Context-as-Action (ConAct) to maintain critical information across extended sequences.

2606.19926 · ▲ 42 · Code

MobileForge enables efficient adaptation of mobile GUI agents through annotation-free learning by combining real app interaction grounding with hierarchical feedback-guided policy optimization.

2606.19930 · ▲ 42 · Code

A unified agentic framework called Qwen-Image-Agent is proposed to address the context gap in text-to-image generation by progressively constructing complete generation context through planning, reasoning, searching, and memory mechanisms.

2606.26907 · ▲ 42

Verification challenges in AI agents arise from the difficulty of aligning proxy signals with human intent, requiring adaptive verification systems that evolve alongside generative capabilities.

2606.26300 · ▲ 41

Masked diffusion language models with fully bidirectional attention outperform autoregressive counterparts on various benchmarks while maintaining competitiveness with established models.

2606.25331 · ▲ 41

Unlimited OCR introduces Reference Sliding Window Attention to eliminate growing memory consumption during long-sequence OCR tasks, enabling efficient transcription of multiple pages in a single forward pass.

2606.23050 · ▲ 38 · Code

This survey explores multimodal code intelligence systems that generate and reason with code based on visual inputs, categorizing approaches across GUI, scientific visualization, structured graphics, and emerging frameworks while identifying verification-centered research directions.

2606.15932 · ▲ 38 · Code

ViQ presents a visual quantization framework that balances semantic richness and detail preservation in discrete representations, enabling efficient multimodal training with native-resolution inputs.

2606.27313 · ▲ 38 · Code

A novel-view video synthesis method that enhances motion-aware diffusion models through multi-view point tracking supervision to improve geometric consistency and motion fidelity.

2606.26087 · ▲ 34 · Code

EvoEmbedding is a dynamic embedding model that generates adaptive representations by maintaining a continuously updated latent memory, enabling improved retrieval performance in long-context scenarios.

2606.21649 · ▲ 32 · Code

JetSpec is a speculative decoding framework that combines efficient forward drafting with causal conditioning to improve LLM inference speed and acceptance rates across various benchmarks.

2606.18394 · ▲ 31 · Code

AOHP presents an Android-based operating system framework that treats AI agents as first-class entities, enhancing task completion rates and reducing execution costs through specialized agent-oriented mechanisms.

2606.23449 · ▲ 30 · Code

A novel label-free framework for visual reasoning called V-Zero is presented, which uses contrastive evidence gating to improve fine-grained visual reasoning without requiring annotated answer labels, achieving faster training than traditional methods.

2606.25319 · ▲ 26 · Code

UnityShots is a memory-driven audio-video generation system that maintains consistent subject appearance and audio across video cuts using fixed-size long-term and short-term memory slots with boundary-conditioned gates and discrete cut-type priors.

2606.21661 · ▲ 24 · Code

BioMatrix is a novel multimodal foundation model that integrates molecular sequences, structures, and natural language into a unified decoder-only architecture for diverse biological tasks.

2606.22138 · ▲ 24 · Code

Autoregressive generation in large language models traditionally uses the final layer for token prediction, but a new decoding strategy dynamically selects more reliable intermediate layers based on entropy-guided search, improving reasoning performance with minimal computational overhead.

2606.21906 · ▲ 24 · Code

Fast-LeWM accelerates visual planning by replacing autoregressive rollout with parallel action-prefix prediction, reducing computational costs and latency accumulation during long-horizon predictions.

2606.26217 · ▲ 23 · Code

A large-scale multi-agent benchmark for evaluating LLMs in Chinese psychiatric diagnosis is introduced, highlighting challenges in dynamic consultation and the gap between consultation quality and diagnostic accuracy.

2602.09379 · ▲ 23 · Code

EBench is a comprehensive simulation benchmark for evaluating generalist mobile manipulation policies across diverse tasks and dimensions, revealing distinct capability profiles and generalization patterns among state-of-the-art models.

2606.18239 · ▲ 15 · Code

A novel RL training approach for terminal agents achieves superior performance using a simplified recipe and expanded dataset, enabling effective training with fewer parameters than previous methods.

2606.23321 · ▲ 13 · Code

A novel online data mixing framework called Holistic Data Scheduler uses reinforcement learning with a multi-objective reward function to optimize large language model pre-training efficiency and performance.

2606.24133 · ▲ 11 · Code

Language models should assist causal discovery workflows by providing contextual support and explanations rather than generating causal conclusions, as demonstrated through a platform that integrates data analysis and expert knowledge.

2606.23608 · ▲ 7 · Code

Efficient sampling of molecular systems at thermodynamic equilibrium is a hallmark challenge in statistical physics. This challenge has driven the development of Boltzmann Generators (BGs), which allow rapid generation of uncorrelated equilibrium samples by combining a generative model with exact likelihoods and an importance sampling correction. However, modern BGs predominantly rely on normalizing flows (NFs), which either suffer from limited expressivity due to strict invertibility constraints (discrete time) or computationally expensive likelihoods (continuous time). In this paper, we propose Autoregressive Boltzmann Generators (ArBG) -- a novel autoregressive modelling framework -- that overcomes these limitations by departing from the flow-based BG paradigm. ArBG circumvents the topological constraints of flows and enables sequential inference-time interventions, while offering enhanced scalability by leveraging architectures effective in Large Language Models. We empirically demonstrate that ArBG leads to significant improvements over flow-based models across all benchmarks, but particularly in larger peptide systems such as the 10-residue Chignolin. Furthermore, we introduce Robin, a 132 million parameter transferable model trained with the ArBG framework which improves over the previous state-of-the-art, reducing the zero-shot energy error, E-W$_2$, on 8-residue systems by over 60$\%$. The code can be found at the following link: https://github.com/danyalrehman/autobg.

2606.27361v1

Neural surrogate models offer fast approximate mappings from PDE parameters to solutions, but they typically treat solving as a purely statistical task: once trained, they struggle to correct their own constraint violations and extrapolate beyond the training distribution. Recent hybrid methods promote physical correctness by targeting the PDE residual via gradient descent or Gauss--Newton steps, but inherit the compute cost and instability of the underlying classical optimizers. We show, theoretically and empirically, that numerically minimizing the PDE residual can be an unreliable proxy for reconstruction accuracy in ill-conditioned systems, explaining why these methods often do not make accurate predictions despite achieving low residuals. We propose error-conditioned Neural Solvers (ENS), built on a different principle: rather than an optimization target, the PDE residual field is passed as a direct input to the network at each iteration, enabling it to read the spatial structure of its own errors and learn an update policy to iteratively correct its predictions. Across four PDE families, ENS attains the highest prediction accuracy in the large majority of settings, with gains reaching $10\times$ on turbulent Kolmogorov flow, while avoiding the expensive compute cost of hybrid methods. ENS's learned correction policy generalizes under distribution shift, including zero-shot parameter changes and cross-equation transfer, where its relative advantage is largest in the ill-conditioned regimes where residual minimization is least reliable. Project website: https://neuralsolver.github.io/.

2606.27354v1

Entity Matching (EM) is a core operation in the data integration pipeline, where records from different sources are compared to determine whether they refer to the same real-world entity. Recent work has incorporated domain information and low-resource learning techniques to better adapt EM systems to realistic settings. While these approaches have demonstrated strong performance, it remains unclear how they behave under varying data constraints and levels of supervision in practice. In this paper, we investigate a state-of-the-art method for low-resource, domain-aware EM--BEACON--and study how its performance is affected by different algorithmic choices and data availability conditions. We conduct a series of targeted experiments to evaluate these variations, providing deeper insight into the role of distribution alignment and the behavior of the BEACON framework.

2606.27342v1

Multimodal web agents can assist humans in operating repetitive GUI tasks, where effective task planning is essential for decomposing complex tasks into executable actions. While small open source MLLMs are cost efficient and privacy preserving compared with commercial large models, they suffer from weak planning and limited cross website generalization. To address these limitations, we introduce the planning experience exploration and utilization (PEEU) method, which autonomously explores environments to discover experiences and utilizes hindsight experience to synthesize strictly aligned, high level training data. To quantitatively analyze the generalization behaviors driving this performance, we propose the task decomposition hierarchical analysis framework (TDHAF) to systematically study compositional generalization across three task granularities: low, middle and high levels. Our analysis reveals that mastering low level atomic skills does not guarantee high level planning competence, while high level task training yields stronger OOD generalization. Experiments on real world benchmarks demonstrate PEEU's superior effectiveness: our 7B model achieves 30.6% accuracy, outperforming the much larger Qwen2.5-VL-32B model. These demonstrate constructing hindsight high level tasks and leveraging experiences is crucial for OOD planning abilities of small MLLMs.

2606.27330v1

Modern image generation demands a single model that unifies diverse capabilities, including text-to-image (T2I), local editing, and global editing. However, these capabilities are rarely naturally aligned and often conflict. For instance, editing tends to degrade T2I performance, while global and local editing interfere with each other. Consequently, effectively composing these capabilities has become a central challenge for image generation model training. To tackle this, we introduce DanceOPD, an on-policy generative field distillation framework for flow-matching models that routes each sample to one capability field, queries one low-noise student-induced state, and trains with a simple velocity MSE objective. With each capability source defined as a velocity field over the shared flow state space, the student learns from fields queried on its own rollout states to compose expert capabilities. This formulation also absorbs operator-defined fields such as classifier-free guidance. Comprehensive experiments on T2I, editing, realism-field absorption, and CFG absorption show that our approach improves multi-capability composition, strengthening target capabilities while preserving anchor generation quality. We believe this work establishes a practical route for generative field distillation in flow-matching models.

2606.27377v1

Whether political elites organise into rent-seeking coalitions that capture public resources or civic networks that sustain governance is a central question in comparative politics. Yet observing these complex, informal, and adversarial ties at scale has historically required intensive manual coding, while automated text-as-data methods have largely been limited to simple co-occurrence. Recent large language model (LLM) approaches offer a path forward but often rely on proprietary APIs, lack cross-lingual capability, and struggle with scalable entity resolution. We present a modular, fully open-weight pipeline for multilingual joint entity-relation extraction that builds signed, temporal knowledge graphs from massive unstructured news corpora. It combines span-based named-entity recognition (NER) with a three-stage linking cascade mapping mentions to language-independent Wikidata identifiers; a high-throughput, ontology-constrained mixture-of-experts model then uses guided decoding to extract directed, signed relationships grounded in a domain ontology. A full-coverage spot-check against a 3491-relation gold standard shows high textual correctness (68.2% strict to 93.7% lenient). Two large-scale case studies validate the pipeline against the public record. In Austria, it reconstructs a political party's complete lifecycle, dating internal fractures and tracking personnel into successor factions and court convictions. In a Polish corpus, it uncovers the overlapping economic and governance networks of state-enterprise patronage, alongside the structurally balanced, signed conflict network of the polarized Civic Platform (Platforma Obywatelska, PO)--Law and Justice (Prawo i Sprawiedliwość, PiS) duopoly. By bridging raw multilingual text and structured relational data, our framework provides a robust, replicable foundation for cross-national empirical computational social science.

2606.27347v1

Verifying the eligibility of securities as collateral is a key responsibility of the German Central Bank. However, manually verifying these assets against legal and financial criteria within lengthy, semi-structured, and often bilingual prospectuses is a resource-intensive task. While previous efforts utilized traditional Named Entity Recognition (NER) for information extraction, these methods can struggle with OCR noise, linguistic variance, and rigid span-based constraints, and the need for manually annotated training data for each relevant annotation type. In this paper, we present the first case study applying Large Language Models (LLMs) to the eligibility examination process, shifting the paradigm toward a generative Information Extraction pipeline. Our approach decomposes the task into extraction, normalization, and interpretation, allowing for greater flexibility in handling noisy text and interleaved German-English content. We further introduce a value-based evaluation methodology using LLM-as-a-judge, which offers a more semantic assessment than location-based metrics. Our results demonstrate that LLM-based systems achieve high precision (up to 91%) in document-level eligibility, exhibiting a conservative operating profile that minimizes false acceptance.

2606.27316v1

To avoid moderation and surveillance on social media, some users routinely invent indirect linguistic expressions (ILE) that camouflage sensitive meanings. Such expressions surface as algospeak, euphemisms, and adversarial obfuscation, depending on intent and context, and they involve recurring encoding mechanisms. We propose a comprehensive, mechanism-oriented taxonomy of ILE that abstracts away from communicative goals and instead categorizes the underlying operations through which meaning is encoded and recovered. We evaluate the taxonomy by incorporating it into LLM prompts and comparing it with four existing taxonomies and a no-taxonomy baseline, using 2,000 manually annotated TikTok and Bluesky posts. The proposed taxonomy attains the strongest document- and span-level performance across the three LLMs, achieving an improvement of 4.7% in accuracy and 5.4% in F1 over the best-performing benchmark. The empirical results reveal the importance of a comprehensive, mechanism-oriented taxonomy as a stable scaffold for detecting emerging coded language and a useful input to content moderation. Disclaimer: This paper contains content that may be profane, vulgar, or offensive.

2606.27314v1

Most unified large multimodal models (LMMs) that support both visual understanding and image generation still rely on curated post-training supervision, such as human annotations, preference labels, or external reward models. We ask whether a unified LMM can improve both abilities autonomously using only unlabeled images. We propose a self-evolving training framework with three internal roles: a Proposer that generates visual questions, a Solver that answers and evaluates them, and a Generator that synthesizes images. Training uses only self-derived consistency signals, without human annotations, preference labels, or task-trained external reward/judge models. To stabilize learning, we introduce Solver Token Entropy (STE), a continuous difficulty signal based on token-level prediction uncertainty that remains useful even when sample-level consistency becomes unreliable. For image generation, we design a multi-scale internal evaluation scheme that combines question-answer fidelity scoring with cycle-consistent captioning. This creates a solver-mediated coupling, where better visual understanding enables more reliable generation assessment and stronger internal training signals. The framework preserves the same role decomposition, reward logic, and training schedule across diffusion-based BLIP3o, rectified-flow BAGEL, and autoregressive VARGPT-v1.1 architectures, requiring only each backbone's native prompting and generation interface. Across eight understanding metrics, our method consistently improves over the corresponding base models. On BAGEL, it achieves a $+3.5\%$ absolute gain on MMMU and improves GenEval image generation performance from $82\%$ to $85\%$. Code and models are publicly released.

2606.27376v1

Going beyond predicting robot actions, World Action Models (WAMs) can also generate future visual observations. We build on this generative capability to propose Recurrent Generative Replay (REGEN), a continual imitation learning framework that synthesizes pseudo-replay trajectories, enabling a robot policy to rehearse previously learned tasks without storing their original human demonstrations. During continual adaptation, REGEN recursively queries the WAM to synthesize pseudo-replay trajectories conditioned only on prior task instructions and current-task observations. Experiments in both simulation and real-world manipulation settings show that REGEN reduces catastrophic forgetting by up to $50\%$ relative to sequential fine-tuning, while approaching the performance of privileged experience replay methods that require access to real replay data. Finally, we analyze the factors limiting generated replay, identifying long-horizon visual degradation and action-observation inconsistency as the primary bottlenecks. Our results establish WAMs as a promising foundation for continual robot learning without stored demonstrations.

2606.27374v1