❌

Normal view

Received β€” 27 May 2026 ⏭

HoloFair: Unified T2I Fairness Evaluation and Fair-GRPO Debiasing

βœ‡cs.AI, q-bio.NC updates on arXiv.org
By: Ruyi Chen Β· Lu Zhou Β· Xiaogang Xu Β· Chiyu Zhang Β· Jiafei Wu Β· Liming Fang
26 May 2026 at 12:00
arXiv:2605.24687v1 Announce Type: cross Abstract: Text-to-Image (T2I) models have made significant strides in visual realism and semantic consistency, yet they often perpetuate and amplify societal biases. Existing evaluation methods typically address only single-dimensional biases, lacking perspectives to uncover model biases at social-related deeper semantic levels. We introduce HoloFair, a comprehensive benchmark framework for multidimensional demographic bias analysis. Built upon our large-scale fairness-oriented dataset and the SpaFreq (Spatial-Frequency) attribute classifier, this framework proposes the Multi-attribute, Group-wise Bias Index (MGBI) metric, designed to assess both intrinsic diversity and conditional biases. Beyond evaluation, we further introduce Fair-GRPO, a reinforcement-learning-based debiasing method that alters the distribution of generative models through a designed multi-objective reward function. E.g., experiments on the SD3.5-Medium model demonstrate that Fair-GRPO significantly improves multidimensional fairness while maintaining high image quality. We also analyze potential reward hacking phenomena and provide corresponding mitigation strategies. Code and dataset are available at https://github.com/1059684669/HoloFair
Received β€” 13 April 2026 ⏭

UBTF-HSP90A-MIF stress circuit drives lenvatinib resistance and immune exclusion in hepatocellular carcinoma

βœ‡Omics in Hepatocellular
By: Shiping Chen Β· Biao Wang Β· Yanfei Wu Β· Zongjuan Li Β· Qiuyi Zheng Β· Zhigang Fu Β· Weifeng Hong Β· Shisuo Du Β· Yang Zhang
7 April 2026 at 18:00

J Adv Res. 2026 Apr 5:S2090-1232(26)00280-8. doi: 10.1016/j.jare.2026.04.002. Online ahead of print.

ABSTRACT

INTRODUCTION: The clinical benefit of combining lenvatinib with PD-1 blockade in HCC is frequently constrained by adaptive resistance and the development of an immune-cold tumor microenvironment.

OBJECTIVES: This study aimed to elucidate the molecular mechanisms underlying adaptive resistance and immune exclusion during lenvatinib-PD-1 therapy in HCC, with a particular focus on a UBTF/HSP90A/MIF regulatory circuit. We examined whether genetic or pharmacologic targeting of macrophage migration inhibitory factor (MIF) could restore lenvatinib sensitivity, remodel the tumor immune microenvironment, and serve as a predictive biomarker in clinical cohorts.

METHODS: Paired lenvatinib-sensitive and -resistant HCC models were interrogated using integrated multi-omic and functional approaches, including RNA sequencing, promoter pull-down assays, ChIP, luciferase reporter assays, PLA, and flow cytometry. Key findings were validated in patient-derived organoids and xenografts, as well as in an immunocompetent hydrodynamic HCC mouse model. Clinical relevance was evaluated in independent cohorts treated with lenvatinib plus anti-PD-1 therapy.

RESULTS: UBTF directly bound to and transcriptionally activated the HSP90A promoter, resulting in increased HSP90A expression and stabilization of MIF. MIF signaling through CD74 co-activated the PI3K-AKT and MAPK pathways, sustaining tumor cell proliferation under lenvatinib pressure. Single-cell RNA sequencing and multiplex immunohistochemistry revealed macrophage enrichment and CD8+ T-cell exclusion in resistant tumors. Genetic ablation of Mif (Alb-Cre; Mifflox/flox) or pharmacologic inhibition with 4-IPP (4-Iodo-6-phenylpyrimidine) restored lenvatinib sensitivity, reprogrammed the tumor immune microenvironment, and, when combined with PD-1 blockade, achieved superior tumor control and prolonged survival. In clinical datasets, low pretreatment MIF expression was associated with improved responses to lenvatinib plus PD-1 therapy.

CONCLUSIONS: These findings define a UBTF/HSP90A/MIF axis linking proteostasis and cytokine signaling to immune-metabolic dysfunction and lenvatinib resistance in HCC. MIF emerges as both a mechanistic driver and a predictive biomarker, supporting prospective evaluation of therapeutic strategies combining lenvatinib-PD-1 with MIF- or HSP90A-targeted interventions to personalize TKI-ICI therapy.

PMID:41946392 | DOI:10.1016/j.jare.2026.04.002

Received β€” 17 March 2026 ⏭

CA-HFP: Curvature-Aware Heterogeneous Federated Pruning with Model Reconstruction

βœ‡cs.AI, q-bio.NC updates on arXiv.org
By: Gang Hu Β· Yinglei Teng Β· Pengfei Wu Β· Shijun Ma
16 March 2026 at 12:00
arXiv:2603.12591v1 Announce Type: cross Abstract: Federated learning on heterogeneous edge devices requires personalized compression while preserving aggregation compatibility and stable convergence. We present Curvature-Aware Heterogeneous Federated Pruning (CA-HFP), a practical framework that enables each client perform structured, device-specific pruning guided by a curvature-informed significance score, and subsequently maps its compact submodel back into a common global parameter space via a lightweight reconstruction. We derive a convergence bound for federated optimization with multiple local SGD steps that explicitly accounts for local computation, data heterogeneity, and pruning-induced perturbations; from which a principled loss-based pruning criterion is derived. Extensive experiments on FMNIST, CIFAR-10, and CIFAR-100 using VGG and ResNet architectures under varying degrees of data heterogeneity demonstrate that CA-HFP preserves model accuracy while significantly reducing per-client computation and communication costs, outperforming standard federated training and existing pruning-based baselines.
Received β€” 11 March 2026 ⏭

From Thinker to Society: Security in Hierarchical Autonomy Evolution of AI Agents

βœ‡cs.AI, q-bio.NC updates on arXiv.org
By: Xiaolei Zhang Β· Lu Zhou Β· Xiaogang Xu Β· Jiafei Wu Β· Tianyu Du Β· Heqing Huang Β· Hao Peng Β· Zhe Liu
10 March 2026 at 12:00
arXiv:2603.07496v1 Announce Type: cross Abstract: Artificial Intelligence (AI) agents have evolved from passive predictive tools into active entities capable of autonomous decision-making and environmental interaction, driven by the reasoning capabilities of Large Language Models (LLMs). However, this evolution has introduced critical security vulnerabilities that existing frameworks fail to address. The Hierarchical Autonomy Evolution (HAE) framework organizes agent security into three tiers: Cognitive Autonomy (L1) targets internal reasoning integrity; Execution Autonomy (L2) covers tool-mediated environmental interaction; Collective Autonomy (L3) addresses systemic risks in multi-agent ecosystems. We present a taxonomy of threats spanning cognitive manipulation, physical environment disruption, and multi-agent systemic failures, and evaluate existing defenses while identifying key research gaps. The findings aim to guide the development of multilayered, autonomy-aware defense architectures for trustworthy AI agent systems.
Received β€” 17 February 2026 ⏭

Preventing Rank Collapse in Federated Low-Rank Adaptation with Client Heterogeneity

βœ‡cs.AI, q-bio.NC updates on arXiv.org
By: Fei Wu Β· Jia Hu Β· Geyong Min Β· Shiqiang Wang
17 February 2026 at 13:00
arXiv:2602.13486v1 Announce Type: cross Abstract: Federated low-rank adaptation (FedLoRA) has facilitated communication-efficient and privacy-preserving fine-tuning of foundation models for downstream tasks. In practical federated learning scenarios, client heterogeneity in system resources and data distributions motivates heterogeneous LoRA ranks across clients. We identify a previously overlooked phenomenon in heterogeneous FedLoRA, termed rank collapse, where the energy of the global update concentrates on the minimum shared rank, resulting in suboptimal performance and high sensitivity to rank configurations. Through theoretical analysis, we reveal the root cause of rank collapse: a mismatch between rank-agnostic aggregation weights and rank-dependent client contributions, which systematically suppresses higher-rank updates at a geometric rate over rounds. Motivated by this insight, we propose raFLoRA, a rank-partitioned aggregation method that decomposes local updates into rank partitions and then aggregates each partition weighted by its effective client contributions. Extensive experiments across classification and reasoning tasks show that raFLoRA prevents rank collapse, improves model performance, and preserves communication efficiency compared to state-of-the-art FedLoRA baselines.
There are no more articles
❌