❌

Normal view

Received — 27 May 2026 ⏭

A Sober Look at Agentic Misalignment in Automated Workflows

✇cs.AI, q-bio.NC updates on arXiv.org
By: Wenqian Ye · Bo Yuan · Zhichao Xu · Yijun Tian · Yawei Wang · Henry Kautz · Aidong Zhang
26 May 2026 at 12:00
arXiv:2605.24197v1 Announce Type: new Abstract: We study a class of emergent misalignment in multi-agent systems (MAS), with a focus on automated workflows, which we refer to agentic misalignment. Although these systems can solve complex tasks, they often fail because agents act according to implicit proxy utilities that do not align with the intended human goals. We formally define these behaviors and analyze them within a Bayesian framework, showing that generic utilities naturally lead to posterior collapse of agents in automated workflows. To address this issue, we propose Agentic Evidence Attribution (AEA), a novel alignment paradigm that improves agent posteriors using context-specific evidence. AEA reasons over agent actions and provides structured evidence to correct misaligned behavior during collaboration. To better understand the role of evidence, we study two instantiations of AEA: self-reflection (internal evidence from the model) and weak-to-strong generalization (external evidence on the agentic trajectory). We show that a small evidence model effectively aligns the MAS by providing orthogonal failure attribution. Our results clarify the sources of agentic misalignment in automated workflows and show that evidence-based alignment can effectively improve agent collaboration and leads to reliable multi-agent systems built on automated workflows.
Received — 4 April 2026 ⏭

Scale over Preference: The Impact of AI-Generated Content on Online Content Ecology

✇cs.AI, q-bio.NC updates on arXiv.org
By: Tianhao Shi · Yang Zhang · Xiaoyan Zhao · Fengbin Zhu · Chenyi Lei · Han Li · Wenwu Ou · Yang Song · Yongdong Zhang · Fuli Feng
3 April 2026 at 12:00
arXiv:2604.01690v1 Announce Type: new Abstract: The rapid proliferation of Artificial Intelligence-Generated Content (AIGC) is fundamentally restructuring online content ecologies, necessitating a rigorous examination of its behavioral and distributional implications. Leveraging a comprehensive longitudinal dataset comprising tens of millions of users from a leading Chinese video-sharing platform, this study elucidated the distinct creation and consumption behaviors characterizing AIGC versus Human-Generated Content (HGC). We identified a prevalent scale-over-preference dynamic, wherein AIGC creators achieve aggregate engagement comparable to HGC creators through high-volume production, despite a marked consumer preference for HGC. Deeper analysis uncovered the ability of the algorithmic content distribution mechanism in moderating these competing interests regarding AIGC. These findings advocated for the implementation of AIGC-sensitive distribution algorithms and precise governance frameworks to ensure the long-term health of the online content platforms.

Stabilizing Rubric Integration Training via Decoupled Advantage Normalization

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zelin Tan · Zhouliang Yu · Bohan Lin · Zijie Geng · Hejia Geng · Yudong Zhang · Mulei Zhang · Yang Chen · Shuyue Hu · Zhenfei Yin · Chen Zhang · Lei Bai
3 April 2026 at 12:00
arXiv:2603.26535v2 Announce Type: replace Abstract: We propose Process-Aware Policy Optimization (PAPO), a method that integrates process-level evaluation into Group Relative Policy Optimization (GRPO) through decoupled advantage normalization, to address two limitations of existing reward designs. Outcome reward models (ORM) evaluate only final-answer correctness, treating all correct responses identically regardless of reasoning quality, and gradually lose the advantage signal as groups become uniformly correct. Process reward models (PRM) offer richer supervision, but directly using PRM scores causes reward hacking, where models exploit verbosity to inflate scores while accuracy collapses. PAPO resolves both by composing the advantage from an outcome component Aout, derived from ORM and normalized over all responses, and a process component Aproc, derived from a rubric-based PRM and normalized exclusively among correct responses. This decoupled design ensures that Aout anchors training on correctness while Aproc differentiates reasoning quality without distorting the outcome signal. Experiments across multiple model scales and six benchmarks demonstrate that PAPO consistently outperforms ORM, reaching 51.3% vs.\ 46.3% on OlympiadBench while continuing to improve as ORM plateaus and declines.
Received — 26 March 2026 ⏭

SortedRL: Accelerating RL Training for LLMs through Online Length-Aware Scheduling

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yiqi Zhang · Huiqiang Jiang · Xufang Luo · Zhihe Yang · Chengruidong Zhang · Yifei Shen · Dongsheng Li · Yuqing Yang · Lili Qiu · Yang You
25 March 2026 at 12:00
arXiv:2603.23414v1 Announce Type: cross Abstract: Scaling reinforcement learning (RL) has shown strong promise for enhancing the reasoning abilities of large language models (LLMs), particularly in tasks requiring long chain-of-thought generation. However, RL training efficiency is often bottlenecked by the rollout phase, which can account for up to 70% of total training time when generating long trajectories (e.g., 16k tokens), due to slow autoregressive generation and synchronization overhead between rollout and policy updates. We propose SortedRL, an online length-aware scheduling strategy designed to address this bottleneck by improving rollout efficiency and maintaining training stability. SortedRL reorders rollout samples based on output lengths, prioritizing short samples forming groups for early updates. This enables large rollout batches, flexible update batches, and near on-policy micro-curriculum construction simultaneously. To further accelerate the pipeline, SortedRL incorporates a mechanism to control the degree of off-policy training through a cache-based mechanism, and is supported by a dedicated RL infrastructure that manages rollout and update via a stateful controller and rollout buffer. Experiments using LLaMA-3.1-8B and Qwen-2.5-32B on diverse tasks, including logical puzzles, and math challenges like AIME 24, Math 500, and Minerval, show that SortedRL reduces RL training bubble ratios by over 50%, while attaining 3.9% to 18.4% superior performance over baseline given same amount of data.

Hypoxia-related and immune phenotype-related fusion model for non-invasive prognostication of hepatocellular carcinoma treated by TACE: a multicentre study

✇Omics in Hepatocellular
By: Yusheng Guo · Guilin Zhang · Xiaona Fu · Shanshan Jiang · Yi Li · Shanmei Li · Xiaofang Guo · Xiaolin Zhang · Chang Zhao · Rong Ding · Lei Yu · Xuegang Yang · Kai Zhao · Yuxin Sun · QiuPing Liu · YuDong Zhang · Xuhua Duan · Hui Zhao · Jiahua Zou · Bin Liang · Lian Yang · Chuansheng Zheng · Xuefeng Kan
19 March 2026 at 18:00

Gut. 2026 Mar 30:gutjnl-2025-337938. doi: 10.1136/gutjnl-2025-337938. Online ahead of print.

ABSTRACT

BACKGROUND: Survival outcomes after transarterial chemoembolisation (TACE) vary in hepatocellular carcinoma (HCC) patients, and existing prognostic scores and imaging models often lack generalisability and biological interpretability.

OBJECTIVE: To develop and validate a multimodal prognostication model for HCC that allows for a precise assessment of survival outcomes of HCC patients receiving TACE therapy.

DESIGN: This study enrolled 1448 HCC patients, including a TACE cohort (n=1349), a biomarker subset from a randomised trial (n=41), a single-cell RNA sequencing cohort and The Cancer Genome Atlas (TCGA) HCC cohort (n=50). Pre-treatment contrast-enhanced CT images were used to construct deep learning and conventional radiomic models. The early-fusion and late-fusion models (LFMs) were compared, and a clinical-radiologic model (CRM) was formed by integrating the better-performing LFM with clinical variables. Using TCGA data and single-cell transcriptomic profiles, the differences between high-score and low-score groups in tumour immune microenvironment, cellular functional states and key signalling pathways were investigated.

RESULTS: The CRM effectively stratified patients' survival across multiple independent cohorts and achieved more granular risk stratification than the existing clinical models. Multi-omic analyses revealed that in the LFM high-score group, myelocytomatosis oncogene was activated, epithelial-mesenchymal transition enhanced, glycolysis upregulated and hypoxia pathway activated. Single-cell transcriptomic data confirmed that virtually all cell types in high-risk patients scored high in hypoxia, and cytotoxic T cells had a reduced cytotoxic activity.

CONCLUSION: The CRM model can non-invasively predict the prognosis of HCC patients treated by TACE therapy.

PMID:41856522 | DOI:10.1136/gutjnl-2025-337938

There are no more articles
❌