Chengzhi Liu

Chengzhi Liu, Yuzhe Yang, Sophia Xiao Pu et al.

Multimodal large language models are increasingly deployed as long-horizon agents, where memory must do more than recall: it must track an evolving world, revise what has gone stale, and surface the right evidence at decision time. Existing benchmarks measure recall over static dialogue, collapse memory into a single end-of-task accuracy, and reduce visual observations to captions, leaving us unable to localize failures to writing, maintenance, retrieval, or use. The rise of agent harnesses that author their own memory sharpens this gap, since we have no principled way to compare hand-designed pipelines with self-managing alternatives. To close these gaps, we formulate multimodal agent memory as an Action-World Interaction Loop with an observable four-stage lifecycle, and instantiate it in WorldMemArena: 400 multi-session multimodal tasks spanning Lifelong Evolution (evolving personal and task states) and Agentic Execution (memory from real observations, actions, and feedback), annotated with gold memory points, updates, distractors, and evidence chains for stage-level diagnosis. This enables the first head-to-head comparison of long-context, manually designed (RAG and external memory systems), and harness-based memory agents. Results show that: (1) better memory writing and storage do not guarantee better performance; (2) multimodal memory still struggles to fully use visual evidence; (3) systems are unstable across domains and degrade on realistic agentic trajectories; and (4) harness memory is more flexible but remains costly and less reliable.

Sophia Xiao Pu, Zhaotian Weng, Chengzhi Liu et al.

Self-play reinforcement learning trains language models on their own generated tasks, co-evolving a proposer and solver without human labels. Recent systems report strong reasoning gains, but collapse and instability are widely observed and poorly understood. The dominant response treats this as a reward-design problem. We argue instead that self-play stability is governed by two distinct levers: a data-level gate that decides which proposer-generated tasks enter the training pool, and the reward signal that updates the policy on tasks already admitted. Through controlled experiments on a Python output-prediction task and a deterministic-DSL twin task that strips pretraining priors, output ambiguity, and executor noise, we find the two levers are asymmetric. A strict gate is sufficient for stability under every reward variant we test, including a self-consistency reward with no access to ground truth; while no reward variant is sufficient once the gate is removed. This asymmetry exposes a counter-intuitive coupling we call the Grounded Proposer Paradox: a proposer with ground-truth access accelerates collapse faster than an ungrounded one when paired with a self-consistency solver, by concentrating training on clean tasks that form the fastest path to a spurious self-consistent attractor. Replacing the binary gate with a continuous strictness parameter $\varepsilon$ further reveals a two-stage phase transition: training-side metrics decouple at low $\varepsilon$, while validation accuracy holds until $\varepsilon$ is much higher. Data-level gating, not reward calibration, is the binding constraint on self-play stability.

Kaiwen Zhou, Chengzhi Liu, Xuandong Zhao et al. · berkeley

The rapid development of large reasoning models (LRMs), such as OpenAI-o3 and DeepSeek-R1, has led to significant improvements in complex reasoning over non-reasoning large language models~(LLMs). However, their enhanced capabilities, combined with the open-source access of models like DeepSeek-R1, raise serious safety concerns, particularly regarding their potential for misuse. In this work, we present a comprehensive safety assessment of these reasoning models, leveraging established safety benchmarks to evaluate their compliance with safety regulations. Furthermore, we investigate their susceptibility to adversarial attacks, such as jailbreaking and prompt injection, to assess their robustness in real-world applications. Through our multi-faceted analysis, we uncover four key findings: (1) There is a significant safety gap between the open-source reasoning models and the o3-mini model, on both safety benchmark and attack, suggesting more safety effort on open LRMs is needed. (2) The stronger the model's reasoning ability, the greater the potential harm it may cause when answering unsafe questions. (3) Safety thinking emerges in the reasoning process of LRMs, but fails frequently against adversarial attacks. (4) The thinking process in R1 models poses greater safety concerns than their final answers. Our study provides insights into the security implications of reasoning models and highlights the need for further advancements in R1 models' safety to close the gap.

Zhen Zhang, Changyi Yang, Zijie Xia et al.

Token serves as the fundamental unit of computation in modern autoregressive models, and generation length directly influences both inference cost and reasoning performance. Despite its importance, existing approaches lack fine-grained length modeling, operating primarily at the coarse-grained sequence level. We introduce the Length Value Model (LenVM), a token-level framework that models the remaining generation length. By formulating length modeling as a value estimation problem and assigning a constant negative reward to each generated token, LenVM predicts a bounded, discounted return that serves as a monotone proxy for the remaining generation horizon. This formulation yields supervision that is annotation-free, dense, unbiased, and scalable. Experiments on LLMs and VLMs demonstrate LenVM provides a highly effective signal at inference time. On the LIFEBench exact length matching task, applying LenVM to a 7B model improves the length score from 30.9 to 64.8, significantly outperforming frontier closed-source models. Furthermore, LenVM enables continuous control over the trade off between performance and efficiency. On GSM8K at a budget of 200 tokens, LenVM maintains 63% accuracy compared to 6 percent for token budget baseline. It also accurately predicts total generation length from the prompt boundary. Finally, LenVM's token-level values offer an interpretable view of generation dynamics, revealing how specific tokens shift reasoning toward shorter or longer regimes. Results demonstrate that LenVM supports a broad range of applications and token length can be effectively modeled as a token-level value signal, highlighting the potential of LenVM as a general framework for length modeling and as a length-specific value signal that could support future RL training. Code is available at https://github.com/eric-ai-lab/Length-Value-Model.

Chengzhi Liu, Yichen Guo, Yepeng Liu et al.

LLM agents increasingly run inside execution harnesses that dispatch tools, allocate resources, and route messages between specialized components. However, a harness can return a correct, benign answer over a trajectory that accesses unauthorized resources or leaks context to the wrong agent. Output-level evaluation cannot see these failures, yet most safety benchmarks score only final outputs or terminal states, even though many violations occur mid-trajectory rather than at termination. The central question is whether the harness respects user intent, permission boundaries, and information-flow constraints throughout execution. To address this gap, we propose HarnessAudit, a framework that audits full execution trajectories across boundary compliance, execution fidelity, and system stability, with a focus on multi-agent harnesses where these risks are most pronounced. We further introduce HarnessAudit-Bench, a benchmark of 210 tasks across eight real-world domains, instantiated in both single-agent and multi-agent configurations with embedded safety constraints. Evaluating ten harness configurations across frontier models and three multi-agent frameworks, we find that: (i) task completion is misaligned with safe execution, and violations accumulate with trajectory length; (ii) safety risks vary across domains, task types, and agent roles; (iii) most violations concentrate in resource access and inter-agent information transfer; and (iv) multi-agent collaboration expands the safety risk surface, while harness design sets the upper bound of safe deployment.

Haochen Xue, Chong Zhang, Chengzhi Liu et al.

The rapid development of the Internet has profoundly changed human life. Humans are increasingly expressing themselves and interacting with others on social media platforms. However, although artificial intelligence technology has been widely used in many aspects of life, its application in social media content creation is still blank. To solve this problem, we propose a new prompt word generation framework based on multi-modal information fusion, which combines multiple tasks including topic classification, sentiment analysis, scene recognition and keyword extraction to generate more comprehensive prompt words. Subsequently, we use a template containing a set of prompt words to guide ChatGPT to generate high-quality tweets. Furthermore, in the absence of effective and objective evaluation criteria in the field of content generation, we use the ChatGPT tool to evaluate the results generated by the algorithm, making large-scale evaluation of content generation algorithms possible. Evaluation results on extensive content generation demonstrate that our cue word generation framework generates higher quality content compared to manual methods and other cueing techniques, while topic classification, sentiment analysis, and scene recognition significantly enhance content clarity and its consistency with the image.

Chengzhi Liu, Zheng Tao, Zihong Luo et al.

Time series analysis and modelling constitute a crucial research area. Traditional artificial neural networks struggle with complex, non-stationary time series data due to high computational complexity, limited ability to capture temporal information, and difficulty in handling event-driven data. To address these challenges, we propose a Multi-modal Time Series Analysis Model Based on Spiking Neural Network (MTSA-SNN). The Pulse Encoder unifies the encoding of temporal images and sequential information in a common pulse-based representation. The Joint Learning Module employs a joint learning function and weight allocation mechanism to fuse information from multi-modal pulse signals complementary. Additionally, we incorporate wavelet transform operations to enhance the model's ability to analyze and evaluate temporal information. Experimental results demonstrate that our method achieved superior performance on three complex time-series tasks. This work provides an effective event-driven approach to overcome the challenges associated with analyzing intricate temporal information. Access to the source code is available at https://github.com/Chenngzz/MTSA-SNN}{https://github.com/Chenngzz/MTSA-SNN

Zihong Luo, Zheng Tao, Yuxuan Huang et al.

Recent advancements in multi-modal artificial intelligence (AI) have revolutionized the fields of stock market forecasting and heart rate monitoring. Utilizing diverse data sources can substantially improve prediction accuracy. Nonetheless, additional data may not always align with the original dataset. Interpolation methods are commonly utilized for handling missing values in modal data, though they may exhibit limitations in the context of sparse information. Addressing this challenge, we propose a Modality Completion Deep Belief Network-Based Model (MC-DBN). This approach utilizes implicit features of complete data to compensate for gaps between itself and additional incomplete data. It ensures that the enhanced multi-modal data closely aligns with the dynamic nature of the real world to enhance the effectiveness of the model. We conduct evaluations of the MC-DBN model in two datasets from the stock market forecasting and heart rate monitoring domains. Comprehensive experiments showcase the model's capacity to bridge the semantic divide present in multi-modal data, subsequently enhancing its performance. The source code is available at: https://github.com/logan-0623/DBN-generate

Yiheng Zhong, Zihong Luo, Chengzhi Liu et al.

Segment Anything Model (SAM) demonstrates powerful zero-shot capabilities; however, its accuracy and robustness significantly decrease when applied to medical image segmentation. Existing methods address this issue through modality fusion, integrating textual and image information to provide more detailed priors. In this study, we argue that the granularity of text and the domain gap affect the accuracy of the priors. Furthermore, the discrepancy between high-level abstract semantics and pixel-level boundary details in images can introduce noise into the fusion process. To address this, we propose Prior-Guided SAM (PG-SAM), which employs a fine-grained modality prior aligner to leverage specialized medical knowledge for better modality alignment. The core of our method lies in efficiently addressing the domain gap with fine-grained text from a medical LLM. Meanwhile, it also enhances the priors' quality after modality alignment, ensuring more accurate segmentation. In addition, our decoder enhances the model's expressive capabilities through multi-level feature fusion and iterative mask optimizer operations, supporting unprompted learning. We also propose a unified pipeline that effectively supplies high-quality semantic information to SAM. Extensive experiments on the Synapse dataset demonstrate that the proposed PG-SAM achieves state-of-the-art performance. Our code is released at https://github.com/logan-0623/PG-SAM.

Chong Zhang, Mingyu Jin, Qinkai Yu et al.

Current large language models (LLMs) provide a strong foundation for large-scale user-oriented natural language tasks. A large number of users can easily inject adversarial text or instructions through the user interface, thus causing LLMs model security challenges. Although there is currently a large amount of research on prompt injection attacks, most of these black-box attacks use heuristic strategies. It is unclear how these heuristic strategies relate to the success rate of attacks and thus effectively improve model robustness. To solve this problem, we redefine the goal of the attack: to maximize the KL divergence between the conditional probabilities of the clean text and the adversarial text. Furthermore, we prove that maximizing the KL divergence is equivalent to maximizing the Mahalanobis distance between the embedded representation $x$ and $x'$ of the clean text and the adversarial text when the conditional probability is a Gaussian distribution and gives a quantitative relationship on $x$ and $x'$. Then we designed a simple and effective goal-guided generative prompt injection strategy (G2PIA) to find an injection text that satisfies specific constraints to achieve the optimal attack effect approximately. It is particularly noteworthy that our attack method is a query-free black-box attack method with low computational cost. Experimental results on seven LLM models and four datasets show the effectiveness of our attack method.

Chengzhi Liu, Zhongxing Xu, Qingyue Wei et al.

Test-time compute has empowered multimodal large language models to generate extended reasoning chains, yielding strong performance on tasks such as multimodal math reasoning. However, this improved reasoning ability often comes with increased hallucination: as generations become longer, models tend to drift away from image-grounded content and rely more heavily on language priors. Attention analysis shows that longer reasoning chains lead to reduced focus on visual inputs, which contributes to hallucination. To systematically study this phenomenon, we introduce RH-AUC, a metric that quantifies how a model's perception accuracy changes with reasoning length, allowing us to evaluate whether the model preserves visual grounding during reasoning. We also release RH-Bench, a diagnostic benchmark that spans a variety of multimodal tasks, designed to assess the trade-off between reasoning ability and hallucination. Our analysis reveals that (i) larger models typically achieve a better balance between reasoning and perception, and (ii) this balance is influenced more by the types and domains of training data than by its overall volume. These findings underscore the importance of evaluation frameworks that jointly consider both reasoning quality and perceptual fidelity.

Haochen Xue, Feilong Tang, Ming Hu et al.

Recent multimodal large language models (MLLMs) have demonstrated significant potential in open-ended conversation, generating more accurate and personalized responses. However, their abilities to memorize, recall, and reason in sustained interactions within real-world scenarios remain underexplored. This paper introduces MMRC, a Multi-Modal Real-world Conversation benchmark for evaluating six core open-ended abilities of MLLMs: information extraction, multi-turn reasoning, information update, image management, memory recall, and answer refusal. With data collected from real-world scenarios, MMRC comprises 5,120 conversations and 28,720 corresponding manually labeled questions, posing a significant challenge to existing MLLMs. Evaluations on 20 MLLMs in MMRC indicate an accuracy drop during open-ended interactions. We identify four common failure patterns: long-term memory degradation, inadequacies in updating factual knowledge, accumulated assumption of error propagation, and reluctance to say no. To mitigate these issues, we propose a simple yet effective NOTE-TAKING strategy, which can record key information from the conversation and remind the model during its responses, enhancing conversational capabilities. Experiments across six MLLMs demonstrate significant performance improvements.

Feilong Tang, Chengzhi Liu, Zhongxing Xu et al.

Recent advancements in multimodal large language models (MLLMs) have significantly improved performance in visual question answering. However, they often suffer from hallucinations. In this work, hallucinations are categorized into two main types: initial hallucinations and snowball hallucinations. We argue that adequate contextual information can be extracted directly from the token interaction process. Inspired by causal inference in the decoding strategy, we propose to leverage causal masks to establish information propagation between multimodal tokens. The hypothesis is that insufficient interaction between those tokens may lead the model to rely on outlier tokens, overlooking dense and rich contextual cues. Therefore, we propose to intervene in the propagation process by tackling outlier tokens to enhance in-context inference. With this goal, we present FarSight, a versatile plug-and-play decoding strategy to reduce attention interference from outlier tokens merely by optimizing the causal mask. The heart of our method is effective token propagation. We design an attention register structure within the upper triangular matrix of the causal mask, dynamically allocating attention to capture attention diverted to outlier tokens. Moreover, a positional awareness encoding method with a diminishing masking rate is proposed, allowing the model to attend to further preceding tokens, especially for video sequence tasks. With extensive experiments, FarSight demonstrates significant hallucination-mitigating performance across different MLLMs on both image and video benchmarks, proving its effectiveness.

Chengzhi Liu, Zile Huang, Zhe Chen et al.

Ophthalmologists typically require multimodal data sources to improve diagnostic accuracy in clinical decisions. However, due to medical device shortages, low-quality data and data privacy concerns, missing data modalities are common in real-world scenarios. Existing deep learning methods tend to address it by learning an implicit latent subspace representation for different modality combinations. We identify two significant limitations of these methods: (1) implicit representation constraints that hinder the model's ability to capture modality-specific information and (2) modality heterogeneity, causing distribution gaps and redundancy in feature representations. To address these, we propose an Incomplete Modality Disentangled Representation (IMDR) strategy, which disentangles features into explicit independent modal-common and modal-specific features by guidance of mutual information, distilling informative knowledge and enabling it to reconstruct valuable missing semantics and produce robust multimodal representations. Furthermore, we introduce a joint proxy learning module that assists IMDR in eliminating intra-modality redundancy by exploiting the extracted proxies from each class. Experiments on four ophthalmology multimodal datasets demonstrate that the proposed IMDR outperforms the state-of-the-art methods significantly.

Zile Huang, Chong Zhang, Mingyu Jin et al.

While deep learning-based general object detection has made significant strides in recent years, the effectiveness and efficiency of small object detection remain unsatisfactory. This is primarily attributed not only to the limited characteristics of such small targets but also to the high density and mutual overlap among these targets. The existing transformer-based small object detectors do not leverage the gap between accuracy and inference speed. To address challenges, we propose methods enhancing sampling within an end-to-end framework. Sample Points Refinement (SPR) constrains localization and attention, preserving meaningful interactions in the region of interest and filtering out misleading information. Scale-aligned Target (ST) integrates scale information into target confidence, improving classification for small object detection. A task-decoupled Sample Reweighting (SR) mechanism guides attention toward challenging positive examples, utilizing a weight generator module to assess the difficulty and adjust classification loss based on decoder layer outcomes. Comprehensive experiments across various benchmarks reveal that our proposed detector excels in detecting small objects. Our model demonstrates a significant enhancement, achieving a 2.9\% increase in average precision (AP) over the state-of-the-art (SOTA) on the VisDrone dataset and a 1.7\% improvement on the SODA-D dataset.

Xinkun Wang, Yifang Wang, Senwei Liang et al.

This paper discusses how ophthalmologists often rely on multimodal data to improve diagnostic accuracy. However, complete multimodal data is rare in real-world applications due to a lack of medical equipment and concerns about data privacy. Traditional deep learning methods typically address these issues by learning representations in latent space. However, the paper highlights two key limitations of these approaches: (i) Task-irrelevant redundant information (e.g., numerous slices) in complex modalities leads to significant redundancy in latent space representations. (ii) Overlapping multimodal representations make it difficult to extract unique features for each modality. To overcome these challenges, the authors propose the Essence-Point and Disentangle Representation Learning (EDRL) strategy, which integrates a self-distillation mechanism into an end-to-end framework to enhance feature selection and disentanglement for more robust multimodal learning. Specifically, the Essence-Point Representation Learning module selects discriminative features that improve disease grading performance. The Disentangled Representation Learning module separates multimodal data into modality-common and modality-unique representations, reducing feature entanglement and enhancing both robustness and interpretability in ophthalmic disease diagnosis. Experiments on multimodal ophthalmology datasets show that the proposed EDRL strategy significantly outperforms current state-of-the-art methods.

Chengzhi Liu, Yuzhe Yang, Yue Fan et al.

Recent advancements in Multimodal Large Language Models (MLLMs) have significantly enhanced cross-modal understanding and reasoning by incorporating Chain-of-Thought (CoT) reasoning in the semantic space. Building upon this, recent studies extend the CoT mechanism to the visual modality, enabling models to integrate visual information during reasoning through external tools or explicit image generation. However, these methods remain dependent on explicit step-by-step reasoning, unstable perception-reasoning interaction and notable computational overhead. Inspired by human cognition, we posit that thinking unfolds not linearly but through the dynamic interleaving of reasoning and perception within the mind. Motivated by this perspective, we propose DMLR, a test-time Dynamic Multimodal Latent Reasoning framework that employs confidence-guided latent policy gradient optimization to refine latent think tokens for in-depth reasoning. Furthermore, a Dynamic Visual Injection Strategy is introduced, which retrieves the most relevant visual features at each latent think token and updates the set of best visual patches. The updated patches are then injected into latent think token to achieve dynamic visual-textual interleaving. Experiments across seven multimodal reasoning benchmarks and various model architectures demonstrate that DMLR significantly improves reasoning and perception performance while maintaining high inference efficiency.

Chengzhi Liu, Yuzhe Yang, Kaiwen Zhou et al.

The promotion of academic papers has become an important means of enhancing research visibility. However, existing automated methods struggle limited storytelling, insufficient aesthetic quality, and constrained self-adjustment, making it difficult to achieve efficient and engaging dissemination. At the heart of those challenges is a simple principle: \emph{there is no way to improve it when you cannot evaluate it right}. To address this, we introduce \textbf{EvoPresent}, a self-improvement agent framework that unifies coherent narratives, aesthetic-aware designs, and realistic presentation delivery via virtual characters. Central to EvoPresent is \textbf{PresAesth}, a multi-task reinforcement learning (RL) aesthetic model that provides reliable aesthetic scoring, defect adjustment, and comparative feedback, enabling iterative self-improvement even under limited aesthetic training data. To systematically evaluate the methods, we introduce \textbf{EvoPresent Benchmark}, a comprehensive benchmark comprising: \textit{Presentation Generation Quality}, built on 650 top-tier AI conference papers with multimodal resources (slides, videos and scripts) to assess both content and design; and \textit{Aesthetic Awareness}, consisting of 2,000 slide pairs with varying aesthetic levels, supporting joint training and evaluation on scoring, defect adjustment, and comparison. Our findings highlight that (i) High-quality feedback is essential for agent self-improvement, while initial capability alone does not guarantee effective self-correction. (ii) Automated generation pipelines exhibit a trade-off between visual design and content construction. (iii) Multi-task RL training shows stronger generalization in aesthetic awareness tasks.