Searcharxiv⌕ Search

arXiv subjects

Jialin Guo

Publications and source records attributed to Jialin Guo.

8 recordsLinked to original sources

FrameRouter: Frame Budget Routing for Long-Video Understanding on Video-MME-v2 and Beyond

Long-video question answering benchmarks like Video-MME-v2 typically evaluate systems under a uniform frame budget for every question, despite stark variation in the evidence each query actually demands. We analyze this mismatch on Video-MME-v2 and find that different question subsets prefer different frozen frame-budget policies, while uniformly increasing the frame count does not reliably improve all metrics. These findings motivate fixed-mean frame-budget routing, a controlled inference setting in which each question may receive a different frame budget while the average budget matches a uniform reference. We introduce FrameRouter, a training-free test-time framework that combines an evidence-demand router, a budget-constrained allocation rule, and a frozen bank of frame-sampling policies. By controlling how much visual evidence each question receives while leaving the underlying samplers unchanged, FrameRouter is orthogonal to query-aware frame selection. Across Video-MME-v2 and several long-video understanding benchmarks, FrameRouter improves long-video question answering by using the same average visual budget more selectively across questions. Code is available at https://github.com/bingjunluo/FrameRouter.

cs.AI↗

VideoEvolve: Evolving Agent Harnesses for Video Temporal Grounding

Video temporal grounding aims to localize events in videos from natural-language queries. For agents built around frozen video-language models, the harness determines how queries guide temporal predictions and how those predictions are refined. Manually refining these harnesses requires diagnosing grounding failures and coordinating changes to both agent workflows and instructions. We introduce VideoEvolve, a framework that automatically evolves agent harnesses for video temporal grounding. VideoEvolve uses a Cloze-Structured Harness Representation that preserves stage interfaces while leaving agent workflows and instructions open to evolution. Branch-Guided Harness Evolution preserves promising code branches for continued refinement, using execution feedback to guide local edits and validation to determine which improvements are carried forward. Experiments demonstrate improved grounding performance across multiple benchmarks. Component analyses identify instruction refinement as a consistent source of gains, while the benefits of evolved code vary across evaluation settings. Together, these results support automated harness evolution as an effective approach to improving video temporal grounding. Code is available at https://github.com/bingjunluo/VideoEvolve .

cs.AI↗

Video-RSI: Recursive Self-Improvement of Video Understanding Agents via Harness Evolution

Video understanding agents acquire evidence through an executable harness that controls what they observe and how they use those observations. However, execution traces contain only the evidence acquired by the current harness, leaving competing explanations for failure unresolved and limiting the basis for self-improvement. We introduce Video-RSI, a framework for recursive self-improvement in which a video understanding agent uses its own language model to revise its harness. Through active video investigation, the model revisits the original training videos to test competing failure explanations with additional observations, grounding proposed changes in evidence beyond the existing trace. Cost-aware harness evolution turns these diagnoses into reusable revisions and determines which revisions to retain by considering both answer accuracy and visual cost. Across our evaluation settings on video understanding benchmarks, the evolved agent improves accuracy while processing fewer frames and achieves competitive accuracy-efficiency trade-offs against existing video understanding agents. These results demonstrate the potential for video understanding agents to improve their own evidence acquisition and use through harness evolution. Code is available at https://github.com/bingjunluo/Video-RSI .

cs.AI↗

UnifiedAttack: Evaluating the Safety of Large Multimodal Models in Synergistic Harmful Image-Text Generation

As Large Multimodal Models (LMMs) transition toward natively unified architectures, evaluating their safety in synergistic harmful image-text generation tasks becomes a critical challenge. Unlike unimodal threats, synergistic risks emerge when text and image modalities are coordinated to produce harm that significantly exceeds their individual components. We introduce UnifiedAttack, a novel benchmark designed to evaluate LMM safety in collaborative scenarios by focusing on the harmfulness gain achieved through cross-modal synergy. The benchmark incorporates samples filtered for their multimodal potential alongside a novel subset of synthesized disinformation queries. To verify identified vulnerabilities, we propose a synergistic hijacking framework featuring In-Context Reskinning (ICR) and Cognitive Planning Injection (CPI). ICR utilizes few-shot learning to wrap adversarial intent in benign virtual shells to desensitize safety filters, while CPI hijacks the reasoning path by enforcing a plan-then-execute paradigm. By compelling the system to commit to a neutral logical plan, we exploit its internal drive for consistency to induce the synchronized generation of harmful multimodal content. Extensive evaluations on state-of-the-art architectures demonstrate that UnifiedAttack consistently bypasses modern alignment. Our findings reveal that the structural helpfulness and logical coherence of unified models can be systematically weaponized, highlighting the urgent need for logic-aware defenses in synergistic generation tasks. Code is available at https://github.com/bingjunluo/UnifiedAttack .

cs.CR↗

Temporal State Transport in Video Generation: Diagnosing and Correcting Spectral Imbalance

Reliable video generation requires more than high-quality frames to form a coherent story: a model must maintain a persistent state, transporting visual attributes such as identity, scene layout, motion, and fine details across time. Existing training-free methods mainly strengthen cross-frame attention or analyze local attention entropy, but these views do not reveal whether temporal interactions stay in a healthy transport regime. In this work, we study video generation through the perspective of Temporal State Transport. We introduce Spectral Tension, a signed diagnostic that compares local attention diffuseness with global spectral diversity, and use it to identify two opposite temporal failures: fragmented transport and over-mixing hotspots. Based on this diagnosis, we propose Spectral Transport Homeostasis, a training-free regulator that softly corrects pathological temporal states while largely preserving balanced ones. Experiments on pretrained video generation models show that the original model often occupies imbalanced temporal regimes, whereas our method selectively applies larger corrections to the worst temporal hotspots and improves temporal consistency and visual quality without finetuning. Code: https://github.com/lytang63/temporal-state-transport

cs.CV↗

Adversarial Style Optimization: Enhancing VLM Jailbreaks by GRPO-based Stylistic Triggers Optimization

Multimodal Large Language Models (MLLMs) have achieved impressive performance, but their safety alignment remains vulnerable to jailbreak attacks. Existing content-based jailbreaks are often inconsistent and show unsatisfying performance against the rapidly evolving MLLMs, failing to exploit non-content-based vulnerabilities. Unlike previous research, we empirically find that MLLMs exhibit a Stylistic Inconsistency between their comprehension ability and safety ability: MLLMs can robustly understand content regardless of visual style, yet their defense mechanisms can be easily bypassed by specific stylistic triggers. Based on this finding, we propose Adversarial Style Optimization (ASO), a plug-and-play enhancement module to amplify existing visual jailbreaks. ASO fine-tunes an image-editing model to superimpose an optimized stylistic modification onto a given adversarial image, using a Group Relative Policy Optimization (GRPO) agent guided by a Structurally-Tiered Reward Function that combines a logit-based signal for detecting explicit refusals with a high-fidelity semantic evaluation from a powerful judge model. Extensive experiments show that ASO significantly enhances the ASR of SOTA attacks, demonstrating that stylistic biases are a scalable vector for red-teaming MLLMs. Our code is available at https://github.com/bingjunluo/ASO.

cs.CL↗

Gender Bias in YouTube Exposure: Allocative and Structural Inequalities in Political Information Environments

Recommendation algorithms have become the dominant mechanism for information distribution on digital platforms, profoundly shaping personalized information consumption environments. However, gender bias, as a significant form of algorithmic discrimination, may cause users to experience unequal exposure within different political information environments. Taking YouTube as a case, we conduct a controlled social-bot field experiment, where male-coded and female-coded profiles are constructed. We track the exposure and click patterns of these bots to analyze their recommendation trajectories. We analyze the distribution of recommended content from two dimensions: allocative bias and structural bias. First, we find statistically significant differences in allocative bias across male-coded and female-coded profiles, particularly in terms of issue distribution, ideological orientation, and political entities. Secondly, we observe structural bias in the political information environments, characterized by distinct clustering patterns. Additionally, time-series analysis shows that exposure pathways continue to be shaped over time by both communities detected in the co-occurrence network and individual profile-level dynamics. Finally, we construct a simple collaborative-filtering model that reproduces the observed gender bias. We argue that gender bias in recommendation systems is reflected not only in the allocation of political content, but also in how community structures shape these environments, reinforcing societal inequalities and highlighting the need for algorithmic fairness.

cs.SI↗

Dual-Branch Temperature Scaling Calibration for Long-Tailed Recognition

The calibration for deep neural networks is currently receiving widespread attention and research. Miscalibration usually leads to overconfidence of the model. While, under the condition of long-tailed distribution of data, the problem of miscalibration is more prominent due to the different confidence levels of samples in minority and majority categories, and it will result in more serious overconfidence. To address this problem, some current research have designed diverse temperature coefficients for different categories based on temperature scaling (TS) method. However, in the case of rare samples in minority classes, the temperature coefficient is not generalizable, and there is a large difference between the temperature coefficients of the training set and the validation set. To solve this challenge, this paper proposes a dual-branch temperature scaling calibration model (Dual-TS), which considers the diversities in temperature parameters of different categories and the non-generalizability of temperature parameters for rare samples in minority classes simultaneously. Moreover, we noticed that the traditional calibration evaluation metric, Excepted Calibration Error (ECE), gives a higher weight to low-confidence samples in the minority classes, which leads to inaccurate evaluation of model calibration. Therefore, we also propose Equal Sample Bin Excepted Calibration Error (Esbin-ECE) as a new calibration evaluation metric. Through experiments, we demonstrate that our model yields state-of-the-art in both traditional ECE and Esbin-ECE metrics.

cs.LG↗