❌

Normal view

Received — 15 September 2026 ⏭

Tracing and Coordinating Cross-Layer Influence for Multimodal Model Merging

✇cs.AI, q-bio.NC updates on arXiv.org
By: Pengyang Zhou · Xiaobin Tu · Zhengxi Liu · Rongkun Xue · Haochen Li · Miancan Liu · Ziyuan Chen · Yinggui Wang · Jinkui Ren · Xiantao Zhang
14 September 2026 at 12:00
arXiv:2609.12897v1 Announce Type: new Abstract: Multimodal model merging aims to consolidate task experts into a single model that retains their complementary capabilities. Most unimodal model merging methods combine expert updates within individual layers, and multimodal approaches largely follow this design. However, an expert update changes the representations passed to subsequent layers, allowing its influence to propagate across depth and affect how visual and textual information interact. When visual and language updates are combined, later updates act on inputs already modified by earlier ones, coupling their effects. This poses two challenges: (1) how to characterize the multimodal influence of individual expert updates across depth, and (2) how to jointly combine expert updates based on their multimodal influence. To address these challenges, we propose TAC-Merge for tracing and coordinating cross-layer influence in multimodal model merging. It contains two modules, i.e., multimodal influence mapping (MIM) and coupled merge control (CMC). MIM constructs graphs of update effects and uses Ricci curvature together with expert predictions to define a shared fusion objective. CMC models interactions among coefficient adjustments and jointly optimizes regional weights to synthesize one shared model. Experiments across diverse multimodal tasks demonstrate the effectiveness of TAC-Merge in consolidating complementary expert capabilities and supporting generalization to unseen tasks.

Behavior Quotient Learning for Low-Rank Adaptation of LLM Agents

✇cs.AI, q-bio.NC updates on arXiv.org
By: Pengyang Zhou · Xiaobin Tu · Zhengxi Liu · Rongkun Xue · Haochen Li · Miancan Liu · Ziyuan Chen · Yinggui Wang · Jinkui Ren · Xiantao Zhang
14 September 2026 at 12:00
arXiv:2609.12896v1 Announce Type: cross Abstract: LLM-based agents rely on heterogeneous interaction capabilities to accomplish complex tasks. Existing approaches often distribute these capabilities across multiple LoRA adapters, which increases adapter storage requirements and introduces routing overhead during inference. A single LoRA avoids this overhead, but learning from diverse agent trajectories under a fixed rank budget presents two challenges. First, trajectories with different interaction traces and parameter gradients can induce equivalent changes in decision distributions, causing repeated updates to overemphasize redundant behavioral changes. Second, an aggregated update may exceed the rank budget of the adapter, and approximating it in weight space can distort the decision changes that it is intended to produce. We propose BQ-LoRA, a low-rank adaptation framework that organizes trajectory updates through a local behavior quotient manifold. It contains two modules, i.e., behavior quotient balancing (BQB) and decision preserving compression (DPC). BQB constructs the quotient manifold from decision distributions and reweights trajectory update directions according to their local density in the quotient tangent space. DPC projects the balanced gradient onto the intrinsic fixed rank tangent space and refactorizes the resulting target by jointly controlling effective weight error and distortion of decision distributions. Experiments on AppWorld and BrowseComp-Plus compare BQ-LoRA with standard LoRA and recent low-rank adaptation methods, while separate ablations evaluate the complementary contributions of both components.
Received — 10 September 2026 ⏭

CityPlanner: A Sandbox Agent for Executable Urban Planning

✇cs.AI, q-bio.NC updates on arXiv.org
By: Wentao Zhang · Jingyuan Wang · Zetong Zhou · Yifan Yang · Wenrui Wang
10 September 2026 at 12:00
arXiv:2609.09578v1 Announce Type: new Abstract: Urban planning is a real-world spatial optimization problem that requires selecting feasible actions from large candidate spaces under practical objectives such as cost and service quality. Existing optimization and reinforcement learning methods are effective for fixed formulations, but often depend on task-specific representations and constraint handling. We propose \emph{CityPlanner}, a sandbox-agent framework for executable urban planning. CityPlanner introduces \emph{UrbanSandbox}, a unified file-based environment where agents inspect task files, generate plans, run evaluators, and revise decisions based on executable feedback. To make learning tractable, we further propose atomic-task reinforcement learning, which decomposes long sandbox trajectories into \emph{BuildPlan} for initial construction and \emph{ImprovePlan} for feedback-based refinement. Experiments on a real-world benchmark show that CityPlanner consistently outperforms heuristic, task-specific RL, and general LLM-agent baselines. Ablations verify the contributions of UrbanSandbox, atomic-task RL, and iterative deployment. We release the code and dataset at https://anonymous.4open.science/r/co-agent-C1C8

KernelGenBench: Can LLMs and Agents Write Efficient Kernels Across Operator Sources and Hardware Platforms?

✇cs.AI, q-bio.NC updates on arXiv.org
By: Peiyu Zang · Jian Tao · Jialing Zhang · Yichen Yuan · Wentao Zhang · Guang Liu · Yonghua Lin
10 September 2026 at 12:00
arXiv:2607.27231v3 Announce Type: replace Abstract: Modern AI systems depend on specialized accelerator kernels, whose development is complicated by increasingly diverse operators and hardware. LLMs and agentic systems promise to automate this work, but existing evaluations do not show whether their performance transfers across operator sources and hardware platforms, or what such transfer costs. We present KernelGenBench, the first unified multi-source and multi-chip infrastructure for evaluating LLM- and agent-generated Triton kernels. With a common Triton target spanning six hardware platforms, it provides the broadest cross-vendor hardware coverage among existing kernel-generation benchmarks. We report two controlled analytical views: KernelGenBench-MS (Multi-Source) covers 210 operators from PyTorch ATen, production vLLM operators, and proprietary cuBLAS routines, while KernelGenBench-MC (Multi-Chip) evaluates a semantically stable 110-operator subset across six hardware platforms. Our evaluation consumed over 15 billion tokens. Agentic execution improved correctness, but no method dominated across sources and platforms: vLLM posed the strongest correctness challenge, cuBLAS set the highest performance ceiling, and AutoKernel accuracy fell from 87% on NVIDIA to 25% on Iluvatar CoreX. These improvements were costly: specialized agents averaged 4.99 million tokens per successful operator, rising to 6.25 million for CUDA Optimized Skill. The results establish operator source, hardware platform, and agentic scaffold as distinct dimensions of kernel-generation capability, and show that success in a familiar source-hardware setting is not a reliable proxy for deployment readiness.

Author Correction: A mouse brain stereotaxic topographic atlas with isotropic 1-μm resolution

✇Nature - Issue - nature.com science feeds
By: Zhao Feng · Xiangning Li · Yue Luo · Xin Liu · Ben Long · Tao Jiang · Xueyan Jia · Xiaowei Chen · Jie Luo · Xiaokang Chai · Zhen Wang · Miao Ren · Xin Lu · Gang Yao · Mengting Zhao · Yuxin Li · Zhixiang Liu · Hong Ni · Chuhao Dou · Shengda Bao · Shicheng Yang · Zoutao Zhang · Jiandong Zhou · Lingyi Cai · Qi Zhang · Ayizuohere Tudi · Chaozhen Tan · Zhengchao Xu · Siqi Chen · Wenxiang Ding · Wenjuan Shi · Anan Li · Hong-wei Dong · Hui Gong · Qingming Luo
7 September 2026 at 08:00

Nature, Published online: 07 September 2026; doi:10.1038/s41586-026-11094-2

Author Correction: A mouse brain stereotaxic topographic atlas with isotropic 1-μm resolution
Received — 27 May 2026 ⏭

ATIC Promotes LIHC Progression and Serves as an Independent Prognostic Marker: A Pan-cancer Transcriptomic Analysis

✇Omics in Hepatocellular
By: Hui Li · Wentao Zhang · Xiaojie Zhou · Weiming Wu · Fuwen Yao · Haojie Zhong · Naiyang Zhan · JianWei Chen · Yongqiang Zhan · Xinping Yang
19 May 2026 at 18:00

Curr Mol Med. 2026 May 11. doi: 10.2174/0115665240438824260113042223. Online ahead of print.

ABSTRACT

BACKGROUND: 5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase/ IMP cyclohydrolase(ATIC) is a 64-kDa bifunctional enzyme, 5-aminoimidazole- 4-carboxamide ribonucleotide formyltransferase (AICART) and IMP cyclohydrolase, respectively. catalyzes the last two steps of the purine ab initio biosynthetic pathway. ATIC has been implicated in cancer progression, but its pan-cancer profile and specific prognostic utility in liver hepatocellular carcinoma (LIHC) remain incompletely defined.

METHODS: We analyzed TCGA RNA-seq data across 33 tumor types to assess ATIC expression, diagnostic performance (ROC/AUC), and prognostic associations (OS, DSS, PFI). We correlated ATIC expression with immune infiltration, TMB, MSI, and predicted neoantigen load, and constructed a LIHC-specific prognostic nomogram integrating ATIC and clinicopathologic features. Enrichment analyses (STRING, GO/KEGG, GSEA) and pharmacogenomic correlations (GDSC, CTRP) were performed to explore mechanisms and drug sensitivities.

RESULTS: ATIC was significantly upregulated in 16 tumor types, including LIHC (p<0.001). Pan-cancer ROC analyses showed high diagnostic accuracy in several cancers (examples: CHOL AUC=1.000, LIHC AUC=0.936, LUAD AUC=0.947). High ATIC expression associated with poorer OS in ACC, HNSC, LIHC, and PAAD (eg, LIHC: HR=1.39(1.04-1.85), p=0.028). In LIHC, ATIC correlated with advanced T stage, higher grade, elevated AFP, and shorter OS. Multivariable Cox regression identified ATIC expression and pathological T stage as independent predictors; time-dependent ROC for the LIHC nomogram showed AUCs of 0.711, 0.649, and 0.653 at 1, 3, and 5 years, respectively. GSEA indicated enrichment of PI3K-AKT-mTOR, MYC targets, and cell-cycle pathways in ATIC-high LIHC. High ATIC expression correlated with predicted increased sensitivity to sorafenib, doxorubicin, cisplatin, epothilone, and mitomycin in the TCGA-LIHC cohort.

DISCUSSION: ATIC upregulation across cancers links to tumor progression, immune modulation, and prognosis (LIHC), suggesting oncogenic roles in pan-cancer contexts. TCGA multi-omics show ATIC associates with immune/molecular subtypes, MSI/TMB/neoantigens, and predicts drug sensitivity, indicating diagnostic/prognostic potential.

CONCLUSION: ATIC is broadly upregulated across cancers and functions as an independent prognostic biomarker in LIHC. The ATIC-integrated nomogram shows modest predictive accuracy for LIHC survival. Our results implicate ATIC in oncogenic signaling (PI3K-AKT-mTOR, MYC, and cell-cycle) and suggest ATIC as a candidate biomarker to guide targeted and chemotherapeutic strategies in LIHC. Further in vitro and in vivo validation is warranted.

PMID:42152649 | DOI:10.2174/0115665240438824260113042223

Imaging interface-controlled bulk oxygen spillover

✇Nature - Issue - nature.com science feeds
By: Weijue Wang · Hongbin Xu · Shuhui Liu · Xiaofeng Yang · Wei Liu · Yang-Gang Wang · Yanqiang Huang · Tao Zhang
15 April 2026 at 08:00

Nature, Published online: 15 April 2026; doi:10.1038/s41586-026-10324-x

In situ microscopic single-particle imaging demonstrates the significance of rationally engineered metal–support interfaces for activating the oxygen in bulk catalyst, helping elucidate reaction pathways in catalytic conversions.
Received — 8 April 2026 ⏭

Document Parsing Unveiled: Techniques, Challenges, and Prospects for Structured Information Extraction

✇cs.AI, q-bio.NC updates on arXiv.org
By: Qintong Zhang · Bin Wang · Victor Shea-Jay Huang · Junyuan Zhang · Zhengren Wang · Hao Liang · Conghui He · Wentao Zhang
7 April 2026 at 12:00
arXiv:2410.21169v5 Announce Type: replace-cross Abstract: Document parsing (DP) transforms unstructured or semi-structured documents into structured, machine-readable representations, enabling downstream applications such as knowledge base construction and retrieval-augmented generation (RAG). This survey provides a comprehensive and timely review of document parsing research. We propose a systematic taxonomy that organizes existing approaches into modular pipeline-based systems and unified models driven by Vision-Language Models (VLMs). We provide a detailed review of key components in pipeline systems, including layout analysis and the recognition of heterogeneous content such as text, tables, mathematical expressions, and visual elements, and then systematically track the evolution of specialized VLMs for document parsing. Additionally, we summarize widely adopted evaluation metrics and high-quality benchmarks that establish current standards for parsing quality. Finally, we discuss key open challenges, including robustness to complex layouts, reliability of VLM-based parsing, and inference efficiency, and outline directions for building more accurate and scalable document intelligence systems.
Received — 4 April 2026 ⏭

Infeasibility Aware Large Language Models for Combinatorial Optimization

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yakun Wang · Min Chen · Zeguan Wu · Junyu Liu · Sitao Zhang · Zhenwen Shao
3 April 2026 at 12:00
arXiv:2604.01455v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly explored for NP-hard combinatorial optimization problems, but most existing methods emphasize feasible-instance solution generation and do not explicitly address infeasibility detection. We propose an infeasibility-aware framework that combines certifiable dataset construction, supervised fine-tuning, and LLM-assisted downstream search. For the minor-embedding problem, we introduce a new mathematical programming formulation together with provable zero-phase infeasibility screening, which enables scalable construction of training instances labeled either as feasible with structured certificates or as certifiably infeasible. Using training data generated through this exact optimization pipeline, we show that an 8B-parameter LLM can be fine-tuned to jointly perform solution generation and infeasibility detection. We further utilize LLM outputs as warm starts for downstream local search, providing a practical way to accelerate optimization even when the LLM outputs are imperfect. Experiments show that our fine-tuned model improves overall accuracy by up to 30\% over GPT-5.2; meanwhile LLM-guided warm starts provide up to $2\times$ speedup compared with starting from scratch in downstream local search.

ContextBudget: Budget-Aware Context Management for Long-Horizon Search Agents

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yong Wu · YanZhao Zheng · TianZe Xu · ZhenTao Zhang · YuanQiang Yu · JiHuai Zhu · Chao Ma · BinBin Lin · BaoHua Dong · HangCheng Zhu · RuoHui Huang · Gang Yu
3 April 2026 at 12:00
arXiv:2604.01664v1 Announce Type: new Abstract: LLM-based agents show strong potential for long-horizon reasoning, yet their context size is limited by deployment factors (e.g., memory, latency, and cost), yielding a constrained context budget. As interaction histories grow, this induces a trade-off between retaining past information and staying within the context limit. To address this challenge, we propose Budget-Aware Context Management (BACM), which formulates context management as a sequential decision problem with a context budget constraint. It enables agents to assess the available budget before incorporating new observations and decide when and how much of the interaction history to compress. We further develop BACM-RL, an end-to-end curriculum-based reinforcement learning approach that learns compression strategies under varying context budgets. Experiments on compositional multi-objective QA and long-horizon web browsing benchmarks show that BACM-RL consistently outperforms prior methods across model scales and task complexities, achieving over $1.6\times$ gains over strong baselines in high-complexity settings, while maintaining strong advantages as budgets shrink, where most methods exhibit a downward performance trend.
Received — 2 April 2026 ⏭

STAT3-mediated transactivation of NOVA2 promotes lung adenocarcinoma metastasis by splicing SMAD4

✇Oncogene - Issue - nature.com science feeds
By: Shengjie Wang · Xin Tong · Runfeng Sun · Xinyi Xia · Donglai Chen · Zhao Wang · Hao Shi · Chenzhuo Wu · Xiaoxiao Guo · Die Hu · Ersuo Jin · Hong-Tao Zhang
1 April 2026 at 08:00

Oncogene, Published online: 01 April 2026; doi:10.1038/s41388-026-03752-6

STAT3-mediated transactivation of NOVA2 promotes lung adenocarcinoma metastasis by splicing SMAD4

Consensus statement on ctDNA minimal residual disease (MRD) testing in early-stage NSCLC - A Delphi study by the Asian Thoracic Oncology Research Group (ATORG)

✇MRD
By: Aaron C Tan · Bin-Chi Liao · Molly Li · David Lee · Yuji Uehara · Lucksamon Thamlikitkul · Jia-Tao Zhang · Meimei Zheng · Chee Khoon Lee · Nick Pavlakis · Thomas John · Ross A Soo · Anders Skanderup · Pei Jye Voon · Beung-Chul Ahn · Sehhoon Park · Hidetoshi Hayashi · Yasushi Goto · Hidehito Horinouchi · Yasushi Yatabe · Thanyanan Reungwetwattana · James Chih-Hsin Yang · Dong-Wan Kim · Tony Mok · Daniel S W Tan · Yi-Long Wu · Myung-Ju Ahn
28 March 2026 at 18:00

J Thorac Oncol. 2026 Mar 26:103696. doi: 10.1016/j.jtho.2026.103696. Online ahead of print.

ABSTRACT

INTRODUCTION: Minimal residual disease (MRD) detection using liquid biopsy is an emerging tool for risk stratification and monitoring for recurrence in resected early-stage NSCLC. There is increasing need for clear guidance on its optimal clinical implementation.

METHODS: The Asian Thoracic Oncology Research Group (ATORG) convened a multi-disciplinary panel of 27 experts to develop a consensus statement on the clinical application of ctDNA-based MRD testing in early-stage resected NSCLC, using a structured Delphi methodology. Statements were organized into broad thematic domains: Assay validity and standardization; Harmonization in research and trials; Clinical application; Challenges in implementation; Consensus recommendations; Infrastructure for regional MRD adoption; and Roadmap for pragmatic trials.

RESULTS: A total of 23 position statements were developed, of which all except one achieved strong consensus. The consensus highlighted the need to define minimum analytical performance thresholds for MRD assays, improve standardization of reporting metrics, and clear guidelines for pre-analytical handling. Harmonization of blood sampling timepoints and terminology across clinical trials is also essential to confirm the prognostic value of MRD assays. While current MRD assays demonstrate high specificity and positive predictive value, variable sensitivity precludes routine use for adjuvant therapy de-escalation outside clinical trials. Broader access, sustainable funding, ongoing consensus building and collaborative real-world data generation are also critical to support clinical implementation and adoption. Future clinical trials must account for the distinct biology and changing standards of care associated with different driver genes.

CONCLUSION: These consensus recommendations provide a pragmatic framework to guide the responsible integration of MRD testing into clinical research and practice.

PMID:41903701 | DOI:10.1016/j.jtho.2026.103696

Received — 11 March 2026 ⏭

Annealed Co-Generation: Disentangling Variables via Progressive Pairwise Modeling

✇cs.AI, q-bio.NC updates on arXiv.org
By: Hantao Zhang · Jieke Wu · Mingda Xu · Xiao Hu · Yingxuan You · Pascal Fua
10 March 2026 at 12:00
arXiv:2603.06615v1 Announce Type: cross Abstract: For multivariate co-generation in scientific applications, we advocate pairwise block rather than joint modeling of all variables. This design mitigates the computational burden and data imbalance. To this end, we propose an Annealed Co-Generation (ACG) framework that replaces high-dimensional diffusion modeling with a low-dimensional diffusion model, which enables multivariate co-generation by composing pairwise variable generations. We first train an unconditional diffusion model over causal variables that are disentangled into pairs. At inference time, we recover the joint distribution by coupling these pairwise models through shared common variables, enabling coherent multivariate generation without any additional training. By employing a three-stage annealing process-Consensus, Heating, and Cooling-our method enforces consistency across shared common variables and progressively constrains each pairwise data distribution to lie on a learnable manifold, while maintaining high likelihood within each pair. We demonstrate the framework's flexibility and efficacy on two distinct scientific tasks: flow-field completion and antibody generation. All datasets and code will be made publicly available upon publication.

HybridStitch: Pixel and Timestep Level Model Stitching for Diffusion Acceleration

✇cs.AI, q-bio.NC updates on arXiv.org
By: Desen Sun · Jason Hon · Jintao Zhang · Sihang Liu
10 March 2026 at 12:00
arXiv:2603.07815v1 Announce Type: cross Abstract: Diffusion models have demonstrated a remarkable ability in Text-to-Image (T2I) generation applications. Despite the advanced generation output, they suffer from heavy computation overhead, especially for large models that contain tens of billions of parameters. Prior work has illustrated that replacing part of the denoising steps with a smaller model still maintains the generation quality. However, these methods only focus on saving computation for some timesteps, ignoring the difference in compute demand within one timestep. In this work, we propose HybridStitch, a new T2I generation paradigm that treats generation like editing. Specifically, we introduce a hybrid stage that jointly incorporates both the large model and the small model. HybridStitch separates the entire image into two regions: one that is relatively easy to render, enabling an early transition to the smaller model, and another that is more complex and therefore requires refinement by the large model. HybridStitch employs the small model to construct a coarse sketch while exploiting the large model to edit and refine the complex regions. According to our evaluation, HybridStitch achieves 1.83$\times$ speedup on Stable Diffusion 3, which is faster than all existing mixture of model methods.

Taming Modality Entanglement in Continual Audio-Visual Segmentation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yuyang Hong · Qi Yang · Tao Zhang · Zili Wang · Zhaojin Fu · Kun Ding · Bin Fan · Shiming Xiang
10 March 2026 at 12:00
arXiv:2510.17234v2 Announce Type: replace-cross Abstract: Recently, significant progress has been made in multi-modal continual learning, aiming to learn new tasks sequentially in multi-modal settings while preserving performance on previously learned ones. However, existing methods mainly focus on coarse-grained tasks, with limitations in addressing modality entanglement in fine-grained continual learning settings. To bridge this gap, we introduce a novel Continual Audio-Visual Segmentation (CAVS) task, aiming to continuously segment new classes guided by audio. Through comprehensive analysis, two critical challenges are identified: 1) multi-modal semantic drift, where a sounding objects is labeled as background in sequential tasks; 2) co-occurrence confusion, where frequent co-occurring classes tend to be confused. In this work, a Collision-based Multi-modal Rehearsal (CMR) framework is designed to address these challenges. Specifically, for multi-modal semantic drift, a Multi-modal Sample Selection (MSS) strategy is proposed to select samples with high modal consistency for rehearsal. Meanwhile, for co-occurence confusion, a Collision-based Sample Rehearsal (CSR) mechanism is designed, allowing for the increase of rehearsal sample frequency of those confusable classes during training process. Moreover, we construct three audio-visual incremental scenarios to verify effectiveness of our method. Comprehensive experiments demonstrate that our method significantly outperforms single-modal continual learning methods.

Rethinking Driving World Model as Synthetic Data Generator for Perception Tasks

✇cs.AI, q-bio.NC updates on arXiv.org
By: Kai Zeng · Zhanqian Wu · Kaixin Xiong · Xiaobao Wei · Xiangyu Guo · Zhenxin Zhu · Kalok Ho · Lijun Zhou · Bohan Zeng · Ming Lu · Haiyang Sun · Bing Wang · Guang Chen · Hangjun Ye · Wentao Zhang
10 March 2026 at 12:00
arXiv:2510.19195v4 Announce Type: replace-cross Abstract: Recent advancements in driving world models enable controllable generation of high-quality RGB videos or multimodal videos. Existing methods primarily focus on metrics related to generation quality and controllability. However, they often overlook the evaluation of downstream perception tasks, which are $\mathbf{really\ crucial}$ for the performance of autonomous driving. Existing methods usually leverage a training strategy that first pretrains on synthetic data and finetunes on real data, resulting in twice the epochs compared to the baseline (real data only). When we double the epochs in the baseline, the benefit of synthetic data becomes negligible. To thoroughly demonstrate the benefit of synthetic data, we introduce Dream4Drive, a novel synthetic data generation framework designed for enhancing the downstream perception tasks. Dream4Drive first decomposes the input video into several 3D-aware guidance maps and subsequently renders the 3D assets onto these guidance maps. Finally, the driving world model is fine-tuned to produce the edited, multi-view photorealistic videos, which can be used to train the downstream perception models. Dream4Drive enables unprecedented flexibility in generating multi-view corner cases at scale, significantly boosting corner case perception in autonomous driving. To facilitate future research, we also contribute a large-scale 3D asset dataset named DriveObj3D, covering the typical categories in driving scenarios and enabling diverse 3D-aware video editing. We conduct comprehensive experiments to show that Dream4Drive can effectively boost the performance of downstream perception models under various training epochs. Page: https://wm-research.github.io/Dream4Drive/ GitHub Link: https://github.com/wm-research/Dream4Drive
Received — 5 March 2026 ⏭

CoFL: Continuous Flow Fields for Language-Conditioned Navigation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Haokun Liu · Zhaoqi Ma · Yicheng Chen · Masaki Kitagawa · Wentao Zhang · Jinjie Li · Moju Zhao
4 March 2026 at 13:00
arXiv:2603.02854v1 Announce Type: cross Abstract: Language-conditioned navigation pipelines often rely on brittle modular components or costly action-sequence generation. To address these limitations, we present CoFL, an end-to-end policy that directly maps a bird's-eye view (BEV) observation and a language instruction to a continuous flow field for navigation. Instead of predicting discrete action tokens or sampling action chunks via iterative denoising, CoFL outputs instantaneous velocities that can be queried at arbitrary 2D projected locations. Trajectories are obtained by numerical integration of the predicted field, producing smooth motion that remains reactive under closed-loop execution. To enable large-scale training, we build a dataset of over 500k BEV image-instruction pairs, each procedurally annotated with a flow field and a trajectory derived from BEV semantic maps built on Matterport3D and ScanNet. By training on a mixed distribution, CoFL significantly outperforms modular Vision-Language Model (VLM)-based planners and generative policy baselines on strictly unseen scenes. Finally, we deploy CoFL zero-shot in real-world experiments with overhead BEV observations across multiple layouts, maintaining reliable closed-loop control and a high success rate.
Received — 25 February 2026 ⏭

AlphaForgeBench: Benchmarking End-to-End Trading Strategy Design with Large Language Models

✇cs.AI, q-bio.NC updates on arXiv.org
By: Wentao Zhang · Mingxuan Zhao · Jincheng Gao · Jieshun You · Huaiyu Jia · Yilei Zhao · Bo An · Shuo Sun
24 February 2026 at 13:00
arXiv:2602.18481v1 Announce Type: cross Abstract: The rapid advancement of Large Language Models (LLMs) has led to a surge of financial benchmarks, evolving from static knowledge tests to interactive trading simulations. However, current evaluations of real-time trading performance overlook a critical failure mode: severe behavioral instability in sequential decision-making under uncertainty. We empirically show that LLM-based trading agents exhibit extreme run-to-run variance, inconsistent action sequences even under deterministic decoding, and irrational action flipping across adjacent time steps. These issues stem from stateless autoregressive architectures lacking persistent action memory, as well as sensitivity to continuous-to-discrete action mappings in portfolio allocation. As a result, many existing financial trading benchmarks produce unreliable, non-reproducible, and uninformative evaluations. To address these limitations, we propose AlphaForgeBench, a principled framework that reframes LLMs as quantitative researchers rather than execution agents. Instead of emitting trading actions, LLMs generate executable alpha factors and factor-based strategies grounded in financial reasoning. This design decouples reasoning from execution, enabling fully deterministic and reproducible evaluation while aligning with real-world quantitative research workflows. Experiments across multiple state-of-the-art LLMs show that AlphaForgeBench eliminates execution-induced instability and provides a rigorous benchmark for assessing financial reasoning, strategy formulation, and alpha discovery.

Synthesizing Multimodal Geometry Datasets from Scratch and Enabling Visual Alignment via Plotting Code

✇cs.AI, q-bio.NC updates on arXiv.org
By: Haobo Lin · Tianyi Bai · Chen Chen · Jiajun Zhang · Bohan Zeng · Wentao Zhang · Binhang Yuan
24 February 2026 at 13:00
arXiv:2602.18745v1 Announce Type: cross Abstract: Multimodal geometry reasoning requires models to jointly understand visual diagrams and perform structured symbolic inference, yet current vision--language models struggle with complex geometric constructions due to limited training data and weak visual--symbolic alignment. We propose a pipeline for synthesizing complex multimodal geometry problems from scratch and construct a dataset named \textbf{GeoCode}, which decouples problem generation into symbolic seed construction, grounded instantiation with verification, and code-based diagram rendering, ensuring consistency across structure, text, reasoning, and images. Leveraging the plotting code provided in GeoCode, we further introduce code prediction as an explicit alignment objective, transforming visual understanding into a supervised structured prediction task. GeoCode exhibits substantially higher structural complexity and reasoning difficulty than existing benchmarks, while maintaining mathematical correctness through multi-stage validation. Extensive experiments show that models trained on GeoCode achieve consistent improvements on multiple geometry benchmarks, demonstrating both the effectiveness of the dataset and the proposed alignment strategy. The code will be available at https://github.com/would1920/GeoCode.
❌