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.

Traditional epidemic intelligence relies heavily on human epidemiologists for data interpretation and reporting, which makes it resource intensive, slow to respond, and vulnerable to variability in professional expertise. To overcome these limitations, we propose an expanded conceptual epidemic intelligence quadripartite framework that extends the classical trinity of (1) surveillance, (2) risk evaluation, and (3) early warning with a fourth pillar, (4) decision support and intervention optimization through AI agents. Acting as 24/7 digital epidemiologists, multiagent systems can integrate heterogeneous signals from multisource surveillance systems, conduct contextual risk evaluation and adaptive forecasting, generate tailored early warnings, and provide actionable recommendations for targeted control—closing the loop between detection and response. Embedding interpretability and mandatory human-in-the-loop oversight enhances trust and accountability. Nonetheless, real-world deployment requires addressing context-specific challenges of data quality, interoperability, robustness, governance, circular reporting, and equity. If designed with transparency, inclusiveness, and resilience, AI agents have the potential to transform epidemic intelligence into a continuously adaptive and globally connected system.

Oncogene, Published online: 06 March 2026; doi:10.1038/s41388-026-03707-x

PMM2 interacts with TRIM28 to recruit E2F4 and promote KIFC3-mediated tumor glycolysis and colorectal cancer progression