Feng Liu

Jiacheng Zhang, Feng Liu, Chao Du et al.

A line of recent training-free methods for mitigating hallucinations in large vision-language models (LVLMs) operates by amplifying attention to visual tokens during autoregressive generation within a single forward pass. We refer to this paradigm as visual attention amplification (VAA). In this paper, we identify a dual failure pattern in existing VAA methods caused by their use of a fixed amplification factor across generation steps: it can be too weak at some steps, leaving hallucinations unresolved, while too strong at others, introducing new hallucinations. Motivated by this finding, we propose Step-wise Adaptive Visual Attention Amplification (SAVAA), a new VAA framework that estimates hallucination risk for each generated token and uses the estimated risk to adaptively amplify visual attention at the next generation step. Specifically, we introduce Visual Grounding Entropy (VGE), a lightweight hallucination-risk estimator that augments predictive entropy with visual grounding, assigning higher risk to tokens that are uncertain, weakly grounded in the image, or both. Guided by VGE, SAVAA uses the estimated risk to calibrate the VAA factor for the next generation step, applying stronger amplification to higher-risk steps and weaker amplification to lower-risk steps. Across LLaVA-NeXT-7B, Qwen3-VL-8B, and InternVL3.5-8B, SAVAA significantly outperforms baseline methods on generative hallucination benchmarks such as CHAIR, SHR and AMBER. Code is available at: https://github.com/JiachengZ01/SAVVA.

In Chong Choi, Jiacheng Zhang, Feng Liu et al.

Multi-turn jailbreak attacks are effective against text-only large language models (LLMs) by gradually introducing malicious content across turns. When extended to large vision-language models (LVLMs), we find that naively adding visual inputs can cause existing multi-turn jailbreaks to be easily defended. For example, overly malicious visual input will easily trigger the defense mechanism of safety-aligned LVLMs, making the response more conservative. To address this, we propose MAPA: a multi-turn adaptive prompting attack that 1) at each turn, alternates text-vision attack actions to elicit the most malicious response; and 2) across turns, adjusts the attack trajectory through iterative back-and-forth refinement to gradually amplify response maliciousness. This two-level design enables MAPA to consistently outperform state-of-the-art methods, improving attack success rates by 11-35% on recent benchmarks against LLaVA-V1.6-Mistral-7B, Qwen2.5-VL-7B-Instruct, Llama-3.2-Vision-11B-Instruct and GPT-4o-mini.

Mingfei Lu, Mengjia Wu, Feng Liu et al.

Memory is critical for enabling large language model (LLM) based agents to maintain coherent behavior over long-horizon interactions. However, existing agent memory systems suffer from two key gaps: they rely on a one-size-fits-all memory structure and do not model memory structure selection as a context-adaptive decision, limiting their ability to handle heterogeneous interaction patterns and resulting in suboptimal performance. We propose a unified framework, FluxMem, that enables adaptive memory organization for LLM agents. Our framework equips agents with multiple complementary memory structures. It explicitly learns to select among these structures based on interaction-level features, using offline supervision derived from downstream response quality and memory utilization. To support robust long-horizon memory evolution, we further introduce a three-level memory hierarchy and a Beta Mixture Model-based probabilistic gate for distribution-aware memory fusion, replacing brittle similarity thresholds. Experiments on two long-horizon benchmarks, PERSONAMEM and LoCoMo, demonstrate that our method achieves average improvements of 9.18% and 6.14%.

Tongze Zhang, Jun-En Ding, Melik Ozolcer et al.

Alzheimer's disease (AD) has become a prevalent neurodegenerative disease worldwide. Traditional diagnosis still relies heavily on medical imaging and clinical assessment by physicians, which is often time-consuming and resource-intensive in terms of both human expertise and healthcare resources. In recent years, large language models (LLMs) have been increasingly applied to the medical field using electronic health records (EHRs), yet their application in Alzheimer's disease assessment remains limited, particularly given that AD involves complex multifactorial etiologies that are difficult to observe directly through imaging modalities. In this work, we propose leveraging LLMs to perform Chain-of-Thought (CoT) reasoning on patients' clinical EHRs. Unlike direct fine-tuning of LLMs on EHR data for AD classification, our approach utilizes LLM-generated CoT reasoning paths to provide the model with explicit diagnostic rationale for AD assessment, followed by structured CoT-based predictions. This pipeline not only enhances the model's ability to diagnose intrinsically complex factors but also improves the interpretability of the prediction process across different stages of AD progression. Experimental results demonstrate that the proposed CoT-based diagnostic framework significantly enhances stability and diagnostic performance across multiple CDR grading tasks, achieving up to a 15% improvement in F1 score compared to the zero-shot baseline method.