❌

Normal view

Received — 27 May 2026 ⏭

When Does Multi-Agent RL Improve LLM Workflows? Workflow, Scale, and Policy-Sharing Tradeoffs

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yifan Zeng · Yiran Wu · Yaolun Zhang · Wentian Zhao · Kun Wan · Qingyun Wu · Huazheng Wang
26 May 2026 at 12:00
arXiv:2605.24202v1 Announce Type: new Abstract: Multi-agent LLM workflows route inference through specialized roles to lift end-task accuracy, but jointly training those roles with reinforcement learning is unstable in ways that are poorly understood. We study when end-to-end RL training of multi-agent LLM workflows improves over their base models, comparing Shared-Policy training, where all roles update one policy, with Isolated-Policy training, where each role has its own parameters. Our experimental matrix spans Eval-Opt, Voting, and Orch-Workers workflows, math and code tasks, and three model scales (0.6B, 1.7B, 4B). We find that multi-agent RL usually improves over base models, but gains depend jointly on workflow, task, and scale, not on policy sharing alone. Isolated-Policy tends to reach higher peak accuracy yet more often falls off a terminal accuracy cliff, while Shared-Policy training does not eliminate failure; it redistributes failure into qualitatively different patterns. We then explain the strongest of these patterns through role-level gradient dynamics induced by workflow topology and policy routing: under Isolated-Policy, parallel same-role agents on shared prompts amplify per-role gradients and drive terminal degradation in Voting and Orch-Workers workflows; under Shared-Policy, asymmetric per-step gradient mass causes the shared policy to be captured by the dominant role, producing different failure signatures by task and workflow. Together, the empirical map and its underlying mechanisms show that policy sharing routes training pressure through different channels rather than offering uniform stability, making it a design choice with workflow- and task-conditional tradeoffs.

A pathogen lncRNA secreted into rice sequesters a host miRNA for virulence

✇Nature - Issue - nature.com science feeds
By: Min He · Jia Su · Xiaogang Zhou · Tuo Qi · Jiazheng Wang · Tianxin Zhang · Jinhua Chen · Mawsheng Chern · Youpin Xu · Xiang Lu · Qingqing Hou · Hongrui Liu · Wenfang Huang · Jiawei Liu · Shiying Li · Yunfei Zhu · Xue Chen · Senyu Ran · Han Li · Renju Liu · Mingliang Lei · Guosong Shu · Haicheng Liao · Junjie Yin · Xiaobo Zhu · Yongyan Tang · Li Song · Long Wang · Qing Xiong · Jiali Liu · Yu Bi · Yihua Yang · Hai Qing · Bingtian Ma · Yan Li · Jing Fan · Peng Qin · Yuping Wang · Jianming Wu · Wenming Wang · Shigui Li · Weitao Li · Jing Wang · Xuewei Chen
20 May 2026 at 08:00

Nature, Published online: 20 May 2026; doi:10.1038/s41586-026-10572-x

A fungal long non-coding RNA from Magnaporthe oryzae translocates into rice cells to sequester a host microRNA that normally represses PKR1, a negative immunity regulator, thereby facilitating infection and revealing a widespread RNA-based pathogen–host interaction mechanism.

Integrated radiopathomics nomogram for predicting angiogenic microvascular patterns in NSCLC: a dual-center validation study

✇Omics In Lung
By: Ronghua Wang · Lin Wang · Xinzheng Wang · Shuai Quan · Linning E · Jiangfeng Du
17 April 2026 at 18:00

Ann Med. 2026 Dec;58(1):2654291. doi: 10.1080/07853890.2026.2654291. Epub 2026 Apr 17.

ABSTRACT

BACKGROUND: To develop and validate an integrated radiopathomics nomogram combining multiphase CT images, H&E-stained slides, and clinicopathological variables for predicting microvascular patterns (MVPs) in non-small cell lung cancer (NSCLC).

METHODS: We retrospectively included consecutive surgically resected NSCLC patients from two centers (n = 258). Patients from center 1 were randomly divided into training and internal validation cohorts, while patients from center 2 formed external validation cohort. CD34-immunohistochemistry was used as the reference standard for MVPs to classify patients into non-angiogenic alveolar (NAA) and non-NAA groups. Radiomics and pathomics features were extracted to construct single-phase radiomics, combined radiomics, and pathomics models. Rad-score and Path-score were derived from combined radiomics and pathomics models, respectively. Rad-score, Path-score, and clinicopathological independent predictors were integrated to develop a nomogram. Model performance was assessed by area under the curve (AUC), calibration curve, decision curve analysis (DCA), and DeLong test.

RESULTS: On multivariable analysis, histological grade was an independent predictor of NAA MVP. Combined radiomics model for predicting MVPs achieved AUCs of 0.863, 0.856, and 0.849 in training, internal validation, and external validation cohorts, showing better performance than single-phase models. Pathomics model yielded AUCs of 0.878, 0.860, and 0.833, however, its specificity markedly decreased in validation cohorts. Nomogram model achieved the superior performance across all cohorts, with AUCs of 0.911, 0.903, and 0.901, outperforming single-modality models (DeLong test: all p < 0.05).

CONCLUSION: The nomogram demonstrated high accuracy and robustness in predicting MVPs in NSCLC, offering a promising tool for characterizing the tumor microenvironment and supporting individualized treatment.

PMID:41992828 | DOI:10.1080/07853890.2026.2654291

Received — 11 March 2026 ⏭

Condition-Gated Reasoning for Context-Dependent Biomedical Question Answering

✇cs.AI, q-bio.NC updates on arXiv.org
By: Jash Rajesh Parekh · Wonbin Kweon · Joey Chan · Rezarta Islamaj · Robert Leaman · Pengcheng Jiang · Chih-Hsuan Wei · Zhizheng Wang · Zhiyong Lu · Jiawei Han
10 March 2026 at 12:00
arXiv:2602.17911v2 Announce Type: replace-cross Abstract: Current biomedical question answering (QA) systems often assume that medical knowledge applies uniformly, yet real-world clinical reasoning is inherently conditional: nearly every decision depends on patient-specific factors such as comorbidities and contraindications. Existing benchmarks do not evaluate such conditional reasoning, and retrieval-augmented or graph-based methods lack explicit mechanisms to ensure that retrieved knowledge is applicable to given context. To address this gap, we propose CondMedQA, the first benchmark for conditional biomedical QA, consisting of multi-hop questions whose answers vary with patient conditions. Furthermore, we propose Condition-Gated Reasoning (CGR), a novel framework that constructs condition-aware knowledge graphs and selectively activates or prunes reasoning paths based on query conditions. Our findings show that CGR more reliably selects condition-appropriate answers while matching or exceeding state-of-the-art performance on biomedical QA benchmarks, highlighting the importance of explicitly modeling conditionality for robust medical reasoning.
Received — 5 March 2026 ⏭

Adaptive Social Learning via Mode Policy Optimization for Language Agents

✇cs.AI, q-bio.NC updates on arXiv.org
By: Minzheng Wang · Yongbin Li · Haobo Wang · Xinghua Zhang · Nan Xu · Bingli Wu · Fei Huang · Haiyang Yu · Wenji Mao
4 March 2026 at 13:00
arXiv:2505.02156v5 Announce Type: replace-cross Abstract: Effective social intelligence simulation requires language agents to dynamically adjust reasoning depth, a capability notably absent in current studies. Existing methods either lack explicit reasoning or employ lengthy Chain-of-Thought reasoning uniformly across all scenarios, resulting in excessive token usage and inflexible social behaviors in tasks such as negotiation or collaboration. To address this, we propose an $\textbf{A}$daptive $\textbf{S}$ocial $\textbf{L}$earning ($\textbf{ASL}$) framework in this paper, aiming to improve the adaptive reasoning ability of language agents in dynamic social interactions. To this end, we first identify the hierarchical reasoning modes under such context, ranging from intuitive response to deep deliberation based on the cognitive control theory. We then develop the $\textbf{A}$daptive $\textbf{M}$ode $\textbf{P}$olicy $\textbf{O}$ptimization ($\textbf{AMPO}$) algorithm to learn the context-aware mode adaptation and reasoning. Our framework advances existing research in three key aspects: (1) Multi-granular reasoning mode design, (2) Context-aware mode switching in rich social interaction, and (3) Token-efficient reasoning with depth adaptation. Extensive experiments on the benchmark social intelligence environment verify that ASL achieves 15.6% higher task performance than GPT-4o. Notably, our AMPO outperforms GRPO by 7.0% with 32.8% shorter thinking chains, demonstrating the advantages of our AMPO and the learned adaptive reasoning ability over GRPO's solution.
Received — 17 February 2026 ⏭

Mitigating the Safety-utility Trade-off in LLM Alignment via Adaptive Safe Context Learning

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yanbo Wang · Minzheng Wang · Jian Liang · Lu Wang · Yongcan Yu · Ran He
17 February 2026 at 13:00
arXiv:2602.13562v1 Announce Type: cross Abstract: While reasoning models have achieved remarkable success in complex reasoning tasks, their increasing power necessitates stringent safety measures. For safety alignment, the core challenge lies in the inherent trade-off between safety and utility. However, prevailing alignment strategies typically construct CoT training data with explicit safety rules via context distillation. This approach inadvertently limits reasoning capabilities by creating a rigid association between rule memorization and refusal. To mitigate the safety-utility trade-off, we propose the Adaptive Safe Context Learning (ASCL) framework to improve the reasoning given proper context. ASCL formulates safety alignment as a multi-turn tool-use process, empowering the model to autonomously decide when to consult safety rules and how to generate the ongoing reasoning. Furthermore, to counteract the preference for rule consultation during RL, we introduce Inverse Frequency Policy Optimization (IFPO) to rebalance advantage estimates. By decoupling rule retrieval and subsequent reasoning, our method achieves higher overall performance compared to baselines.

CT-Bench: A Benchmark for Multimodal Lesion Understanding in Computed Tomography

✇cs.AI, q-bio.NC updates on arXiv.org
By: Qingqing Zhu · Qiao Jin · Tejas S. Mathai · Yin Fang · Zhizheng Wang · Yifan Yang · Maame Sarfo-Gyamfi · Benjamin Hou · Ran Gu · Praveen T. S. Balamuralikrishna · Kenneth C. Wang · Ronald M. Summers · Zhiyong Lu
17 February 2026 at 13:00
arXiv:2602.14879v1 Announce Type: cross Abstract: Artificial intelligence (AI) can automatically delineate lesions on computed tomography (CT) and generate radiology report content, yet progress is limited by the scarcity of publicly available CT datasets with lesion-level annotations. To bridge this gap, we introduce CT-Bench, a first-of-its-kind benchmark dataset comprising two components: a Lesion Image and Metadata Set containing 20,335 lesions from 7,795 CT studies with bounding boxes, descriptions, and size information, and a multitask visual question answering benchmark with 2,850 QA pairs covering lesion localization, description, size estimation, and attribute categorization. Hard negative examples are included to reflect real-world diagnostic challenges. We evaluate multiple state-of-the-art multimodal models, including vision-language and medical CLIP variants, by comparing their performance to radiologist assessments, demonstrating the value of CT-Bench as a comprehensive benchmark for lesion analysis. Moreover, fine-tuning models on the Lesion Image and Metadata Set yields significant performance gains across both components, underscoring the clinical utility of CT-Bench.

Less is More: Improving LLM Alignment via Preference Data Selection

✇cs.AI, q-bio.NC updates on arXiv.org
By: Xun Deng · Han Zhong · Rui Ai · Fuli Feng · Zheng Wang · Xiangnan He
17 February 2026 at 13:00
arXiv:2502.14560v4 Announce Type: replace-cross Abstract: Direct Preference Optimization (DPO) has emerged as a promising approach for aligning large language models with human preferences. While prior work mainly extends DPO from the aspect of the objective function, we instead improve DPO from the largely overlooked but critical aspect of data selection. Specifically, we address the issue of parameter shrinkage caused by noisy data by proposing a novel margin-maximization principle for dataset curation in DPO training. To further mitigate the noise in different reward models, we propose a Bayesian Aggregation approach that unifies multiple margin sources (external and implicit) into a single preference probability. Extensive experiments in diverse settings demonstrate the consistently high data efficiency of our approach. Remarkably, by using just 10\% of the Ultrafeedback dataset, our approach achieves 3\% to 8\% improvements across various Llama, Mistral, and Qwen models on the AlpacaEval2 benchmark. Furthermore, our approach seamlessly extends to iterative DPO, yielding a roughly 3\% improvement with 25\% online data, revealing the high redundancy in this presumed high-quality data construction manner. These results highlight the potential of data selection strategies for advancing preference optimization.
There are no more articles
❌