❌

Reading view

↗️

Engineering inflammation-responsive proteins through nitric oxide-caged amino acids

✇Nature Biomedical Engineering
By: Wenkang Cai · Junhao Cui · Zhiying Zeng · Zexian Xiang · Yuanzhe Xie · Yeyu Su · Yi Zuo · Yingze Liu · Haoyu Wang · Liying Chang · Xue Wang · Jingjing Wang · Jun-An Ma · Tao Liu

Nature Biomedical Engineering, Published online: 31 August 2026; doi:10.1038/s41551-026-01782-9

A protein engineering strategy enables nitric oxide-triggered reactivation of proteins using genetically encoded caged amino acids, allowing inflammation-localized control of protein activity, viral gene delivery and biosensing in vivo.
  •  
  • ↗️
↗️

FlowCPO: A Unified Divergence View of Preference Alignment for Flow Models

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yansen Han · Shengyi Liao · Peng Sun · Deyuan Liu · Yuanxing Zhang · Pengfei Wan · Tao Lin
arXiv:2609.09905v1 Announce Type: cross Abstract: Preference alignment for flow and diffusion models now spans online reinforcement learning and offline preference optimization, but the relation between these methods remains unclear. In particular, existing forward-process alignment methods require fresh samples from the current model, while offline methods based on fixed preference pairs rely primarily on positive-only fine-tuning or DPO-style likelihood-ratio surrogates. We organize these approaches through a divergence-based framework and introduce FlowCPO, an offline forward-KL objective that uses both preferred and dispreferred samples without online rollouts. For linear interpolation, we show under explicit regularity conditions that the forward-KL objective is bounded by a contrastive flow matching loss, yielding a tractable surrogate on fixed data. We further show that this loss is nonnegative, whereas the signed regression loss of simplified FlowDPO can be unbounded below. In the in-domain setting, FlowCPO achieves higher mean GenEval and OCR scores than the evaluated baselines, reaching 0.84 and 0.87 versus 0.81 and 0.74 for FlowDPO at CFG 3.0. In the out-of-domain setting, the results are mixed, with the best GenEval result but lower reward scores than RFT on several metrics.
  •  
  • ↗️
↗️

Ancient proteins identify various Denisovan remains from Southwest China

✇Nature - Issue - nature.com science feeds
By: Huiyun Rao · Song Xing · Qijun Ruan · Fan Bai · Yuhao Zhao · Zhongyou Wu · Siyuan Tang · E. Andrew Bennett · Xiaotian Feng · Sungui Lin · Qingfeng Shao · Xiong Yang · Zitao Li · Kai Zhou · Feng Liu · Qiaomei Fu

Nature, Published online: 09 September 2026; doi:10.1038/s41586-026-10976-9

Identification and proteomic analysis of bone fragments and teeth from an excavation in Southwest China provide insight into the evolution and phenotype of Denisovans and fill a geographical gap in their documented distribution.
  •  
  • ↗️
↗️

Integrated single-cell multi-omics characterization reveals lipid-associated macrophage-mediated immunosuppression in neoadjuvant immunotherapy of hepatocellular carcinoma

✇Omics in Hepatocellular
By: Shengxuan Peng · Chang Liu · Junyu Long · Jincheng Tian · Han Li · Donghai Lu · Qihang Cao · Daolin Zhang · Qiao He · Jisen Jia · Yuxuan Wang · Zhaoru Dong · Haitao Zhao · Lei Zhao · Dongxu Wang · Tao Li

Nat Commun. 2026 Jul 31;17(1):9381. doi: 10.1038/s41467-026-75949-y.

ABSTRACT

Hepatocellular carcinoma (HCC) is a cancer with high incidence and mortality rate. Although immune checkpoint inhibitors (ICIs) improved survival outcomes for HCC patients, limited objective response rate highlights the urgency of investigating determinants of immunotherapy. Here, we explore HCC resistance mechanisms following neoadjuvant αPD-1 immunotherapy by constructing a comprehensive multi-modal single-cell transcriptomic atlas consisting of 14 HCC patients treated with αPD-1 from our cohort (ClinicalTrials.gov ID: NCT06571396) and 60 external HCC cases with heterogeneous treatment backgrounds. Supervised by clinical outcomes of our cohort, we identify positive and negative regulators of immunotherapy within the tumor immune microenvironment (TIME), especially lipid-associated macrophages (LAM) with increased lipid metabolic state in non-responders and characterized by C1QA, FABP1, and APOA1 expression. We further show the presence, exogenous inducements and immunosuppressive functions of LAM, along with regulation strategies of its lipid-associated condition, including lycopene and chiglitazar. Furthermore, we construct interaction networks of immune regulators across responders and non-responders, showing distinct ligand-receptor landscapes with intervention targets. We reveal the TIME components including immunosuppressive LAMs that influence immunotherapy outcomes, thus providing evidence and insights for exploring immune landscape and therapeutic strategies for HCC immunotherapy. ClinicalTrials.gov ID: NCT06571396.

PMID:42680737 | PMC:PMC13534469 | DOI:10.1038/s41467-026-75949-y

  •  
  • ↗️
↗️

When the Manual Lies: A Realistic Benchmark to Evaluate MCP Poisoning Attacks for LLM Agents

✇cs.AI, q-bio.NC updates on arXiv.org
By: Shi Liu · Xuehai Tang · Xikang Yang · Liang Lin · Biyu Zhou · Wenjie Xiao · Wantao Liu
arXiv:2605.24069v1 Announce Type: cross Abstract: The rise of tool-using Large Language Model (LLM) agents, standardized by protocols like the Model Context Protocol (MCP), has unlocked unprecedented autonomous execution capabilities for LLM Agents by integrating external open-domain knowledge and tools. However, this interoperability introduces a covert attack surface targeting the agent's cognitive planning layer. This paper systematically investigates Tool Description Poisoning (TDP), a novel semantic attack. In TDP, malicious instructions are not embedded in a tool's executable code, but rather covertly injected into its descriptive metadata, the very "manual" an agent relies on for secure planning and decision-making. To rigorously and systematically evaluate this emerging threat, we introduce the MCP-TDP Security Benchmark. This high-fidelity sandbox environment comprises 32 realistic, real-world test cases spanning 6 distinct risk categories. Our evaluation of 8 mainstream LLMs reveals severe vulnerabilities, with leading models like GPT-4o exhibiting a nearly 100% Attack Success Rate (ASR) in six high-risk scenarios. Furthermore, our findings demonstrate that common prompt-guardrail defenses are largely ineffective and can, counterintuitively, even be counterproductive (a phenomenon which we term the "Firewall Fallacy"). Crucially, we also propose a defense mechanism: "Reactive Self-Correction," where an agent autonomously detects and reverts its own malicious actions post-execution. This work provides the first specialized security benchmark tailored for TDP, offering essential insights for securing the cognitive and planning layers of advanced agentic systems.
  •  
  • ↗️
↗️

Security in the Fine-Tuning Lifecycle of Large Language Models: Threats, Defenses,Evaluation, and Future Directions

✇cs.AI, q-bio.NC updates on arXiv.org
By: Wenjuan Li · Yitao Liu · Runze Chen · Rajkumar Buyya
arXiv:2605.25073v1 Announce Type: cross Abstract: Background: Fine-tuning is central to adapting pre-trained Large Language Models (LLMs) to downstream tasks, but its reliance on training data, parameter updates, and reusable components opens entry points for attackers. Threats have evolved from data poisoning and weight tampering to agent manipulation and interface exploitation, yet existing reviews lack a unified framework spanning the full fine-tuning lifecycle. Objective: This paper presents a systematic survey of LLM fine-tuning security and establishes a lifecycle-based framework for comparing attacks and defenses, complemented by unified empirical evaluation. Methods: We divide attack and defense mechanisms into three phases by intervention timing: pre-tuning, during-tuning, and post-tuning. Within each phase, strategies are reviewed and contrasted to expose their evolution and limitations. Representative methods are then evaluated under a unified model, hardware, and protocol setup, with cross-phase experiments pairing attacks and defenses from different phases. Results: Attack effectiveness is highly model-dependent and non-monotonic with scale: weight-editing attacks effective on earlier models lose impact on modern open-source LLMs; cross-lingual backdoor transfer, reported as near-perfect at larger scales, fails entirely on tested 1B-4B models; and purely benign samples can compromise safety alignment in instruction-tuned models. Single-phase defenses rarely generalize across phases, and defense effectiveness depends jointly on model architecture and alignment state. Conclusion: We identify key open problems (configuration-robust defense, cross-phase defense composition, and embedding-space attacks beyond behavioral assumptions) and propose concrete future research directions.
  •  
  • ↗️
↗️

Hide to Guide: Learning via Semantic Masking

✇cs.AI, q-bio.NC updates on arXiv.org
By: Ruitao Liu · Qinghao Hu · Alex Hu · Yecheng Wu · Shang Yang · Luke J. Huang · Zhuoyang Zhang · Han Cai · Song Han
arXiv:2605.25198v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has become a powerful paradigm for improving language models on reasoning-intensive tasks, but its effectiveness is often limited by exploration. For example, models often fail on hard problems, leaving little useful reward signal. External expert traces offer a natural source of guidance, yet they may also expose reward-relevant content along the critical path to the verifier target, such as final answers, intermediate values, executable implementations, or answer-related entities. This content can create an unintended reward hacking channel, allowing the policy to obtain reward by copying the trace rather than learning the underlying reasoning or agentic behavior. Existing guided-RL methods reduce this risk by using partial trajectories, but they mainly control how much expert information is shown heuristically rather than which parts should be hidden. To this end, we propose Semantic Masked Expert Policy Optimization (SMEPO), a fine-grained semantic masking strategy for expert-guided RLVR. Instead of truncating traces coarsely or revealing them unchanged, SMEPO masks reward-relevant semantic spans along the critical path while preserving the expert's decomposition, plan, and procedural structure. This turns hard problems from reasoning from scratch into a fill-in-the-blank process: the policy can follow the expert's problem-solving route, but must still reconstruct the missing values, code, or entities by itself. SMEPO is simple to apply and requires no changes to the reward function or RL objective. Across diverse domains, including math, code, and agentic search, SMEPO improves accuracy by up to 3.2 points over GRPO and reduces training time by up to 4.2x. The code is available at https://github.com/mit-han-lab/SMEPO.
  •  
  • ↗️
↗️

NSR-Boost: A Neuro-Symbolic Residual Boosting Framework for Industrial Legacy Models

✇cs.AI, q-bio.NC updates on arXiv.org
By: Ziming Dai · Dabiao Ma · Jinle Tong · Mengyuan Han · Jian Yang · Hongtao Liu · Haojun Fei · Qing Yang
arXiv:2601.10457v3 Announce Type: replace Abstract: Although the Gradient Boosted Decision Trees (GBDTs) dominate industrial tabular applications, upgrading legacy models in high-concurrency production environments still faces prohibitive retraining costs and systemic risks. To address this problem, we present NSR-Boost, a neuro-symbolic residual boosting framework designed specifically for industrial scenarios. Its core advantage lies in being ``non-intrusive''. It treats the legacy model as a frozen model and performs targeted repairs on "hard regions" where predictions fail. The framework comprises three key stages: First, finding hard regions through residuals, then generating interpretable experts by generating symbolic code structures using Large Language Model (LLM) and fine-tuning parameters using Bayesian optimization, and finally dynamically integrating experts with legacy model output through a lightweight aggregator. Experimental results demonstrate that the framework significantly outperforms state-of-the-art (SOTA) baselines across six public datasets and one private dataset. More importantly, we report the successful deployment of NSR-Boost within the core financial risk control system of Qfin Holdings, where empirical results on real-world online traffic exhibit superior performance improvements and a significant reduction in the bad rate. In conclusion, it effectively captures long-tail risks missed by traditional models and offers a safe, low-cost evolutionary paradigm for industry.
  •  
  • ↗️
↗️

Bridging the Semantic-Action Gap in Visual Token Pruning for Efficient VLA Inference

✇cs.AI, q-bio.NC updates on arXiv.org
By: Ziyan Liu · Yeqiu Chen · Hongyi Cai · Tao Lin · Shuo Yang · Zheng Liu · Bo Zhao
arXiv:2511.16449v5 Announce Type: replace-cross Abstract: Vision-Language-Action (VLA) models have shown great potential for embodied AI by integrating visual perception, language understanding, and action execution. In real-time deployment, these models must process continuous visual streams, incurring substantial computational overhead. Visual token pruning -- a mainstream technique for accelerating Vision-Language Models (VLMs) by retaining salient tokens while discarding redundant ones -- offers a natural candidate solution to this challenge. However, directly applying VLM-oriented pruning methods to VLA inference can cause severe degradation in manipulation performance. Our analysis attributes this degradation to a key mismatch: VLA inference exhibits distinct attention patterns between the vision-language prefill stage and the action-decode stage, so pruning based only on context-prefill semantic salience is biased toward semantic cues and may remove action-critical visual tokens. Motivated by this observation, we propose VLA-Pruner, an effective plug-and-play token pruning method grounded in the visual requirements of VLA inference, further exploiting the temporal continuity of robot manipulation. Specifically, VLA-Pruner estimates visual-token importance from both semantic prefilling and temporally smoothed action relevance, and then applies a Combine-then-Filter strategy to retain compact, non-redundant tokens under the compute budget. Experiments show that VLA-Pruner outperforms state-of-the-art approaches across multiple VLA architectures, achieving up to 1.99x speedup with comparable manipulation quality.
  •  
  • ↗️
↗️

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

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.
  •  
  • ↗️
↗️

Effect of the Maxing Huoqiao granule on nonsevere community-acquired pneumonia: A multicenter, double-blind, placebo-controlled randomized trial

✇(Multiomics OR Omics) AND (Lung OR gastric OR Hepatocellular)
By: Qingming Ke · Caiping Zhao · Shengda Luo · Yuntao Liu · Ye Ye · Xingdong Lin · Jing Zeng · Xinghua Liu · Yang Liao · Hongtao Li · Shu Chen · Hong Ren · Liang Liu · Hudan Pan

Pharmacol Res. 2026 Apr 9:108186. doi: 10.1016/j.phrs.2026.108186. Online ahead of print.

ABSTRACT

Community-acquired pneumonia (CAP) remains a major global public health challenge with substantial morbidity and mortality. Although preclinical studies suggest that Maxing Huoqiao (MXHQ) granule may have therapeutic potential for pneumonia, high-quality clinical evidence is still limited. We conducted a multicenter, double-blind, randomized, placebo-controlled trial at two tertiary hospitals in China to evaluate the clinical efficacy of MXHQ as adjunctive therapy and to explore its potential mechanisms in adults with nonsevere CAP receiving standard moxifloxacin treatment. A total of 96 patients were enrolled and randomized (1:1:1) to receive standard-dose MXHQ, low-dose MXHQ, or placebo in addition to moxifloxacin for 7 days, with a 14-day follow-up. The primary endpoint was clinical cure, defined as composite recovery of major respiratory symptoms, lung rales, and fever; secondary endpoints included symptom relief, radiographic improvement, and safety. Compared with placebo, standard-dose MXHQ was associated with a higher day-14 clinical cure rate (30.78% vs. 68.97%; RR = 0.45, 95% CI = 0.24-0.83; P < 0.01). Furthermore, the standard-dose intervention was correlated with a shorter time to relief and recovery of cough and sputum (P < 0.05), as well as improvements in symptom scores (P < 0.05) and promoting lesion absorption on chest CT (P < 0.05). Low-dose MXHQ showed no significant clinical benefit, whereas safety profiles were comparable across all groups. Transcriptomic analyses of peripheral blood mononuclear cells, complemented by a Streptococcus pneumonia animal model, indicated that the clinical benefits of MXHQ are linked to the modulation of inflammation and innate immunity. These omics and in vivo observations suggest a potential mechanism underlying the protective effects of MXHQ against inflammatory injury and promotion of tissue repair, involving the regulation of anti-inflammatory mediators and tissue repair-related factors. (Chictr.org.cn, ID Number: ChiCTR2400082095).

PMID:41966499 | DOI:10.1016/j.phrs.2026.108186

  •  
  • ↗️
↗️

Talk to Right Specialists: Iterative Routing in Multi-agent Systems for Question Answering

✇cs.AI, q-bio.NC updates on arXiv.org
By: Feijie Wu · Zitao Li · Fei Wei · Yaliang Li · Bolin Ding · Jing Gao
arXiv:2501.07813v2 Announce Type: replace-cross Abstract: Retrieval-augmented generation (RAG) agents are increasingly deployed to answer questions over local knowledge bases that cannot be centralized due to knowledge-sovereignty constraints. This results in two recurring failures in production: users do not know which agent to consult, and complex questions require evidence distributed across multiple agents. To overcome these challenges, we propose RIRS, a training-free orchestration framework to enable a multi-agent system for question answering. In detail, RIRS summarizes each agent's local corpus in an embedding space, enabling a user-facing server to route queries only to the most relevant agents, reducing latency and avoiding noisy "broadcast-to-all" contexts. For complicated questions, the server can iteratively aggregate responses to derive intermediate results and refine the question to bridge the gap toward a comprehensive answer. Extensive experiments demonstrate the effectiveness of RIRS, including its ability to precisely select agents and provide accurate responses to single-hop queries, and its use of an iterative strategy to achieve accurate, multi-step resolutions for complex queries.
  •  
  • ↗️
↗️

From Multi-Agent to Single-Agent: When Is Skill Distillation Beneficial?

✇cs.AI, q-bio.NC updates on arXiv.org
By: Binyan Xu · Dong Fang · Haitao Li · Kehuan Zhang
arXiv:2604.01608v1 Announce Type: new Abstract: Multi-agent systems (MAS) tackle complex tasks by distributing expertise, though this often comes at the cost of heavy coordination overhead, context fragmentation, and brittle phase ordering. Distilling a MAS into a single-agent skill can bypass these costs, but this conversion lacks a principled answer for when and what to distill. Instead, the empirical outcome is surprisingly inconsistent: skill lift ranges from a 28% improvement to a 2% degradation across metrics of the exact same task. In this work, we reveal that skill utility is governed not by the task, but by the evaluation metric. We introduce Metric Freedom ($F$), the first a priori predictor of skill utility. $F$ measures the topological rigidity of a metric's scoring landscape by quantifying how output diversity couples with score variance via a Mantel test. Guided by $F$, we propose a two-stage adaptive distillation framework. Stage 1 acts as a selective extraction mechanism, extracting tools and knowledge while discarding restrictive structures on "free" metrics to preserve exploration. Stage 2 targets computationally intensive iterative refinement exclusively toward "rigid" metrics ($F \lesssim 0.6$) to eliminate trajectory-local overfitting. Evaluating across 4 tasks, 11 datasets, and 6 metrics, $F$ strongly predicts skill utility ($\rho = -0.62$, $p
  •  
  • ↗️
↗️

SleepVLM: Explainable and Rule-Grounded Sleep Staging via a Vision-Language Model

✇cs.AI, q-bio.NC updates on arXiv.org
By: Guifeng Deng · Pan Wang · Jiquan Wang · Shuying Rao · Junyi Xie · Wanjun Guo · Tao Li · Haiteng Jiang
arXiv:2603.26738v2 Announce Type: replace-cross Abstract: While automated sleep staging has achieved expert-level accuracy, its clinical adoption is hindered by a lack of auditable reasoning. We introduce SleepVLM, a rule-grounded vision-language model (VLM) designed to stage sleep from multi-channel polysomnography (PSG) waveform images while generating clinician-readable rationales based on American Academy of Sleep Medicine (AASM) scoring criteria. Utilizing waveform-perceptual pre-training and rule-grounded supervised fine-tuning, SleepVLM achieved Cohen's kappa scores of 0.767 on an held out test set (MASS-SS1) and 0.743 on an external cohort (ZUAMHCS), matching state-of-the-art performance. Expert evaluations further validated the quality of the model's reasoning, with mean scores exceeding 4.0/5.0 for factual accuracy, evidence comprehensiveness, and logical coherence. By coupling competitive performance with transparent, rule-based explanations, SleepVLM may improve the trustworthiness and auditability of automated sleep staging in clinical workflows. To facilitate further research in interpretable sleep medicine, we release MASS-EX, a novel expert-annotated dataset.
  •  
  • ↗️
↗️

Ran Score: a LLM-based Evaluation Score for Radiology Report Generation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Ran Zhang · Yucong Lin · Zhaoli Su · Bowen Liu · Danni Ai · Tianyu Fu · Deqiang Xiao · Jingfan Fan · Yuanyuan Wang · Mingwei Gao · Yuwan Hu · Shuya Gao · Jingtao Li · Jian Yang · Hong Song · Hongliang Sun
arXiv:2603.22935v1 Announce Type: new Abstract: Chest X-ray report generation and automated evaluation are limited by poor recognition of low-prevalence abnormalities and inadequate handling of clinically important language, including negation and ambiguity. We develop a clinician-guided framework combining human expertise and large language models for multi-label finding extraction from free-text chest X-ray reports and use it to define Ran Score, a finding-level metric for report evaluation. Using three non-overlapping MIMIC-CXR-EN cohorts from a public chest X-ray dataset and an independent ChestX-CN validation cohort, we optimize prompts, establish radiologist-derived reference labels and evaluate report generation models. The optimized framework improves the macro-averaged score from 0.753 to 0.956 on the MIMIC-CXR-EN development cohort, exceeds the CheXbert benchmark by 15.7 percentage points on directly comparable labels, and shows robust generalization on the ChestX-CN validation cohort. Here we show that clinician-guided prompt optimization improves agreement with a radiologist-derived reference standard and that Ran Score enables finding-level evaluation of report fidelity, particularly for low-prevalence abnormalities.
  •  
  • ↗️
↗️

PaLMR: Towards Faithful Visual Reasoning via Multimodal Process Alignment

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yantao Li · Qiang Hui · Chenyang Yan · Kanzhi Cheng · Fang Zhao · Chao Tan · Huanling Gao · Jianbing Zhang · Kai Wang · Xinyu Dai · Shiguo Lian
arXiv:2603.06652v1 Announce Type: cross Abstract: Reinforcement learning has recently improved the reasoning ability of Large Language Models and Multimodal LLMs, yet prevailing reward designs emphasise final-answer correctness and consequently tolerate process hallucinations--cases where models reach the right answer while misperceiving visual evidence. We address this process-level misalignment with PaLMR, a framework that aligns not only outcomes but also the reasoning process itself. PaLMR comprises two complementary components: a perception-aligned data layer that constructs process-aware reasoning data with structured pseudo-ground-truths and verifiable visual facts, and a process-aligned optimisation layer that constructs a hierarchical reward fusion scheme with a process-aware scoring function to encourage visually faithful chains-of-thought and improve training stability. Experiments on Qwen2.5-VL-7B show that our approach substantially reduces reasoning hallucinations and improves visual reasoning fidelity, achieving state-of-the-art results on HallusionBench while maintaining strong performance on MMMU, MathVista, and MathVerse. These findings indicate that PaLMR offers a principled and practical route to process-aligned multimodal reasoning, advancing the reliability and interpretability of MLLMs.
  •  
  • ↗️
↗️

Not All Candidates are Created Equal: A Heterogeneity-Aware Approach to Pre-ranking in Recommender Systems

✇cs.AI, q-bio.NC updates on arXiv.org
By: Pengfei Tong · Siyuan Chen · Chenwei Zhang · Bo Wang · Qi Pi · Pixun Li · Zuotao Liu
arXiv:2603.03770v1 Announce Type: cross Abstract: Most large-scale recommender systems follow a multi-stage cascade of retrieval, pre-ranking, ranking, and re-ranking. A key challenge at the pre-ranking stage arises from the heterogeneity of training instances sampled from coarse-grained retrieval results, fine-grained ranking signals, and exposure feedback. Our analysis reveals that prevailing pre-ranking methods, which indiscriminately mix heterogeneous samples, suffer from gradient conflicts: hard samples dominate training while easy ones remain underutilized, leading to suboptimal performance. We further show that the common practice of uniformly scaling model complexity across all samples is inefficient, as it overspends computation on easy cases and slows training without proportional gains. To address these limitations, this paper presents Heterogeneity-Aware Adaptive Pre-ranking (HAP), a unified framework that mitigates gradient conflicts through conflict-sensitive sampling coupled with tailored loss design, while adaptively allocating computational budgets across candidates. Specifically, HAP disentangles easy and hard samples, directing each subset along dedicated optimization paths. Building on this separation, it first applies lightweight models to all candidates for efficient coverage, and further engages stronger models on the hard ones, maintaining accuracy while reducing cost. This approach not only improves pre-ranking effectiveness but also provides a practical perspective on scaling strategies in industrial recommender systems. HAP has been deployed in the Toutiao production system for 9 months, yielding up to 0.4% improvement in user app usage duration and 0.05% in active days, without additional computational cost. We also release a large-scale industrial hybrid-sample dataset to enable the systematic study of source-driven candidate heterogeneity in pre-ranking.
  •  
  • ↗️
↗️

DisenReason: Behavior Disentanglement and Latent Reasoning for Shared-Account Sequential Recommendation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Jiawei Cheng · Min Gao · Zongwei Wang · Xiaofei Zhu · Zhiyi Liu · Wentao Li · Wei Li · Huan Wu
arXiv:2603.03782v1 Announce Type: cross Abstract: Shared-account usage is common on streaming and e-commerce platforms, where multiple users share one account. Existing shared-account sequential recommendation (SSR) methods often assume a fixed number of latent users per account, limiting their ability to adapt to diverse sharing patterns and reducing recommendation accuracy. Recent latent reasoning technique applied in sequential recommendation (SR) generate intermediate embeddings from the user embedding (e.g, last item embedding) to uncover users' potential interests, which inspires us to treat the problem of inferring the number of latent users as generating a series of intermediate embeddings, shifting from inferring preferences behind user to inferring the users behind account. However, the last item cannot be directly used for reasoning in SSR, as it can only represent the behavior of the most recent latent user, rather than the collective behavior of the entire account. To address this, we propose DisenReason, a two-stage reasoning method tailored to SSR. DisenReason combines behavior disentanglement stage from frequency-domain perspective to create a collective and unified account behavior representation, which serves as a pivot for latent user reasoning stage to infer the number of users behind the account. Experiments on four benchmark datasets show that DisenReason consistently outperforms all state-of-the-art baselines across four benchmark datasets, achieving relative improvements of up to 12.56\% in MRR@5 and 6.06\% in Recall@20.
  •  
  • ↗️
↗️

EDU-MATRIX: A Society-Centric Generative Cognitive Digital Twin Architecture for Secondary Education

✇cs.AI, q-bio.NC updates on arXiv.org
By: Wenjing Zhai · Jianbin Zhang · Tao Liu
arXiv:2602.18705v1 Announce Type: cross Abstract: Existing multi-agent simulations often suffer from the "Agent-Centric Paradox": rules are hard-coded into individual agents, making complex social dynamics rigid and difficult to align with educational values. This paper presents EDU-MATRIX, a society-centric generative cognitive digital twin architecture that shifts the paradigm from simulating "people" to simulating a "social space with a gravitational field." We introduce three architectural contributions: (1) An Environment Context Injection Engine (ECIE), which acts as a "social microkernel," dynamically injecting institutional rules (Gravity) into agents based on their spatial-temporal coordinates; (2) A Modular Logic Evolution Protocol (MLEP), where knowledge exists as "fluid" capsules that agents synthesize to generate new paradigms, ensuring high dialogue consistency (94.1%); and (3) Endogenous Alignment via Role-Topology, where safety constraints emerge from the agent's position in the social graph rather than external filters. Deployed as a digital twin of a secondary school with 2,400 agents, the system demonstrates how "social gravity" (rules) and "cognitive fluids" (knowledge) interact to produce emergent, value-aligned behaviors (Social Clustering Coefficient: 0.72).
  •  
  • ↗️
↗️

MolReasoner: Toward Effective and Interpretable Reasoning for Molecular LLMs

✇cs.AI, q-bio.NC updates on arXiv.org
By: Guojiang Zhao · Zixiang Lu · Yutang Ge · Sihang Li · Zheng Cheng · Haitao Lin · Lirong Wu · Hanchen Xia · Hengxing Cai · Wentao Guo · Hongshuai Wang · Mingjun Xu · Siyu Zhu · Guolin Ke · Linfeng Zhang · Zhifeng Gao
arXiv:2508.02066v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have shown impressive performance across various domains, but their ability to perform molecular reasoning remains underexplored. Existing methods mostly rely on general-purpose prompting, which lacks domain-specific molecular semantics, or fine-tuning, which faces challenges in interpretability and reasoning depth, often leading to structural and textual hallucinations. To address these issues, we introduce MolReasoner, a two-stage framework that transitions LLMs from memorization to high-fidelity chemical reasoning. In the Mol-SFT stage, knowledge-enhanced Chain-of-Thought (CoT) data provides a strong foundation, while the Mol-RL stage refines reasoning using a novel, task-adaptive reward system to mitigate hallucinations. Extensive evaluations demonstrate that MolReasoner significantly outperforms a wide range of strong baselines in both molecule generation and captioning tasks. Further analyses highlight the framework's synergistic design and its ability to produce more interpretable outputs. Our work presents a principled and effective new approach for advancing high-fidelity molecular reasoning.
  •  
  • ↗️
❌