❌

Normal view

Received — 15 September 2026 ⏭

Occamy-1.0: Open Pareto-frontier 35B Intelligence for Co-work

✇cs.AI, q-bio.NC updates on arXiv.org
By: Wenhui Chen · Shiwen Cheng · Hao Dong · Chenda Duan · Ruixiang Feng · Zhong Guan · Boqiang Guo · Xueyuan Han · Haojie Hao · Liangmeng Huang · Zhelong Huang · Xinke Kong · Hongyu Li · Jiazheng Li · Junbo Li · Qingchuan Li · Yukun Lian · Chang Liu · Tianyu Liu · Zicheng Liu · Shuyi Ouyang · Yijun Pan · Kunyu Shi · Xiaojun Tang · Bingquan Wang · Kesu Wang · Yuchen Wang · Sibo Wei · Sicong Xie · Xiaoying Xing · Yi Xu · Zhijun Xu · Hongwei Xue · Qingcheng Zeng · Di Zhang · Guannan Zhang · Haochen Zhang · Tianlong Zhang · Tianyu Zhao · Tianyu Zhao · Yanjun Zheng · Jialong Zhu · Zijian Zou
14 September 2026 at 12:00
arXiv:2609.11977v1 Announce Type: new Abstract: Co-work agents execute complex workflows that combine information gathering, tool use, coding, and file manipulation across many model invocations. Because cost and latency accumulate over the full episode, their practical value depends not only on peak capability but also on how efficiently that capability is delivered. Yet many steps in everyday work emphasize state tracking, coordination, recovery, and follow-through rather than frontier-scale reasoning. We present Occamy-1.0, a cost-efficient co-work model obtained by further training the post-trained Qwen3.6-35B-A3B checkpoint. We construct execution-grounded data and environments, capture replayable long-horizon trajectories across multiple harnesses, and use staged post-training to develop and consolidate complementary execution capabilities. Across a broad suite of co-work benchmarks, Occamy-1.0 is consistently among the strongest comparably sized models and remains competitive with substantially larger frontier systems on several tasks. Under our stated evaluation and pricing protocol, its aggregate performance across four representative benchmarks places it at the low-cost knee of the observed cost--performance Pareto frontier. Supporting evaluations in tool calling, coding, and instruction following further show that this specialization preserves broad agentic capability. We release the model weights and a subset of the training data to support research on practical co-work agents and agentic post-training.

BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI Agents

✇cs.AI, q-bio.NC updates on arXiv.org
By: Tong Ye · Kunyang Han · Guozhi Wang · Longqiang Luo · Zhifeng Ding · Yongxiang Zhang · Xiaolei Shen · Yuxuan Zhang · Zhuping Zhang · Tao Xu · Yue Pan · Yucheng Zhao · Yupei Hu · Yuanjiang Ouyang · Danfeng Shen · Runqi Lin · Hongda Cai · Zhaoxiong Wang · Mengjia Yan · Yingjie Zhong · Chen Zhou · Zeyu Zhang · Xuwen Zhu · Penggang Shi · Mingcheng Luo · Ziyang Wu · Min Jin · Mingfu Shen · Zairong Xu · Fan Zhang · Hao Wang · Liang Liu · Zhulin Xie · Lijun Yao · Xiao Liang · Liangmin Wen · Liqiang Feng · Feilong Wu · Min Hu · Min Chen · Guanjing Xiong · Xiaohu Ruan · Xiaoxin Chen
14 September 2026 at 12:00
arXiv:2609.12394v1 Announce Type: new Abstract: Mobile GUI agents are shifting from multi-module frameworks to native models trained end-to-end, yet industrial deployment faces three persistent gaps. Sandbox training produces a distribution mismatch with production environments; expensive real-device failures remain underutilized; and fixed benchmarks saturate, losing the power to guide iteration. We present BlueLM-GUI, a 35B-A3B mobile GUI agent built as a real-device-centric flywheel that closes these gaps through three principles. Every Sample Matters: a dual-track pipeline with Heterogeneous Triple-System Consensus evaluation and an Error Correction \& Derivation Module salvages every trajectory into usable supervision. Every Rollout Is Real: a three-stage recipe---continual pre-training, supervised fine-tuning, and agentic reinforcement learning on hundreds of real phones---grounds every rollout in real production environments, so the capability the model learns transfers directly to deployment. Every Query Evolves: a quota-driven benchmark methodology with three orthogonal axes enables precise attribution and allows the benchmark to be systematically upgraded as the model improves. BlueLM-GUI achieves 87.4 on MobileGUI-VBench, surpassing the best closed-source model by 5.1 points, and 84.9 on AndroidWorld, the best result among open-source models and competitive with closed-source models. These results demonstrate that grounding model training and iterative improvement in both real devices and the three Every principles yields strong, robust, and transferable mobile GUI capability.

K-Bench: A Benchmark for LLM Unlearning in Agentic Deployments

✇cs.AI, q-bio.NC updates on arXiv.org
By: Guangsheng Yu · Yanna Jiang · Qin Wang · Baihe Ma · Xu Wang
14 September 2026 at 12:00
arXiv:2609.12808v1 Announce Type: new Abstract: Unlearning benchmarks such as TOFU and MUSE certify forgetting by reading the model's final answer, where a model that refuses to answer already counts as having forgotten. We show that this model-level certificate does not transfer once the model is deployed as an agent. We introduce K-Bench, a benchmark that scores LLM unlearning under agentic deployment. K-Bench inspects all six channels a ReAct agent exposes, including its chain-of-thought (CoT), tool calls and tool observations, and elicited summary. A query counts as leaked if the secret appears in any of them. Each experiment places the secret in exactly one of the agent's three sources (the weights, the prompt, or the retrieval store). The K-Score is computed separately for each source and credits forgetting only when the agent remains usable. Clearing the answer channel does not make the secret unrecoverable. On structured retrieval, the secret stays verbatim in the tool-observation channel and the aggregate leak rate is unchanged. When the secret lives in the prompt or the retrieval store, TOFU and MUSE report no leakage, while the deployed agent still leaks it on 22--86\% of queries. When the secret is in the weights, none of the twenty evaluated published methods demonstrably removes it, and only an input-corruption intervention reaches selective forgetting under the evaluated observer. The top-ranked method changes across base models. A refusal-tuning method resists the evaluated extraction without verified knowledge removal.

MP-Bench: Evaluating Voice Agents as a Multiparty Conversation Participant

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yi-Jen Shih · Shih-Yun Shan Kuan · Guan-Ting Lin · Kai-Wei Chang · Siddhant Arora · Shu-wen Yang · Abdelrahman Mohamed · Shinji Watanabe · Hung-yi Lee · David Harwath
14 September 2026 at 12:00
arXiv:2609.13076v1 Announce Type: cross Abstract: Conversational voice agents have advanced significantly, offering increasingly natural human-machine interactions through both cascaded and end-to-end architectures. However, while recent benchmarks extensively evaluate dyadic interactions and passive audio comprehension, they largely overlook a prevalent real-world scenario: multi-party conversations. Evaluating agents in these settings is fundamentally more challenging than in dyadic interactions due to the exponentially greater conversational complexity. For voice agents to integrate seamlessly into human group dynamics, they must not only generate contextually appropriate responses but also demonstrate a nuanced understanding of open turn-taking. To address this gap, we introduce Multiparty Bench (MP-Bench), the first benchmark specifically designed to objectively evaluate conversational speech systems as active participants within multi-party contexts. MP-Bench assesses agent behavior along two primary dimensions: turn-taking awareness and response appropriateness. Additionally, we incorporate comprehension-based question-answering tasks as a complementary evaluation. By benchmarking 12 voice agents, we find that real-time voice agents stay at or below 22% on multiparty comprehension and remain near chance on multiparty turn-taking, exposing an open challenge for real-time voice agents under multiparty scenario.

HyQuant: Hybrid-Precision Quantization for LLM Attention

✇cs.AI, q-bio.NC updates on arXiv.org
By: Jiatong Ding · Bingxin Xing · Yu Zhang · Dian Ding · Xiaodong Yi · Xianbin Ouyang · Feihu Zhou · Kun Zhang · Zhenyu Guo · Hao Pan · Guangtao Xue · Yiming Zhang
14 September 2026 at 12:00
arXiv:2608.27875v2 Announce Type: replace Abstract: Quantization has been widely adopted in LLM training and inference to reduce cost and improve efficiency. However, low-bit quantization of the \emph{attention} module often introduces large errors at very low bit-widths, causing performance degradation. Existing methods mainly rely on smoothing techniques to handle outliers, while we propose a hybrid quantization design to better balance accuracy and efficiency. Specifically, we propose \textbf{HyQuant}, an efficient hybrid quantization framework for LLM attention. HyQuant quantizes most attention states into low-bit formats while retaining a small set of vertical-line tokens and local-window states in high precision. These accuracy-critical regions are selected using lightweight vertical-line-aware attention-pattern signals, reducing quantization error with limited overhead. In the Prefill stage, HyQuant uses a hybrid-precision quantized attention operator that preserves vertical-line tokens and a local sliding window in full precision while quantizing the remaining context. In the Decode stage, HyQuant applies the same principle to KV-cache compression and fuses KV dequantization with attention computation to improve memory and hardware efficiency. Across diverse tasks, models, and datasets, HyQuant maintains nearly lossless accuracy with an extremely simple design, demonstrating the efficiency and practical feasibility of hybrid quantization for LLM attention. Code is available at: https://github.com/jerrysfls/HyQuant .

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.

Attributable by Construction: Claim-Anchored Provenance for Multi-Document Summarization

✇cs.AI, q-bio.NC updates on arXiv.org
By: Shuo Guan
14 September 2026 at 12:00
arXiv:2606.23989v5 Announce Type: replace-cross Abstract: Large language models produce fluent multi-document summaries, but their attributions are typically coarse---whole documents or passages---and generated post hoc, leaving each statement hard to verify. We argue that attribution should be a structural property of generation rather than a downstream prediction. We present CAMS, a Claim-Anchored Multi-document Summarization framework that decomposes every source document into atomic claims whose provenance is resolved deterministically from verbatim quotes to token spans, clusters equivalent claims across documents while flagging inter-source conflicts, selects a support-aware and salient subset, and rewrites it so that every summary sentence terminates in claim identifiers resolving back to source spans. This yields a separation we make explicit: provenance is an invariant holding for every emitted sentence independently of model accuracy, whereas faithfulness is an objective that selection, constrained rewriting, and verification only encourage---a distinction end-to-end and post-hoc systems conflate. We evaluate on MultiNews, DiverseSumm, and zero-shot on WCEP under a two-regime protocol separating reference-free citation quality from gold-aligned localization, audited by a support model never used for selection or verification. CAMSmatches strong end-to-end and span-attribution baselines on summary quality while improving faithfulness and citation precision, raising multi-source attribution accuracy from 38% to 64% without inflating the number of cited sources, and cutting human verification time per claim by $3.4\times$. We release code and ${\sim}320$K claim--quote--span annotations over MultiNews as a reusable fine-grained attribution resource.

GigaBrain-WBC-0.5: A Behavior World Model for Robust Whole-Body Control with Environment Interaction

✇cs.AI, q-bio.NC updates on arXiv.org
By: Ziyang Cheng · Tianshu Tang · Jinxin Lan · Xinze Chen · Yuhan Gong · Zhichao Liu · Changzhong Wu · Yahao Mao · Zongyan Deng · Mingxuan Ma · Huasen Xi · Yilong Liu · Yutong Wu · Xiaofeng Wang · Yang Wang · Yun Ye · Guan Huang · Xiaojie Jin · Zheng Zhu · Jiwen Lu
14 September 2026 at 12:00
arXiv:2608.18234v3 Announce Type: replace-cross Abstract: Whole-body motion tracking policies turn a humanoid into a robust control interface: the teleoperator---or an upstream model---only supplies a coarse movement intent, while the low-level policy keeps the robot balanced and physically feasible. Existing trackers deliver this interface only on flat ground: trained in empty scenes, they never learn how contact with terrain and objects reshapes their dynamics, and they attempt to teach the policy to balance under any command by continually enlarging the reference-motion corpus, which stops working once feasible behaviors become environment-dependent. We present GigaBrain-WBC-0.5, the first Behavior World Model (BWM) for humanoid whole-body control. Rather than a purely reactive tracker, we train a causal Transformer to jointly predict its next action, next state, and the distribution over its next latent behavior command, so the network that acts also models how the environment shapes what it can do next. An automatic terrain-annotation pipeline recovers full 3D contact geometry from retargeted motion, enabling terrain annotation at the scale of existing motion datasets. The predicted distribution is reused at deployment to detect implausible commands online and retract them onto learned behaviors, so the robot attempts tasks in a "best-effort" manner. The result is a unified policy that takes real-time command, interacts with environment, and stays robust to implausible commands, falls, and disturbances. GigaBrain-WBC-0.5 achieves the highest success rate across all four regimes among three large-scale tracker baselines: 81.3% on terrain interaction (4.3x the strongest baseline), 83.1% under implausible commands, and 99.3% fall recovery (16.8x the strongest baseline). Hardware trials show robust interaction under missing supports and disturbances; the Unitree G1 checkpoint transfers to the Maker L01 robot with simple fine-tuning.

A2DINOv3: Rethinking Multi-Modal Object Detection via Socialized Collaboration

✇cs.AI, q-bio.NC updates on arXiv.org
By: Jiekang Feng · Zhihe Fan · Yunqi Zhu · Xinjie Yao · Yueying Zhang · Yike Gao · Ranxin Li · Guanzuo Chen · Pengfei Zhu
14 September 2026 at 12:00
arXiv:2608.21099v2 Announce Type: replace-cross Abstract: Multi-modal object detection is essential for robust scene understanding in challenging conditions, including low-light and adverse environments. Recent vision foundation models (e.g., DINOv3) have exhibited strong representation capabilities, yet adapting them to multi-modal scenarios remains challenging. Existing dense cross-modal fusion strategies often force heterogeneous modalities to interact indiscriminately, which may introduce redundant information and disrupt the valuable pre-trained representations. To address this issue, we revisit multi-modal fusion from the perspective of socialized learning and propose adapter to DINOv3 (A2DINOv3), a multi-expert collaboration framework with a Socialized Collaboration Protocol (SCP). Specifically, RGB and infrared branches are modeled as heterogeneous experts that independently preserve their specialized knowledge while exchanging complementary information through selective and constrained interactions. This design mitigates harmful cross-modal interference and prevents degradation of pre-trained priors during adaptation. Furthermore, a zero-initialization strategy is introduced to gradually activate cross-modal collaboration, enabling a smooth transition from modality-specific learning to cooperative representation learning. Extensive experiments on four multi-modal benchmarks, including aerial detection (GAIIC), autonomous driving (FLIR), low-light surveillance (LLVIP), and diverse real-world scenarios (M3FD), demonstrate that A2DINOv3 consistently achieves state-of-the-art performance in multi-modal object detection.

Narrative review of the staging classification controversy in stage N3 small cell lung cancer: from the perspective of overlapping Veterans Administration Lung Study Group and International Association for the Study of Lung Cancer definitions

✇Omics In Lung
By: Xinyu Chen · Qilin Wang · Juncai Ye · Xueping Yang · Jiahao Chen · Guangzeng Zhou · Chuanyu You · Tian Xu · Yan Gui
11 September 2026 at 18:00

J Thorac Dis. 2026 Aug 31;18(8):950. doi: 10.21037/jtd-2026-1704. Epub 2026 Aug 28.

ABSTRACT

BACKGROUND AND OBJECTIVE: Traditionally, two primary systems have been employed for staging small cell lung cancer (SCLC): the Veterans Administration Lung Study Group (VALG) system and the International Association for the Study of Lung Cancer (IASLC) tumor, node, metastasis (TNM) system. The term "limited disease" is defined differently: VALG characterizes it as disease encompassed within a single tolerable radiation field, while IASLC defines it as the lack of distant metastases (M0). Patients with N3 disease frequently satisfy VALG extensive-stage (ES) criteria while meeting IASLC limited-stage (LS) criteria, resulting in a notable staging discrepancy. Therefore, this review aims to clarify the clinical challenges posed by this staging overlap and provide insights for standardizing staging terminology and optimizing therapeutic decision-making in N3 SCLC.

METHODS: A narrative review utilizing a systematized search strategy was conducted. While strict adherence to PRISMA guidelines was not pursued because the extensive heterogeneity of the literature precluded a formal meta-analysis, rigorous search criteria were applied to minimize selection bias. Databases including PubMed, Web of Science, Embase, the Cochrane Library, and China National Knowledge Infrastructure (CNKI) were searched for literature from January 2000 to March 2026. Studies examining stage N3 SCLC, spatial metastatic burden, and definitional inconsistencies between the VALG and IASLC staging systems were analyzed to assess their effects on treatment dosimetry, systemic therapy, and survival outcomes.

KEY CONTENT AND FINDINGS: The staging overlap in N3 SCLC leads to heterogeneous clinical management depending on its spatial metastatic burden, and this highly variable cohort can be stratified into distinct prognostic subgroups based on the anatomical distribution (single-region vs. multi-region) of the involved lymph nodes.

CONCLUSIONS: These findings should guide clinical trial design and terminology. Clinical decision-making must transcend historical paradigms and technical constraints. Future strategies must incorporate spatial evaluations of metastatic burden alongside innovative multimodal tools, such as artificial intelligence (AI) and multi-omics, to facilitate tailored therapy for SCLC.

PMID:42724560 | PMC:PMC13559235 | DOI:10.21037/jtd-2026-1704

MMSDH facilitates ACSL4 propionylation to counteract ferroptosis upon hypoxia and impairs PDAC chemotherapy efficacy

✇Nature Cancer
By: Peixiang Zheng · Yanni Lin · Meiyin Zhang · Zhiyuan Tang · Wenxing Qin · Min Xiao · Xue Sun · Yuli Shen · Zhiqiang Hu · Shuo Ru · Yuran Duan · Bofei Dong · Ting Wen · Mengdie Li · Hong Wu · Yueru Hou · Guimei Ji · Zheng Wang · Xiaohong Wu · Qimin Zhou · Xiaoguang Liu · Gaopeng Li · Yuhao Wang · Shengzhong Duan · Yang Luo · Bo Chu · Xu Qian · Zhimin Lu · Yuan Ding · Chuan Xu · Tong Liu · Daqian Xu
11 September 2026 at 08:00

Nature Cancer, Published online: 11 September 2026; doi:10.1038/s43018-026-01236-w

Zheng et al. describe how hypoxia-induced methylmalonate semialdehyde dehydrogenase lactylation promotes acyl-CoA synthetase long-chain family member 4 propionylation and degradation, thereby suppressing ferroptosis induced by chemotherapy, and develop a blocking peptide that increased chemotherapy efficacy in pancreatic ductal adenocarcinoma.
There are no more articles
❌