论文检索

Linhan Cao, Wei Sun, Weixia Zhang, Xiangyang Zhu, Jun Jia, Kaiwei Zhang, Dandan Zhu, Guangtao Zhai, Xiongkuo Min

Video quality assessment (VQA) aims to objectively quantify perceptual quality degradation in alignment with human visual perception. Despite recent advances, existing VQA models still suffer from two critical limitations: poor generalization to out-of-distribution (OOD) videos and limited explainability, which restrict their applicability in real-world scenarios. To address these challenges, we propose VQAThinker, a reasoning-based VQA framework that leverages large multimodal models (LMMs) with reinforcement learning to jointly model video quality understanding and scoring, emulating human perceptual decision-making. Specifically, we adopt group relative policy optimization (GRPO), a rule-guided reinforcement learning algorithm that enables reasoning over video quality under score-level supervision, and introduce three VQA-specific rewards: (1) a bell-shaped regression reward that increases rapidly as the prediction error decreases and becomes progressively less sensitive near the ground truth; (2) a pairwise ranking reward that guides the model to correctly determine the relative quality between video pairs; and (3) a temporal consistency reward that encourages the model to prefer temporally coherent videos over their perturbed counterparts. Extensive experiments demonstrate that VQAThinker achieves state-of-the-art performance on both in-domain and OOD VQA benchmarks, showing strong generalization for video quality scoring. Furthermore, evaluations on video quality understanding tasks validate its superiority in distortion attribution and quality description compared to existing explainable VQA models and LMMs. These findings demonstrate that reinforcement learning offers an effective pathway toward building generalizable and explainable VQA models solely with score-level supervision.

Hyeongseop Rha, Jeong Hun Yeo, Yeonju Kim, Yong Man Ro

The recent advancement of Multimodal Large Language Models (MLLMs) is transforming human-computer interaction (HCI) from surface-level exchanges into more nuanced and emotionally intelligent communication. To realize this shift, emotion understanding becomes essential allowing systems to capture subtle cues underlying user intent. Furthermore, providing faithful explanations for predicted emotions is crucial to ensure interpretability and build user trust. However, current MLLM-based methods often generate emotion explanations that diverge from the ground-truth (GT) labels and sometimes even contradict their own predicted emotions. This inconsistency poses a critical risk for misunderstanding and erodes reliability in interactive settings. To address this, we propose a novel approach: the Emotional Rationale Verifier (ERV) and an Explanation Reward. Our method guides the model to produce reasoning that is explicitly consistent with the GT emotion during multimodal emotion recognition without modifying the model architecture or requiring paired video–description annotations. Our method significantly improves faithful explanation–prediction consistency and explanation emotion accuracy on the MAFW and DFEW datasets. Through extensive experiments and human evaluations, we show that our approach not only enhances alignment between explanation and prediction but also empowers MLLMs to deliver emotionally coherent, trustworthy interactions, marking a key step toward truly human-like HCI systems.

Siran Chen, Boyu Chen, Yuxiao Luo, Chenyun Yu, Yi Ouyang, Lei Cheng, Chengxiang Zhuo, Zang Li, Yali Wang

Large language model (LLM) agents have emerged as a promising solution for enhancing recommendation systems via user simulation. However, existing studies predominantly resort to prompt-based simulation using frozen LLMs, which frequently results in suboptimal item modeling and user preference learning, thereby ultimately constraining recommendation performance. To address these challenges, we introduce VRAgent-R1, a novel agent-based paradigm that incorporates human-like intelligence in user simulation. Specifically, VRAgent-R1 comprises two distinct agents: the Item Perception (IP) Agent and the User Simulation (US) Agent, designed for interactive user-item modeling. Firstly, the IP Agent emulates human-like progressive thinking based on MLLMs, effectively capturing hidden recommendation semantics in videos. With a more comprehensive multimodal content understanding provided by the IP Agent, the video recommendation system is equipped to provide higher-quality candidate items. Subsequently, the US Agent refines the recommended video sets based on in-depth chain-of-thought (CoT) reasoning and achieves better alignment with real user preferences through reinforcement learning. Experimental results on a large-scale video recommendation benchmark MicroLens-100k have demonstrated the effectiveness of our proposed VRAgent-R1 method, e.g., the IP Agent achieves a 6.0% improvement in NDCG@10, while the US Agent shows approximately 45.0% higher accuracy in user decision simulation compared to state-of-the-art baselines.

Nuo Li, Yuan Xiong, Chengliang Liu, Jie Wen, Chao Huang

The increasing prominence of short video platforms has positioned them as a primary channel for public awareness of current events, while also facilitating the widespread dissemination of fake news, thus highlighting the critical need for automated detection technologies. In contrast to fake news confined to text and images, short video news encompasses multiple modalities and extensive information, presenting heightened challenges. Most existing research emphasizes the analysis of news content or user comments alone, while overlooking the crucial role of publishers, leading to poor model performance when handling fake news lacking obvious false signals. Therefore, we propose a Publisher Profiling Module to identify new false signals. To enable a more comprehensive detection of misinformation, we design a Multi-View Aggregation (MVA) model, simultaneously evaluating news from three distinct perspectives: sentiment analysis, content understanding, and publisher profiling. Late fusion is applied at the decision level to leverage the complementary strengths of these perspectives, addressing the limitations of single-view methods. Our experiments conducted on the FakeSV and FVC datasets demonstrate the superior performance of the proposed method.

Zongsheng Cao, Anran Liu, Yangfan He, Jing Li, Bo Zhang, Zigan Wang

Retrieval-augmented generation (RAG) has greatly improved Large Language Models (LLMs) by adding external knowledge. However, current RAG-based methods face difficulties with long-context video understanding due to two main challenges. First, Current RAG-based methods for long-context video understanding struggle to effectively integrate multimodal and long-range temporal information, resulting in fragmented and context-insensitive knowledge representations. Furthermore, their retrieval mechanisms often rely on static textual matching, failing to dynamically align user queries with the most relevant video segments and leading to suboptimal downstream performance. To overcome these issues, we introduce ViG-RAG, a new framework to enhance long-context video understanding through structured textual knowledge grounding and multi-modal retrieval. Specifically, we segment video transcripts into structured units, extract key entities, form temporal connections, and assign confidence for evidence, enabling coherent long-range reasoning. In this way, it utilizes a knowledge-aware grounding mechanism and a context-aware retrieval process that dynamically builds a probabilistic temporal knowledge graph to organize multi-video content. To improve retrieval accuracy, we propose a hybrid retrieval strategy for semantic and temporal features, with an adaptive distribution modeling the relevance. In this way, it achieves the optimal retrieval distribution for each query, enhancing generation efficiency by reducing unnecessary computations. On top of this, ViG-RAG uses a vision-language model to integrate semantic anchors, expanded contextual fields, and selected video frames, generating an accurate response. We evaluate ViG-RAG on several benchmarks, demonstrating that it significantly surpasses current RAG-based methods.

Jiahao Yang, Zihan Wang, Xiangyang Li, Xing Zhu, Yujun Shen, Yinghao Xu, Shuqiang Jiang

Despite significant progress in Vision-Language Navigation (VLN), existing approaches still rely on dense RGB videos that produce excessive patch tokens and lack explicit spatial structure, resulting in substantial computational overhead and limited spatial reasoning. To address these issues, we introduce the Geometry-Aware BEV (GA-BEV) --a compact, 3D-grounded feature representation that integrates both explicit and implicit geometric cues into multimodal large language model (MLLM)-based navigation systems. We construct BEV spatial maps from RGB-D inputs by projecting visual features into 3D space and aggregating them into an agent-centric layout that preserves geometric consistency while reducing token redundancy. To further enrich geometric understanding, we incorporate features from a pretrained 3D foundation model into the BEV space, injecting structural priors learned from large-scale 3D reconstruction tasks. Together, these complementary cues--explicit depth-based projection and implicit learned priors--yield compact yet spatially expressive representations that substantially improve navigation efficiency and performance. Experiments show that our method achieves state-of-the-art results using only navigation data, without DAgger augmentation or mixed VQA training, demonstrating the robustness and data efficiency of the proposed GA-VLN framework.

Zheng Wang, Haoran Chen, Haoxuan Qin, Zhipeng Wei, Tianwen Qian, Cong Bai

Long video understanding is challenging due to dense visual redundancy, long-range temporal dependencies, and the tendency of chain-of-thought and retrieval-based agents to accumulate semantic drift and correlation-driven errors. We argue that long-video reasoning should begin not with reactive retrieval, but with deliberate task formulation: the model must first articulate what must be true in the video for each candidate answer to hold. This thinking-before-finding principle motivates VideoHV-Agent, a framework that reformulates video question answering as a structured hypothesis-verification process.Based on video summaries, a Thinker rewrites answer candidates into testable hypotheses, a Judge derives a discriminative clue specifying what evidence must be checked, a Verifier grounds and tests the clue using localized, fine-grained video content, and an Answer agent integrates validated evidence to produce the final answer.Experiments on three long-video understanding benchmarks show that VideoHV-Agent achieves state-of-the-art accuracy while providing enhanced interpretability, improved logical soundness, and lower computational cost.

Alexandros Stergiou

How do video understanding models acquire their answers? Although current Vision Language Models (VLMs) reason over complex scenes with diverse objects, action performances, and scene dynamics, understanding and controlling their internal processes remains an open challenge. Motivated by recent advancements in text-to-video (T2V) generative models, this paper introduces a logits-to-video (L2V) task alongside a model-independent approach, TRANSPORTER, to generate videos that capture the underlying rules behind VLMs' predictions. Given the high-visual-fidelity produced by T2V models, TRANSPORTER learns an optimal transport coupling to VLM's high-semantic embedding spaces. In turn, logit scores define embedding directions for conditional video generation. TRANSPORTER generates videos that reflect caption changes over diverse object attributes, action adverbs, and scene context. Quantitative and qualitative evaluations across VLMs demonstrate that L2V can provide a fidelity-rich, novel direction for model interpretability that has not been previously explored.

Ying Shen, Jerry Xiong, Tianjiao Yu, Ismini Lourentzou

Recent advances in generative video modeling, driven by large-scale datasets and powerful architectures, have yielded remarkable visual realism. However, emerging evidence suggests that simply scaling data and model size does not endow these systems with an understanding of the underlying physical laws that govern real-world dynamics. Existing approaches often fail to capture or enforce such physical consistency, resulting in unrealistic motion and dynamics.In this work, we investigate whether integrating the inference of latent physical properties directly into the video generation process can equip models with the ability to produce physically plausible videos. To this end, we propose PHANTOM, a Physics-Infused Video Generation model that jointly models the visual content and latent physical dynamics. Conditioned on observed video frames and inferred physical states, PHANTOM jointly predicts latent physical dynamics and generates future video frames.PHANTOM leverages a physics-aware video representation that serves as an abstract yet informative embedding of the underlying physics, facilitating the joint prediction of physical dynamics alongside video content without requiring an explicit specification of a complex set of physical dynamics and properties. By integrating the inference of physical-aware video representation directly into the video generation process, PHANTOM produces video sequences that are both visually realistic and physically consistent.Quantitative and qualitative results on both standard video generation and physics-aware benchmarks demonstrate that PHANTOM not only outperforms existing methods in terms of adherence to physical dynamics but also delivers competitive perceptual fidelity.

Apoorv Vyas, Heng-Jui Chang, Cheng-Fu Yang, Po-Yao Huang, Luya Gao, Julius Richter, Sanyuan Chen, Matthew Le, Piotr Dollár, Christoph Feichtenhofer 等

We introduce Perception Encoder-Audiovisual, PE-AV, a new family of encoders for audio and video understanding trained with scaled contrastive learning. Building on PE, PE-AV makes several key contributions to extend representations to audio, and natively support joint embeddings across audio-video, audio-text, and video-text modalities. PE-AV's unified cross-modal embeddings enable novel tasks such as speech retrieval, and set a new state of the art across standard audio and video benchmarks. We unlock this by building a strong audiovisual data engine that synthesizes high-quality captions for O(100M) audio-video pairs, enabling large-scale supervision consistent across modalities. Our audio data includes speech, music, and general sound effects--avoiding single-domain limitations common in prior work. We exploit ten pairwise contrastive objectives, showing that scaling cross-modality and caption-type pairs strengthens alignment and improves zero-shot performance. Models and code are available.

Mohamad Alansari, Naufal Suryanto, Divya Velayudhan, Sajid Javed, Naoufel Werghi, Muzammal Naseer

Multimodal large language models (MLLMs) have advanced from image-level reasoning to pixel-level grounding, but extending these capabilities to videos remains challenging as models must achieve spatial precision and temporally consistent reference tracking. Existing video MLLMs often rely on a static segmentation token ([SEG]) for frame-wise grounding, which provides semantics but lacks temporal context, causing spatial drift, identity switches, and unstable initialization when objects move or reappear. We introduce SPARROW, a pixel-grounded video MLLM that unifies spatial accuracy and temporal stability through two key components: (i) Target-Specific Tracked Features (TSF), which inject temporally aligned referent cues during training, and (ii) a dual-prompt design that decodes box ([BOX]) and segmentation ([SEG]) tokens to fuse geometric priors with semantic grounding. SPARROW is supported by a curated referential video dataset of 30,646 videos and 45,231 Q&A pairs and operates end-to-end without external detectors via a class-agnostic SAM2-based proposer. Integrated into three recent open-source video MLLMs (UniPixel, GLUS, and VideoGLaMM), SPARROW delivers consistent gains across six benchmarks, improving up to +8.9 J&F on RVOS, +5 mIoU on visual grounding, and +5.4 CLAIR on GCG. These results demonstrate that SPARROW substantially improves referential stability, spatial precision, and temporal coherence in pixel-grounded video understanding. Project page: https://risys-lab.github.io/SPARROW/

Zefeng He, Xiaoye Qu, Yafu Li, Siyuan Huang, Daizong Liu, Yu Cheng

Reinforcement Learning with Verifiable Reward (RLVR) has substantially advanced the video understanding capabilities of Multimodal Large Language Models (MLLMs). However, the rapid progress of MLLMs is outpacing the complexity of existing video datasets, while the manual annotation of new, high-quality data remains prohibitively expensive.This work investigates a pivotal question: Can the rich, intrinsic information within videos be harnessed to self-generate high-quality, verifiable training data?To investigate this problem, we first introduce three self-supervised pretext tasks for video understanding: Anomaly Grounding, Object Counting, and Temporal Jigsaw. To validate the difficulty of these tasks, we construct the Video Intrinsic Understanding Benchmark (VIUBench), revealing that current state-of-the-art MLLMs struggle significantly on these tasks. Building upon these pretext tasks, we develop the VideoSSR-30K dataset and propose VideoSSR, a novel video self-supervised reinforcement learning framework for RLVR. Extensive experiments across 17 benchmarks, spanning four major video domains (General Video QA, Long Video QA, Temporal Grounding, and Complex Reasoning), demonstrate that our VideoSSR consistently enhances model performance, yielding an average improvement of over 5%. These results establish VideoSSR as a potent foundational framework for developing more advanced video understanding in MLLMs.

Yura Choi, Roy Miles, Rolandos Alexandros Potamias, Ismail Elezi, Jiankang Deng, Stefanos Zafeiriou

Understanding and answering questions based on a user's pointing gesture is essential for next-generation egocentric AI assistants. However, current Multimodal Large Language Models (MLLMs) struggle with such tasks due to the lack of gesture-rich data and their limited ability to infer fine-grained pointing intent from egocentric video.To address this, we introduce EgoPointVQA, a dataset and benchmark for gesture-grounded egocentric question answering, comprising 4000 synthetic and 400 real-world videos across multiple deictic reasoning tasks.Built upon it, we further propose Hand Intent Tokens (HINT), which encode tokens derived from 3D hand keypoints using an off-the-shelf reconstruction model and interleaves them with the model input to provide explicit spatial and temporal context for interpreting pointing intent.We show that our model outperforms others in different backbones and model sizes.In particular, HINT-14Bachieves 68.1% accuracy, on average over 6 tasks, surpassing the state-of-the-art, InternVL3-14B, by 6.6%.To further facilitate the open research, we will release the code, model, and dataset.

Boyu Chen, Zikang Wang, Zhengrong Yue, Kainan Yan, Chenyun Yu, Yi Huang, Zijun Liu, Yafei Wen, Xiaoxin Chen, Yang Liu 等

Most of the multi-agent video understanding frameworks adopt static and non-learnable tool invocation mechanisms, which limit the discovery of diverse clues essential for robust perception and reasoning regarding temporally or spatially complex videos. To address this challenge, we propose a novel Multi-agent system for video understanding, namely VideoChat-M1. Instead of using a single or fixed policy, we adopt a distinct Collaborative Policy Planning (CPP) paradigm with multiple policy agents, which comprises three key processes. (1)Policy Generation: Each agent generates its unique tool invocation policy tailored to the user's query. (2) Policy Execution: Each agent sequentially invokes relevant tools to execute its policy and explore the video content. (3) Policy Communication: During the intermediate stages of policy execution, agents interact with one another to update their respective policies. Through this collaborative framework, all agents work in tandem, dynamically refining their preferred policies based on contextual insights from peers. Moreover, we equip our CPP paradigm with Multi-Agent Reinforcement Learning (MARL). Consequently, policy agents can be jointly optimized to enhance the performance, guided by both the final answer reward and intermediate collaborative process feedback. Extensive experiments demonstrate that VideoChat-M1 achieves SOTA performance across eight benchmarks on four tasks. Notably, on LongVideoBench, our method outperforms Gemini 2.5 pro by 3.6% and GPT-4o by 15.6%.

Jiaxu Wan, Xu Wang, Mengwei Xie, Hang Zhang, Mu Xu, Yang Han, Ding Yuan, Hong Zhang, Yifan Yang

Video-based spatial reasoning -- such as estimating distances, judging directions, or understanding layouts from multiple views -- requires selecting informative frames and, when needed, actively seeking additional viewpoints during inference. Existing multimodal large language models (MLLMs) consume a fixed set of uniformly sampled frames and cannot request new views once reasoning begins, often missing the geometric cues necessary for reliable spatial judgments. We present EagleVision, a dual-stage framework that combines geometry-aware frame selection with active, Bird's-Eye-View (BEV)-grounded reasoning. In the first stage (macro perception), a semantics-perspective-fusion determinantal point process (SPF-DPP) selects a compact set of keyframes that jointly maximize semantic relevance and viewpoint diversity under a fixed token budget. In the second stage (micro verification), the model performs iterative spatial Chain-of-Thought: at each step it can either reason in text or predict a pose on the BEV plane to retrieve the nearest real frame, forming a closed-loop hypothesize-look-verify cycle. The querying policy is trained purely via reinforcement learning with a spatial grounding reward, requiring no human-annotated reasoning traces. On VSI-Bench and SQA3D, EagleVision achieves state-of-the-art performance among open-source vision-language models.

Yiweng Xie, Bo He, Junke Wang, Xiangyu Zheng, Ziyi Ye, Zuxuan Wu

This paper presents FluxMem, a training-free framework for efficient streaming video understanding. FluxMem adaptively compresses redundant visual memory through a hierarchical, two-stage design: (1) a Temporal Adjacency Selection (TAS) module removes redundant visual tokens across adjacent frames, and (2) a Spatial Domain Consolidation (SDC) module further merges spatially repetitive regions within each frame into compact representations. To adapt effectively to dynamic scenes, we introduce a self-adaptive token compression mechanism in both TAS and SDC, which automatically determines the compression rate based on intrinsic scene statistics rather than manual tuning. Extensive experiments demonstrate that FluxMem achieves new state-of-the-art results on existing online video benchmarks, reaching 76.4 on StreamingBench and 67.2 on OVO-Bench under real-time settings, while reducing latency by 69.9% and peak GPU memory by 34.5% on OVO-Bench. Furthermore, it maintains strong offline performance, achieving 73.1 on MLVU while using 65% fewer visual tokens.

Jialiang Zhang, Junlong Tong, Junyan Lin, Hao Wu, Yirong Sun, Yunpu Ma, Xiaoyu Shen

Large Vision Language Models (LVLMs) have demonstrated remarkable capabilities in Chain-of-Thought (CoT) reasoning. However, existing LVLM reasoning paradigms only begin reasoning after the entire video becomes available, introducing unnecessary latency and diminishing attention to early visual cues in dynamic scenes. Inspired by the human ability to think while watching, we introduce a streaming reasoning paradigm for LVLMs, where reasoning unfolds sequentially with incoming frames and deepens after the full video is observed. We instantiate this paradigm through Think-as-You-See (TaYS), a unified framework that enables LVLMs to reason while watching by integrating streaming CoT generation, stream-constrained training, and stream-parallel inference. Specifically, TaYS employs temporally aligned streaming reasoning units with precise CoT supervision, enforces ordered reasoning via streaming attention masks and positional encodings, and utilizes a parallel KV caches mechanism that decouples input encoding from reasoning generation, ensuring alignment and true concurrency. We evaluate TaYS on the Qwen2.5-VL model family across representative video CoT tasks, including event dynamics analysis, causal reasoning, and thematic understanding. Experimental results show that TaYS achieves superior reasoning performance compared with batch-mode CoT, while reducing pre-reasoning latency to under one second and overall answer delay by more than 50%. These findings demonstrate the effectiveness of the streaming paradigm in enabling real-time, human-like reasoning for LVLMs.

Arsha Nagrani, Jasper Uijlings, Shyamal Buch, Tobias Weyand, Sudheendra Vijayanarasimhan, Bo Hu, Ramin Mehran, David A Ross, Cordelia Schmid

Video reasoning models are a core component of egocentric and embodied agents. However, standard benchmarks for assessing models provide only evaluation of the output (e.g. the answer to a question), without evaluation of inter- mediate reasoning steps, and most provide answers only in the text domain. We introduce Minerva-Ego, a bench- mark for evaluating complex egocentric visual reasoning. We extend recent high-quality video data sources recorded from egocentric / embodied settings with a set of challenging, multi-step multimodal questions and spatiotemporally-dense human-annotated reasoning traces. Benchmarking experiments show that state-of-the-art models still have a large gap to human performance. To investigate this gap in detail, we annotate each reasoning trace in the dataset with the objects of interest required to solve the question, as spatio-temporal mask annotations. Through extensive evaluations, we identify that prompting frontier models with hints of 'where' and when to look yields substantial improvements in performance.

Zhiyu Xu, Weilong Yan, Yufei Shi, Xin Meng, Tao He, Huiping Zhuang, Ming Li, Hehe Fan

Recent advancements in multimodal large language models (MLLMs) and video agent systems have significantly improved general video understanding. However, when applied to scientific video understanding and educating--a domain that demands external professional knowledge integration and rigorous step-wise reasoning--existing approaches often struggle. To bridge this gap, we propose SciEducator, an iterative self-evolving multi-agent system for scientific video comprehension and education. Rooted in the classical Deming Cycle from management science, our design reformulates its Plan-Do-Study-Act philosophy into a self-evolving reasoning and feedback mechanism, which facilitates the interpretation of intricate scientific activities in videos. Moreover, SciEducator can produce multimodal educational content tailored to specific scientific processes, including textual instructions, visual guides, audio narrations, and interactive references. To support evaluation, we construct SciVBench, a benchmark consisting of 500 expert-verified and literature-grounded science QA pairs across five categories, covering physical, chemical, and everyday phenomena. Extensive experiments demonstrate that SciEducator substantially outperforms leading closed-source MLLMs (e.g., Gemini, GPT-4o) and state-of-the-art video agents on the benchmark, establishing a new paradigm for the community.

Yuzhi Huang, Kairun Wen, Rongxin Gao, Dongxuan Liu, Yibin Lou, Jie Wu, Jing Xu, Jian Zhang, Zheng Yang, Yunlong Lin 等

Humans inhabit a physical 4D world, where spatial geometry and semantic content evolve over time, forming a dynamic reality. While current Multimodal Large Language Models (MLLMs) demonstrate strong capabilities in understanding static visual inputs, it remains unclear whether they can effectively "think in dynamics," i.e., perceive, track, and reason about spatio-temporal evolution in complex scenes.To systematically evaluate these abilities, we introduce \texttt Dyn-Bench , a large-scale benchmark designed to assess spatio-temporal reasoning and localized dynamics perception. Constructed through multi-stage filtering over massive 2D and 4D data sources, \texttt Dyn-Bench provides a high-quality collection of diverse dynamic scenes, consisting of 1k videos, 7k visual question answering (VQA) pairs, and 3k dynamic object grounding samples.We comprehensively study general-purpose, spatial-aware, and region-level MLLMs to understand how they "think in dynamics" from both linguistic and visual perspectives. Our results reveal that existing models struggle to jointly excel in both spatio-temporal reasoning and dynamic object grounding, often producing inconsistent interpretations of motion and interaction. Conventional prompting strategies i.e., chain-of-thought or caption-based hints) provide only limited improvements.In contrast, structured integration approaches, including Mask-Guided Fusion and the Spatio-Temporal Textual Cognitive Map (ST-TCM), substantially enhance MLLMs' dynamic perception and spatio-temporal reasoning in an evolving 4D world. These findings underscore the importance of explicit spatio-temporal structural cues to bridge the gap between static perception and dynamic reasoning in MLLMs.