❌

Normal view

Received — 15 September 2026 ⏭

Reinforcement Learning over Patient Trajectories for Clinical Reasoning in EHR Foundation Models

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yuxin Xiao · Sheng Zhang · Chandan Singh · Tristan Naumann · Hoifung Poon · Jianfeng Gao · Xiaodong Liu
14 September 2026 at 12:00
arXiv:2609.12277v1 Announce Type: cross Abstract: Electronic health record (EHR) foundation models trained on longitudinal patient trajectories have demonstrated strong performance across diverse clinical prediction tasks. However, their clinical reasoning capabilities remain constrained by next-token prediction on limited and incomplete EHR data. To address this, we propose a reinforcement learning (RL) fine-tuning framework that treats EHR foundation models as generative policies over patient trajectories. We formulate common clinical prediction problems (e.g., hospital readmission) as event-conditioned, time-windowed reasoning tasks. We then design time-aware, rollout-sensitive rewards to account for finite rollout lengths and temporally inconclusive outcomes. We find that RL fine-tuning consistently improves over pre-trained backbones and strong baselines. Notably, it enables smaller models to surpass larger pre-trained models in data-limited regimes and induces positive transfer across tasks. Further analysis shows that RL fine-tuned models generate trajectories with stronger structural and semantic alignment to ground truth and greater downstream utility.

OpenFinGym: A Verifiable Multi-Task Gym Environment for Evaluating Quant Agents

✇cs.AI, q-bio.NC updates on arXiv.org
By: Kaicheng Zhang · Wen Ge · Lei Jiang · Weixin Yang · Jordan Langham-Lopez · Jialin Yu · Lukasz Szpruch · Hao Ni
14 September 2026 at 12:00
arXiv:2606.26350v2 Announce Type: replace Abstract: Although large language model agents are increasingly applied to quantitative-finance workflows, their evaluation remains fragmented across isolated tasks, while the financial relevance of benchmark tasks is often overlooked. Yet financial workflows are inherently multi-stage, spanning interdependent tasks such as forecasting, strategy construction, risk management, and trading. Existing platforms typically focus on a single task, and can therefore overstate agent competence and fail to reveal weaknesses in generalization, real-market interaction, and financially meaningful decision-making. We introduce OpenFinGym, a unified gym environment for quantitative-finance agent development that covers forecasting, market generation, real-time trading, and fraud detection under a single execution and verification interface. OpenFinGym additionally provides an automated task-construction pipeline that turns quantitative finance publications into executable task packages; a containerised runtime with a host-side verifier service that supports scalable agent rollouts and prevents runtime train-test leakage; a paper trading engine with a low-latency data-stream design; deferred-resolution support for long-horizon and event-market forecasts; and integration for SFT and RL post-training

AsyncFlow: An Asynchronous Streaming RL Framework for Efficient LLM Post-Training

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zhenyu Han · Ansheng You · Haibo Wang · Kui Luo · Guang Yang · Wenqi Shi · Menglong Chen · Sicheng Zhang · Zeshun Lan · Chunshi Deng · Huazhong Ji · Wenjie Liu · Yu Huang · Yixiang Zhang · Chenyi Pan · Jing Wang · Xin Huang · Chunsheng Li · Jianping Wu
14 September 2026 at 12:00
arXiv:2507.01663v2 Announce Type: replace-cross Abstract: Reinforcement learning (RL) has become a pivotal technology in the post-training phase of large language models (LLMs). Traditional task-collocated RL frameworks suffer from significant scalability bottlenecks, while task-separated RL frameworks face challenges in managing complex dataflows and resolving resource idling. Furthermore, most existing frameworks are tightly coupled with LLM training or inference engines, making them difficult to support custom-designed engines. To address these challenges, we propose AsyncFlow, an asynchronous streaming RL framework tailored for efficient post-training. Specifically, we introduce a distributed data storage and transfer module that provides panoramic data management and fine-grained scheduling capabilities in a fully streamed manner. This architecture inherently enables automated pipeline overlapping among RL tasks and dynamic load-balancing. Moreover, we propose an asynchronous producer-consumer workflow, which is engineered to minimize computational idleness by strategically deferring the parameter update process within staleness thresholds. Finally, the core capabilities of AsyncFlow are architecturally decoupled from underlying training and inference engines and encapsulated by service-oriented user interfaces, offering a modular and customizable user experience. Extensive experiments demonstrate an average throughput of 1.59x compared to the state-of-the-art baseline. The architecture presented in this work provides actionable insights for designing next-generation RL training systems.

FEAT: A Linear-Complexity Foundation Model for Extremely Large Structured Data

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zhenghang Song · Tang Qian · Lu Chen · Yushuai Li · Zhengke Hu · Bingbing Fang · Yumeng Song · Junbo Zhao · Sheng Zhang · Tianyi Li
14 September 2026 at 12:00
arXiv:2603.16513v4 Announce Type: replace-cross Abstract: Structured data is widely used in domains such as healthcare, finance, and scientific data management. Recent studies on structured data foundation models (SFMs) aim to support data analysis and mining tasks over such data, but still face scalability and generalization challenges when applied to real-world enterprise databases. First, many SFMs rely on full self-attention, which introduces an O(N^2) computational bottleneck and limits the number of tuples that can be processed jointly. Second, directly replacing attention with linear-complexity sequence models may conflict with the permutation-invariant nature of structured data, introducing artificial order bias and degrading representation quality. Moreover, models trained only on synthetic data may struggle to generalize to the heavy-tailed and heterogeneous distributions commonly found in real-world databases. To address these challenges, we propose FEAT, a linear-complexity foundation model for extremely large structured data. FEAT replaces quadratic attention with a multi-layer dual-axis encoding architecture. It integrates an adaptive-fusion bidirectional state-space model (AFBM) with convolutional gated linear attention (Conv-GLA), enabling cross-tuple contextualization in O(N) time while supporting permutation-invariant representation learning. To improve robustness under real-world data skewness, FEAT further adopts a hybrid structural causal pre-training pipeline with a robust reconstruction objective. Experiments on 12 real-world database benchmarks show that FEAT consistently outperforms representative SFMs on zero-shot tasks and scales linearly with structured-data sample length, achieving up to 50x faster inference latency.
Received — 13 September 2026 ⏭

Fibronectin 1 mediated histone lactylation promotes malignant progression of GIST regulated by m<sup>6</sup>A modification

✇Cell Death Discovery nature.com science feeds
By: Ge Zhang · Jiehan Li · Yuhan Yin · Wunan Mi · Hao Liu · Zhao Sun · Jiahao Liu · Yizheng Zhang · Meimei Jiang · Nannan Liu · Guiyun Jia · Yingjie Zhang · Yang Fu
11 September 2026 at 08:00

Cell Death Discovery, Published online: 11 September 2026; doi:10.1038/s41420-026-03338-x

Fibronectin 1 mediated histone lactylation promotes malignant progression of GIST regulated by m6A modification
Received — 10 September 2026 ⏭

Macrophage spatiotemporal plasticity in pulmonary diseases: decoding the niche at single-cell resolution

✇Pulmonary nodule
By: Chenchen Lin · Shaohui Huang · Li Yang · Ziqi Wang · Huanhuan Zhou · Quncheng Zhang · Xiaoju Zhang
3 July 2026 at 18:00

Front Immunol. 2026 Jun 18;17:1855906. doi: 10.3389/fimmu.2026.1855906. eCollection 2026.

ABSTRACT

Pulmonary gas exchange and host defense depend on the dynamic coordination of resident and recruited macrophage populations. Historically, macrophage functions have often been interpreted through the classic M1/M2 dichotomy; however, this binary framework does not capture the heterogeneity and context-dependent plasticity of macrophage states within the lung microenvironment. Advances in single-cell RNA sequencing and spatial multi-omics have substantially refined our understanding of this complex macrophage network. Here, we synthesize evidence from human studies and experimental models to summarize macrophage functional states in homeostasis and across chronic obstructive pulmonary disease, asthma, idiopathic pulmonary fibrosis, pulmonary hypertension, acute lung injury/acute respiratory distress syndrome, and lung cancer. We highlight how macrophage transcriptional programs are shaped by ontogeny, tissue niche, and epigenetic-metabolic regulation, and how these programs are linked to disease-specific remodeling of the pulmonary microenvironment. Across diverse respiratory diseases, persistent tissue injury and microenvironmental stress remodel resident macrophage programs and are frequently accompanied by the expansion and context-dependent differentiation of recruited monocyte-derived macrophages. These macrophage states are associated with inflammatory amplification, epithelial and endothelial barrier dysfunction, extracellular matrix remodeling, and tumor immune evasion. Ligand-receptor and spatial analyses further identify candidate communication axes linking macrophages with stromal, epithelial, endothelial, and immune cells, some of which appear partially conserved across disease contexts. Emerging macrophage-targeted strategies are increasingly being explored beyond broad depletion, with growing interest in context-specific reprogramming and niche modulation, including antibody-based, nanocarrier-mediated, and engineered-cell approaches. Decoding the spatiotemporal trajectories and cell-cell communication networks of specific macrophage subsets, while considering tissue context, species differences, and levels of experimental support, may help clarify mechanisms of tissue remodeling, therapeutic resistance, and macrophage-targeted intervention in complex pulmonary diseases.

PMID:42396453 | PMC:PMC13322945 | DOI:10.3389/fimmu.2026.1855906

Received — 27 May 2026 ⏭

Test-Time Self-Adaptive Conditioning for Stable Audio-Driven Talking-Head Generation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zhicheng Zhang · Lei Wang · Yu Zhang · Yongsheng Gao
26 May 2026 at 12:00
arXiv:2605.25488v1 Announce Type: cross Abstract: Audio-driven talking-head generation has achieved remarkable progress with recent models such as AniTalker, FLOAT, and Sonic. Despite their success, most existing approaches rely on a single static reference image to condition the entire video generation process at inference stage. This static conditioning paradigm often creates a mismatch between fixed identity features and dynamically evolving facial motion, leading to identity drift, temporal inconsistency, and degraded perceptual quality. We introduce Test-Time Self-Adaptive Conditioning (TT-SAC), a parameter-free inference framework that enables pretrained talking-head generators to adapt their conditioning representations during inference without retraining, gradient updates, or additional supervision. Instead of treating the reference portrait as immutable, TT-SAC composes the generator with its encoder in a feedback loop: the generator's own outputs are re-encoded to construct a refined conditioning representation that better aligns with the temporal dynamics of the synthesized sequence. A single adaptation step approximates a self-consistent equilibrium of the generative process, stabilizing identity and motion across time. We further provide theoretical analysis showing that test-time conditioning adaptation reduces feature variance and improves generative stability under mild Lipschitz assumptions, while exhibiting a principled bias-variance tradeoff that governs the optimal strength of adaptation. Extensive experiments on state-of-the-art talking-head generators and benchmark datasets demonstrate consistent improvements in lip-sync accuracy, temporal coherence, identity preservation, and perceptual fidelity. TT-SAC offers a model-agnostic and training-free strategy for enhancing generative video models, establishing test-time conditioning adaptation as an effective mechanism for stabilizing audio-driven portrait animation.

SAMark: A Self-Anchored Text Watermarking with Paragraph-Level Paraphrase Robustness

✇cs.AI, q-bio.NC updates on arXiv.org
By: Jiahao Huo · Wenjie Qu · Yibo Yan · Kening Zheng · Jiaheng Zhang · Xuming Hu · Philip S. Yu · Mingxun Zhou
26 May 2026 at 12:00
arXiv:2605.25796v1 Announce Type: cross Abstract: Semantic-level watermarking (SWM) improves robustness against text modifications by treating sentences as the basic unit. However, robustness to paragraph-level paraphrasing remains difficult because such attacks globally disrupt watermark signals by changing sentence order. In this work, we propose SAMark, a self-anchored watermarking framework that removes the dependency on sentence order by establishing a step-independent green region in semantic space. To improve detectability, we introduce a multi-channel hyperbolic scoring mechanism that amplifies watermark signals while suppressing noise from weakly aligned candidates. We further propose a diversity-aware filtering strategy that combines hard filtering with soft regularization, extending beyond simple n-gram repetition filters to address semantic redundancy. Experimental results show that SAMark achieves up to 90.2% TP@FP1% under typical paragraph-level paraphrasing attacks, outperforming the strongest prior baseline by more than 30% on average, while maintaining generation quality competitive with unwatermarked text and breaking the robustness-quality trade-off that limits prior methods.

Dynamic Dual-Granularity Skill Bank for Agentic RL

✇cs.AI, q-bio.NC updates on arXiv.org
By: Songjun Tu · Chengdong Xu · Qichao Zhang · Yaocheng Zhang · Xiangyuan Lan · Linjing Li · Dong Li · Dongbin Zhao
26 May 2026 at 12:00
arXiv:2603.28716v2 Announce Type: replace Abstract: Agentic RL can benefit substantially from reusable experience, yet existing skill-based methods mainly extract trajectory-level guidance and often lack principled mechanisms for maintaining an evolving skill memory. We propose D2Skill, a dynamic dual-granularity skill bank for agentic RL that organizes reusable experience into task skills for high-level guidance and step skills for fine-grained decision support and error correction. D2Skill jointly trains the policy and skill bank through paired baseline and skill-injected rollouts under the same policy, using their performance gap to derive hindsight utility signals for both skill updating and policy optimization. Built entirely from training-time experience, the skill bank is continuously expanded through reflection and maintained with utility-aware retrieval and pruning. Experiments on ALFWorld, WebShop, and Search-Augmented QA tasks show that D2Skill substantially improves performance over skill-free baselines across models of different scales. Further ablations and analyses show that both dual-granularity skill modeling and dynamic skill maintenance are critical to these gains, while the learned skills exhibit higher utility, transfer across evaluation settings, and introduce only modest training overhead.

AutoResearchClaw: Self-Reinforcing Autonomous Research with Human-AI Collaboration

✇cs.AI, q-bio.NC updates on arXiv.org
By: Jiaqi Liu · Shi Qiu · Mairui Li · Bingzhou Li · Haonian Ji · Siwei Han · Xinyu Ye · Peng Xia · Zihan Dong · Meng Chen · Congyu Zhang · Letian Zhang · Guiming Chen · Haoqin Tu · Xinyu Yang · Lu Feng · Xujiang Zhao · Haifeng Chen · Jiawei Zhou · Xiao Wang · Weitong Zhang · Hongtu Zhu · Yun Li · Jieru Mei · Hongliang Fei · Jiaheng Zhang · Linjie Li · Linjun Zhang · Yuyin Zhou · Sheng Wang · Caiming Xiong · James Zou · Zeyu Zheng · Cihang Xie · Mingyu Ding · Huaxiu Yao
26 May 2026 at 12:00
arXiv:2605.20025v2 Announce Type: replace Abstract: Automating scientific discovery requires more than generating papers from ideas. Real research is iterative: hypotheses are challenged from multiple perspectives, experiments fail and inform the next attempt, and lessons accumulate across cycles. Existing autonomous research systems often model this process as a linear pipeline: they rely on single-agent reasoning, stop when execution fails, and do not carry experience across runs. We present AutoResearchClaw, a multi-agent autonomous research pipeline built on five mechanisms: structured multi-agent debate for hypothesis generation and result analysis, a self-healing executor with a \textsc{Pivot}/\textsc{Refine} decision loop that transforms failures into information, verifiable result reporting that prevents fabricated numbers and hallucinated citations, human-in-the-loop collaboration with seven intervention modes spanning full autonomy to step-by-step oversight, and cross-run evolution that converts past mistakes into future safeguards. On ARC-Bench, a 25-topic experiment-stage benchmark, AutoResearchClaw outperforms AI Scientist v2 by 54.7%. A human-in-the-loop ablation across seven intervention modes reveals that precise, targeted collaboration at high-leverage decision points consistently outperforms both full autonomy and exhaustive step-by-step oversight. We position AutoResearchClaw as a research amplifier that augments rather than replaces human scientific judgment. Code is available at https://github.com/aiming-lab/AutoResearchClaw.

Kolmogorov-Arnold Fourier Networks

✇cs.AI, q-bio.NC updates on arXiv.org
By: Jusheng Zhang · Yijia Fan · Kaitong Cai · Keze Wang · Wenhao Wang
26 May 2026 at 12:00
arXiv:2502.06018v3 Announce Type: replace-cross Abstract: Although Kolmogorov-Arnold-based interpretable networks (KANs) possess strong theoretical expressiveness, they suffer from severe parameter explosion and limited ability to capture high-frequency features in high-dimensional tasks. To address these issues, we propose the Kolmogorov-Arnold Fourier Network (KAF), which fundamentally redefines the KAN paradigm through spectral reparameterization. Our key contributions include: (1) proposing a fundamental basis transformation from the local, grid-based B-spline representation to a global, adaptive spectral representation. This shift changes the network's inductive bias, reducing parameter complexity from $O(G)$ to $O(1)$ while preserving expressiveness; (2) introducing trainable Random Fourier Features (RFF) initialized via a spectral alignment strategy, which allows the model to break the smoothness limitation of fixed kernels and accurately capture high-frequency components; and (3) implementing an adaptive hybrid GELU-Fourier activation mechanism that progressively enhances frequency representation during training. Comprehensive experiments demonstrate the superiority of KAF across computer vision (CV), natural language processing (NLP), audio, and partial differential equation (PDE) solving tasks, achieving state-of-the-art performance with improved efficiency. The code is available at https://github.com/kolmogorovArnoldFourierNetwork/KAF.
Received — 8 April 2026 ⏭

Social Media Intervention Based on the Information-Motivation-Behavioral Skills Model Promotes HIV Testing and Reduces High-Risk Behaviors Among Men Who Have Sex With Men in Resource-Limited Settings in China: Randomized Controlled Trial

✇Journal of Medical Internet Research
By: Huanzhuo Mai · Jianyuan Liu · Liping Cheng · Hao Liang · Ping Cui · Qin Yao · Qin Cao · Heng Zhang · Wenping Liao · Shiran He · Xing Yang · Ping Cen · Jiegang Huang
8 April 2026 at 01:30
Background: Social media intervention may enhance HIV prevention among men who have sex with men, but the effect of this intervention in resource-limited settings remains unclear. Objective: This randomized controlled trial evaluated whether a social media intervention grounded in the information-motivation-behavioral skills (IMB) model could be beneficial for HIV prevention among men who have sex with men in resource-limited settings. Methods: Participants were recruited in Nanning, China, between April 2023 and April 2024. Eligible participants were randomly assigned to either the social media intervention group or the routine HIV prevention services control group. Participants in the intervention group received a 3-month social media intervention, which included completing video-based tasks. Baseline surveys were conducted, followed by follow-up surveys every 3 months, for a total of 2 follow-ups. Outcomes included HIV testing uptake, high-risk behavior, AIDS-related knowledge, safe sex self-efficacy, and attitude. Results: A total of 180 eligible men who have sex with men were enrolled (90 per group). Follow-up rates were 97.8% (88/90) and 95.5% (86/90) for the intervention and control groups, respectively. At the follow-ups, the intervention group demonstrated significantly higher uptake of HIV testing, a lower proportion of participants reporting high-risk sexual behaviors, and higher condom use self-efficacy compared to the control group (all

Combee: Scaling Prompt Learning for Self-Improving Language Model Agents

✇cs.AI, q-bio.NC updates on arXiv.org
By: Hanchen Li · Runyuan He · Qizheng Zhang · Changxiu Ji · Qiuyang Mang · Xiaokun Chen · Lakshya A Agrawal · Wei-Liang Liao · Eric Yang · Alvin Cheung · James Zou · Kunle Olukotun · Ion Stoica · Joseph E. Gonzalez
7 April 2026 at 12:00
arXiv:2604.04247v1 Announce Type: new Abstract: Recent advances in prompt learning allow large language model agents to acquire task-relevant knowledge from inference-time context without parameter changes. For example, existing methods (like ACE or GEPA) can learn system prompts to improve accuracy based on previous agent runs. However, these methods primarily focus on single-agent or low-parallelism settings. This fundamentally limits their ability to efficiently learn from a large set of collected agentic traces. It would be efficient and beneficial to run prompt learning in parallel to accommodate the growing trend of learning from many agentic traces or parallel agent executions. Yet without a principled strategy for scaling, current methods suffer from quality degradation with high parallelism. To improve both the efficiency and quality of prompt learning, we propose Combee, a novel framework to scale parallel prompt learning for self-improving agents. Combee speeds up learning and enables running many agents in parallel while learning from their aggregate traces without quality degradation. To achieve this, Combee leverages parallel scans and employs an augmented shuffle mechanism; Combee also introduces a dynamic batch size controller to balance quality and delay. Evaluations on AppWorld, Terminal-Bench, Formula, and FiNER demonstrate that Combee achieves up to 17x speedup over previous methods with comparable or better accuracy and equivalent cost.

TreeGaussian: Tree-Guided Cascaded Contrastive Learning for Hierarchical Consistent 3D Gaussian Scene Segmentation and Understanding

✇cs.AI, q-bio.NC updates on arXiv.org
By: Jingbin You · Zehao Li · Hao Jiang · Xinzhu Ma · Shuqin Gao · Honglong Zhao · Congcong Zheng · Tianlu Mao · Feng Dai · Yucheng Zhang · Zhaoqi Wang
7 April 2026 at 12:00
arXiv:2604.03309v1 Announce Type: cross Abstract: 3D Gaussian Splatting (3DGS) has emerged as a real-time, differentiable representation for neural scene understanding. However, existing 3DGS-based methods struggle to represent hierarchical 3D semantic structures and capture whole-part relationships in complex scenes. Moreover, dense pairwise comparisons and inconsistent hierarchical labels from 2D priors hinder feature learning, resulting in suboptimal segmentation. To address these limitations, we introduce TreeGaussian, a tree-guided cascaded contrastive learning framework that explicitly models hierarchical semantic relationships and reduces redundancy in contrastive supervision. By constructing a multi-level object tree, TreeGaussian enables structured learning across object-part hierarchies. In addition, we propose a two-stage cascaded contrastive learning strategy that progressively refines feature representations from global to local, mitigating saturation and stabilizing training. A Consistent Segmentation Detection (CSD) mechanism and a graph-based denoising module are further introduced to align segmentation modes across views while suppressing unstable Gaussian points, enhancing segmentation consistency and quality. Extensive experiments, including open-vocabulary 3D object selection, 3D point cloud understanding, and ablation studies, demonstrate the effectiveness and robustness of our approach.

Learning Additively Compositional Latent Actions for Embodied AI

✇cs.AI, q-bio.NC updates on arXiv.org
By: Hangxing Wei · Xiaoyu Chen · Chuheng Zhang · Tim Pearce · Jianyu Chen · Alex Lamb · Li Zhao · Jiang Bian
7 April 2026 at 12:00
arXiv:2604.03340v1 Announce Type: cross Abstract: Latent action learning infers pseudo-action labels from visual transitions, providing an approach to leverage internet-scale video for embodied AI. However, most methods learn latent actions without structural priors that encode the additive, compositional structure of physical motion. As a result, latents often entangle irrelevant scene details or information about future observations with true state changes and miscalibrate motion magnitude. We introduce Additively Compositional Latent Action Model (AC-LAM), which enforces scene-wise additive composition structure over short horizons on the latent action space. These AC constraints encourage simple algebraic structure in the latent action space~(identity, inverse, cycle consistency) and suppress information that does not compose additively. Empirically, AC-LAM learns more structured, motion-specific, and displacement-calibrated latent actions and provides stronger supervision for downstream policy learning, outperforming state-of-the-art LAMs across simulated and real-world tabletop tasks.

Persistent Cross-Attempt State Optimization for Repository-Level Code Generation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Ruwei Pan · Jiangshuai Wang · Qisheng Zhang · Yueheng Zhu · Linhao Wu · Zixiong Yang · Yakun Zhang · Lu Zhang · Hongyu Zhang
7 April 2026 at 12:00
arXiv:2604.03632v1 Announce Type: cross Abstract: Large language models (LLMs) have achieved substantial progress in repository-level code generation. However, solving the same repository-level task often requires multiple attempts, while existing methods still optimize each attempt in isolation and do not preserve or reuse task-specific state across attempts. In this paper, we propose LiveCoder, a novel framework for repository-level code generation based on cross-attempt knowledge optimization. LiveCoder maintains persistent task-specific state from prior attempts to guide subsequent generation. This state includes success knowledge, which captures reusable signals from previously strong repositories, failure knowledge, which records unsuccessful outcomes and their diagnostic signals, and a historical-best repository, which preserves the strongest result found so far and prevents regression. These components collectively transform repeated repository generation into a persistent, knowledge-driven optimization process. We evaluate LiveCoder using four frontier LLMs on two representative repository-level code generation benchmarks. Extensive experimental results demonstrate the effectiveness and efficiency of LiveCoder, improving the functional score by up to 22.94 percentage points, increasing repository reuse to 81.58%, and reducing cost by up to 53.63% on RAL-Bench while maintaining broadly stable non-functional quality.

Agentization of Digital Assets for the Agentic Web: Concepts, Techniques, and Benchmark

✇cs.AI, q-bio.NC updates on arXiv.org
By: Linyao Chen · Bo Huang · Qinlao Zhao · Shuai Shao · Zhi Han · Zicai Cui · Ziheng Zhang · Guangtao Zeng · Wenzheng Tang · Yikun Wang · Yuanjian Zhou · Zimian Peng · Yong Yu · Weiwen Liu · Hiroki Kobayashi · Weinan Zhang
7 April 2026 at 12:00
arXiv:2604.04226v1 Announce Type: cross Abstract: Agentic Web, as a new paradigm that redefines the internet through autonomous, goal-driven interactions, plays an important role in group intelligence. As the foundational semantic primitives of the Agentic Web, digital assets encapsulate interactive web elements into agents, which expand the capacities and coverage of agents in agentic web. The lack of automated methodologies for agent generation limits the wider usage of digital assets and the advancement of the Agentic Web. In this paper, we first formalize these challenges by strictly defining the A2A-Agentization process, decomposing it into critical stages and identifying key technical hurdles on top of the A2A protocol. Based on this framework, we develop an Agentization Agent to agentize digital assets for the Agentic Web. To rigorously evaluate this capability, we propose A2A-Agentization Bench, the first benchmark explicitly designed to evaluate agentization quality in terms of fidelity and interoperability. Our experiments demonstrate that our approach effectively activates the functional capabilities of digital assets and enables interoperable A2A multi-agent collaboration. We believe this work will further facilitate scalable and standardized integration of digital assets into the Agentic Web ecosystem.

KLong: Training LLM Agent for Extremely Long-horizon Tasks

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yue Liu · Yingwei Ma · Yibo Miao · Yanhao Li · Yuchong Xie · Xinlong Yang · Zhiyuan Hu · Flood Sung · Jiaheng Zhang · Bryan Hooi
7 April 2026 at 12:00
arXiv:2602.17547v2 Announce Type: replace Abstract: This paper introduces KLong, an open-source LLM agent trained to solve extremely long-horizon tasks. The principle is to first cold-start the model via trajectory-splitting SFT, then scale it via progressive RL training. Specifically, we first activate basic agentic abilities of a base model with a comprehensive SFT recipe. Then, we introduce Research-Factory, an automated pipeline that generates high-quality training data by collecting research papers and constructing evaluation rubrics. Using this pipeline, we build thousands of long-horizon trajectories distilled from Claude 4.5 Sonnet (Thinking). To train with these extremely long trajectories, we propose a new trajectory-splitting SFT, which preserves early context, progressively truncates later context, and maintains overlap between sub-trajectories. In addition, to further improve long-horizon task-solving capability, we propose a novel progressive RL, which schedules training into multiple stages with progressively extended timeouts. Experiments demonstrate the superiority and generalization of KLong, as shown in Figure 1. Notably, our proposed KLong (106B) surpasses Kimi K2 Thinking (1T) by 11.28% on PaperBench, and the performance improvement generalizes to other coding benchmarks like SWE-bench Verified and MLE-bench.

ToG-Bench: Task-Oriented Spatio-Temporal Grounding in Egocentric Videos

✇cs.AI, q-bio.NC updates on arXiv.org
By: Qi'ao Xu · Tianwen Qian · Yuqian Fu · Kailing Li · Yang Jiao · Jiacheng Zhang · Xiaoling Wang · Liang He
7 April 2026 at 12:00
arXiv:2512.03666v2 Announce Type: replace-cross Abstract: A core capability towards general embodied intelligence lies in localizing task-relevant objects from an egocentric perspective, formulated as Spatio-Temporal Video Grounding (STVG). Despite recent progress, existing STVG studies remain largely confined to object-centric and descriptive instructions, neglecting the task-oriented reasoning that is crucial for embodied agents to accomplish goal-directed interactions. To bridge this gap, we introduce \textbf{ToG-Bench}, the first task-oriented spatio-temporal video grounding benchmark for egocentric videos. ToG-Bench is characterized by three key features: (1) \textbf{Task-oriented Grounding}, which requires identifying and localizing objects based on intended tasks rather than straightforward descriptions; (2) \textbf{Explicit-Implicit Dual Grounding}, where target objects can be either explicitly mentioned or implicitly inferred by contextual reasoning; (3) \textbf{One-to-Many Grounding}, where a single instruction may correspond to multiple objects involved in task execution. Built upon videos sourced from ScanNet, ToG-Bench comprises 100 annotated clips with 2,704 task-oriented grounding instructions, constructed via a semi-automated pipeline that combines foundation model annotation and human refinement. In addition, we introduce a set of task-level evaluation metrics tailored for multi-object and explicit-implicit object grounding, and systematically benchmark seven state-of-the-art MLLMs. Extensive experiments reveal the intrinsic challenges of task-oriented STVG and substantial performance gaps across explicit-implicit and multi-object grounding, highlighting the difficulty of bridging perception and interaction in embodied scenarios. Data and code will be released at: \href{https://github.com/qaxuDev/ToG-Bench}{https://github.com/qaxuDev/ToG-Bench}..
Received — 4 April 2026 ⏭

CogBias: Measuring and Mitigating Cognitive Bias in Large Language Models

✇cs.AI, q-bio.NC updates on arXiv.org
By: Fan Huang · Songheng Zhang · Haewoon Kwak · Jisun An
3 April 2026 at 12:00
arXiv:2604.01366v1 Announce Type: new Abstract: Large Language Models (LLMs) are increasingly deployed in high-stakes decision-making contexts. While prior work has shown that LLMs exhibit cognitive biases behaviorally, whether these biases correspond to identifiable internal representations and can be mitigated through targeted intervention remains an open question. We define LLM cognitive bias as systematic, reproducible deviations from correct answers in tasks with computable ground-truth baselines, and introduce LLM CogBias, a benchmark organized around four families of cognitive biases: Judgment, Information Processing, Social, and Response. We evaluate three LLMs and find that cognitive biases emerge systematically across all four families, with magnitudes and debiasing responses that are strongly family-dependent: prompt-level debiasing substantially reduces Response biases but backfires for Judgment biases. Using linear probes under a contrastive design, we show that these biases are encoded as linearly separable directions in model activation space. Finally, we apply activation steering to modulate biased behavior, achieving 26--32\% reduction in bias score (fraction of biased responses) while preserving downstream capability on 25 benchmarks (Llama: negligible degradation; Qwen: up to $-$19.0pp for Judgment biases). Despite near-orthogonal bias representations across models (mean cosine similarity 0.01), steering reduces bias at similar rates across architectures ($r(246)$=.621, $p$
❌