❌

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 — 8 April 2026 ⏭

Adaptive Stopping for Multi-Turn LLM Reasoning

✇cs.AI, q-bio.NC updates on arXiv.org
By: Xiaofan Zhou · Huy Nguyen · Bo Yu · Chenxi Liu · Lu Cheng
7 April 2026 at 12:00
arXiv:2604.01413v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) increasingly rely on multi-turn reasoning and interaction, such as adaptive retrieval-augmented generation (RAG) and ReAct-style agents, to answer difficult questions. These methods improve accuracy by iteratively retrieving information, reasoning, or acting, but introduce a key challenge: \textbf{When should the model stop?} Existing approaches rely on heuristic stopping rules or fixed turn budgets and provide no formal guarantees that the final prediction still contains the correct answer. This limitation is particularly problematic in high-stakes domains such as finance and healthcare, where unnecessary turns increase cost and latency, while stopping too early risks incorrect decisions. Conformal prediction (CP) provides formal coverage guarantees, but existing LLM-CP methods only apply to a single model output and cannot handle multi-turn pipelines with adaptive stopping. To address this gap, we propose Multi-Turn Language Models with Conformal Prediction (MiCP), the first CP framework for multi-turn reasoning. MiCP allocates different error budgets across turns, enabling the model to stop early while maintaining an overall coverage guarantee. We demonstrate MiCP on adaptive RAG and ReAct, where it achieves the target coverage on both single-hop and multi-hop question answering benchmarks while reducing the number of turns, inference cost, and prediction set size. We further introduce a new metric that jointly evaluates coverage validity and answering efficiency.
Received — 4 April 2026 ⏭

Multi-omics and experimental validation identify USP54 as a prognostic deubiquitinase promoting pancreatic ductal adenocarcinoma progression within the immune microenvironment

✇(Multiomics OR Omics) AND (Pancreatic)
By: Zibo Yuan · Zhiwei Yu · Qiuran Xu · Dongsheng Huang · Di Cui
3 April 2026 at 18:00

Front Immunol. 2026 Mar 18;17:1791707. doi: 10.3389/fimmu.2026.1791707. eCollection 2026.

ABSTRACT

BACKGROUND: Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal malignancy with a complex tumor ecosystem that contributes to its progression. Deubiquitinases (DUBs) are vital regulators in cancer. However, the overall activity of DUBs and their role in driving PDAC progression within immune microenvironment remain largely unknown.

METHODS: We employed an integrative multi-omics strategy combining machine learning (ML) on bulk transcriptomic data, single-cell RNA sequencing and spatial transcriptomic profiling. We applied Coxnet and Fuzzy SVM for prognostic modeling, inferCNV for malignant cell identification, SCENIC for transcription factor regulon analysis, LIANA+ for inferring inter-cellular communication networks and cell2location for spatial deconvolution. USP54 expression was detected by real-time quantitative PCR, western blotting and immunohistochemistry. USP54 function was validated through in vitro and in vivo assays.

RESULTS: ML-based pathway analysis revealed post-translational modification as a major prognostic category, within which elevated DUBs activity emerged as an independent adverse prognostic factor. At the single-cell level, USP54 was upregulated along the trajectory of malignant ductal cells and correlated with an inflamed tumor microenvironment. Cell-cell communication analysis predicted signaling from monocytes/macrophages to tumor cells via the THBS1-integrin ligand-receptor pair. This immune-derived signaling potentially converged on KLF5-positive tumor cells, with KLF5 identified as a putative transcriptional activator of USP54. Spatial transcriptomics validated the co-localization of USP54 expression, elevated DUB activity, and KRAS signaling within specific tumor niches adjacent to THBS1-enriched immune regions. High USP54 expression was frequently observed in PDAC tissues and associated with poor patient survival. More importantly, in both BxPC-3 and PANC-1 cell lines, USP54 knockdown suppressed cell proliferation and metastasis, whereas its overexpression enhanced these malignant phenotypes. Subcutaneous xenograft growth and tail vein injection experiments validated these findings in vivo.

CONCLUSIONS: Our comprehensive multi-omics analysis and experimental validation identify the deubiquitinase USP54 as a novel promoter of PDAC progression within a spatially organized tumor-immune microenvironment. These findings suggest USP54 as both a candidate prognostic biomarker and a potential therapeutic target for this lethal malignancy.

PMID:41929495 | PMC:PMC13038871 | DOI:10.3389/fimmu.2026.1791707

MM-ReCoder: Advancing Chart-to-Code Generation with Reinforcement Learning and Self-Correction

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zitian Tang · Xu Zhang · Jianbo Yuan · Yang Zou · Varad Gunjal · Songyao Jiang · Davide Modolo
3 April 2026 at 12:00
arXiv:2604.01600v1 Announce Type: new Abstract: Multimodal Large Language Models (MLLMs) have recently demonstrated promising capabilities in multimodal coding tasks such as chart-to-code generation. However, existing methods primarily rely on supervised fine-tuning (SFT), which requires the model to learn code patterns through chart-code pairs but does not expose the model to a code execution environment. Moreover, while self-correction through execution feedback offers a potential route to improve coding quality, even state-of-the-art MLLMs have been shown to struggle with effective self-correction. In this work, we introduce MM-ReCoder, a chart-to-code generation model trained with reinforcement learning (RL) and equipped with self-correction ability. We propose a two-stage multi-turn self-correction RL strategy based on Group Relative Policy Optimization (GRPO). The first stage enhances the model's self-correction ability via rolling out a shared first turn, while the second stage improves the coding capability with full-trajectory optimization. MM-ReCoder learns to produce more accurate and executable code through the interaction with the environment and by iteratively correcting its own outputs. Our results on three chart-to-code benchmarks demonstrate the state-of-the-art performance of MM-ReCoder.

Adaptive Stopping for Multi-Turn LLM Reasoning

✇cs.AI, q-bio.NC updates on arXiv.org
By: Xiaofan Zhou · Huy Nguyen · Bo Yu · Chenxi Liu · Lu Cheng
3 April 2026 at 12:00
arXiv:2604.01413v1 Announce Type: cross Abstract: Large Language Models (LLMs) increasingly rely on multi-turn reasoning and interaction, such as adaptive retrieval-augmented generation (RAG) and ReAct-style agents, to answer difficult questions. These methods improve accuracy by iteratively retrieving information, reasoning, or acting, but introduce a key challenge: \textbf{When should the model stop?} Existing approaches rely on heuristic stopping rules or fixed turn budgets and provide no formal guarantees that the final prediction still contains the correct answer. This limitation is particularly problematic in high-stakes domains such as finance and healthcare, where unnecessary turns increase cost and latency, while stopping too early risks incorrect decisions. Conformal prediction (CP) provides formal coverage guarantees, but existing LLM-CP methods only apply to a single model output and cannot handle multi-turn pipelines with adaptive stopping. To address this gap, we propose Multi-Turn Language Models with Conformal Prediction (MiCP), the first CP framework for multi-turn reasoning. MiCP allocates different error budgets across turns, enabling the model to stop early while maintaining an overall coverage guarantee. We demonstrate MiCP on adaptive RAG and ReAct, where it achieves the target coverage on both single-hop and multi-hop question answering benchmarks while reducing the number of turns, inference cost, and prediction set size. We further introduce a new metric that jointly evaluates coverage validity and answering efficiency.
Received — 26 March 2026 ⏭

Grounding Sim-to-Real Generalization in Dexterous Manipulation: An Empirical Study with Vision-Language-Action Models

✇cs.AI, q-bio.NC updates on arXiv.org
By: Ruixing Jin · Zicheng Zhu · Ruixiang Ouyang · Sheng Xu · Bo Yue · Zhizheng Wu · Guiliang Liu
25 March 2026 at 12:00
arXiv:2603.22876v1 Announce Type: cross Abstract: Learning a generalist control policy for dexterous manipulation typically relies on large-scale datasets. Given the high cost of real-world data collection, a practical alternative is to generate synthetic data through simulation. However, the resulting synthetic data often exhibits a significant gap from real-world distributions. While many prior studies have proposed algorithms to bridge the Sim-to-Real discrepancy, there remains a lack of principled research that grounds these methods in real-world manipulation tasks, particularly their performance on generalist policies such as Vision-Language-Action (VLA) models. In this study, we empirically examine the primary determinants of Sim-to-Real generalization across four dimensions: multi-level domain randomization, photorealistic rendering, physics-realistic modeling, and reinforcement learning updates. To support this study, we design a comprehensive evaluation protocol to quantify the real-world performance of manipulation tasks. The protocol accounts for key variations in background, lighting, distractors, object types, and spatial features. Through experiments involving over 10k real-world trials, we derive critical insights into Sim-to-Real transfer. To inform and advance future studies, we release both the robotic platforms and the evaluation protocol for public access to facilitate independent verification, thereby establishing a realistic and standardized benchmark for dexterous manipulation policies.
Received — 17 February 2026 ⏭

ParaCook: On Time-Efficient Planning for Multi-Agent Systems

✇cs.AI, q-bio.NC updates on arXiv.org
By: Shiqi Zhang · Xinbei Ma · Yunqing Xu · Zouying Cao · Pengrui Lu · Haobo Yuan · Tiancheng Shen · Zhuosheng Zhang · Hai Zhao · Ming-Hsuan Yang
17 February 2026 at 13:00
arXiv:2510.11608v2 Announce Type: replace Abstract: Large Language Models (LLMs) exhibit strong reasoning abilities for planning long-horizon, real-world tasks, yet existing agent benchmarks focus on task completion while neglecting time efficiency in parallel and asynchronous operations. To address this, we present ParaCook, a benchmark for time-efficient collaborative planning. Inspired by the Overcooked game, ParaCook provides an environment for various challenging interaction planning of multi-agent systems that are instantiated as cooking tasks, with a simplified action space to isolate the core challenge of strategic parallel planning. Through a comprehensive evaluation of state-of-the-art LLMs, we find that current approaches achieve suboptimal plans, which struggle with parallel actions or coordination. Our analysis also reveals LLMs' potential on abstract tasks where they can focus on high-level parallel optimization. ParaCook provides a scalable evaluation framework with adjustable complexity, establishing a foundation for developing and assessing time efficiency-aware multi-agent planning. The code and data are available at https://github.com/zsq259/ParaCook.
There are no more articles
❌