arXiv subjects

Yutong Zhang

Publications and source records attributed to Yutong Zhang.

At least 73 records · Page 4Linked to original sources

arXiv2026

MP-ISMoE: Mixed-Precision Interactive Side Mixture-of-Experts for Efficient Transfer Learning

Parameter-efficient transfer learning (PETL) has emerged as a pivotal paradigm for adapting pre-trained foundation models to downstream tasks, significantly reducing trainable parameters yet suffering from substantial memory overhead caused by gradient backpropagation during fine-tuning. While memory-efficient transfer learning (METL) circumvents this challenge by bypassing backbone gradient computation via lightweight small side networks, its stringent memory constraint severely limits learning capacity of side networks, thereby significantly compromising performance. To address these limitations, we propose a novel Mixed-Precision Interactive Side Mixture-of-Experts framework (MP-ISMoE). Specifically, we first propose a Gaussian Noise Perturbed Iterative Quantization (GNP-IQ) scheme to quantize weights into lower-bits while effectively decreasing quantization errors. By leveraging memory conserved from GNP-IQ, we subsequently employ Interactive Side Mixture-of-Experts (ISMoE) to scaling up side networks without sacrificing overall memory efficiency. Different from conventional mixture-of-experts, ISMoE learns to select optimal experts by interacting with salient features from frozen backbones, thus suppressing knowledge forgetting and boosting performance. Extensive experiments across diverse vision-language and language-only tasks demonstrate that MP-ISMoE remarkably promotes accuracy compared to state-of-the-art METL approaches, while maintaining comparable parameter and memory efficiency.

cs.LG↗

arXiv2026

Multimodal Structure Learning: Disentangling Shared and Specific Topology via Cross-Modal Graphical Lasso

Learning interpretable multimodal representations inherently relies on uncovering the conditional dependencies between heterogeneous features. However, sparse graph estimation techniques, such as Graphical Lasso (GLasso), to visual-linguistic domains is severely bottlenecked by high-dimensional noise, modality misalignment, and the confounding of shared versus category-specific topologies. In this paper, we propose Cross-Modal Graphical Lasso (CM-GLasso) that overcomes these fundamental limitations. By coupling a novel text-visualization strategy with a unified vision-language encoder, we strictly align multimodal features into a shared latent space. We introduce a cross-attention distillation mechanism that condenses high-dimensional patches into explicit semantic nodes, naturally extracting spatial-aware cross-modal priors. Furthermore, we unify tailored GLasso estimation and Common-Specific Structure Learning (CSSL) into a joint objective optimized via the Alternating Direction Method of Multiplier (ADMM). This formulation guarantees the simultaneous disentanglement of invariant and class-specific precision matrices without multi-step error accumulation. Extensive experiments across eight benchmarks covering both natural and medical domains demonstrate that CM-GLasso establishes a new state-of-the-art in generative classification and dense semantic segmentation tasks.

cs.CV↗

arXiv2026

SAATT Nav: a Socially Aware Autonomous Transparent Transportation Navigation Framework for Wheelchairs

While powered wheelchairs reduce physical fatigue as opposed to manual wheelchairs for individuals with mobility impairment, they demand high cognitive workload due to information processing, decision making and motor coordination. Current autonomous systems lack social awareness in navigation and transparency in decision-making, leading to decreased perceived safety and trust from the user and others in context. This work proposes Socially Aware Autonomous Transparent Transportation (SAATT) Navigation framework for wheelchairs as a potential solution. By implementing a Large Language Model (LLM) informed of user intent and capable of predicting other peoples' intent as a decision-maker for its local controller, it is able to detect and navigate social situations, such as passing pedestrians or a pair conversing. Furthermore, the LLM textually communicates its reasoning at each waypoint for transparency. In this experiment, it is compared against a standard global planner, a representative competing social navigation model, and an Ablation study in three simulated environments varied by social levels in eight metrics categorized under Safety, Social Compliance, Efficiency, and Comfort. Overall, SAATT Nav outperforms in most social situations and equivalently or only slightly worse in the remaining metrics, demonstrating the potential of a socially aware and transparent autonomous navigation system to assist wheelchair users.

cs.RO↗

arXiv2026

MARTI-MARS$^2$: Scaling Multi-Agent Self-Search via Reinforcement Learning for Code Generation

While the complex reasoning capability of Large Language Models (LLMs) has attracted significant attention, single-agent systems often encounter inherent performance ceilings in complex tasks such as code generation. Multi-agent collaboration offers a promising avenue to transcend these boundaries. However, existing frameworks typically rely on prompt-based test-time interactions or multi-role configurations trained with homogeneous parameters, limiting error correction capabilities and strategic diversity. In this paper, we propose a Multi-Agent Reinforced Training and Inference Framework with Self-Search Scaling (MARTI-MARS2), which integrates policy learning with multi-agent tree search by formulating the multi-agent collaborative exploration process as a dynamic and learnable environment. By allowing agents to iteratively explore and refine within the environment, the framework facilitates evolution from parameter-sharing homogeneous multi-role training to heterogeneous multi-agent training, breaking through single-agent capability limits. We also introduce an efficient inference strategy MARTI-MARS2-T+ to fully exploit the scaling potential of multi-agent collaboration at test time. We conduct extensive experiments across varied model scales (8B, 14B, and 32B) on challenging code generation benchmarks. Utilizing two collaborating 32B models, MARTI-MARS2 achieves 77.7%, outperforming strong baselines like GPT-5.1. Furthermore, MARTI-MARS2 reveals a novel scaling law: shifting from single-agent to homogeneous multi-role and ultimately to heterogeneous multi-agent paradigms progressively yields higher RL performance ceilings, robust TTS capabilities, and greater policy diversity, suggesting that policy diversity is critical for scaling intelligence via multi-agent reinforcement learning.

cs.LG↗

arXiv2026

Nested search

I introduce and study a nested search problem modeled as a tree structure that generalizes Weitzman (1979) in two ways: (1) search progresses incrementally, reflecting real-life scenarios where agents gradually acquire information about the prizes; and (2) the realization of prizes can be correlated, capturing similarities among them. I derive the optimal policy, which takes the form of an index solution. I apply this result to study monopolistic competition in a market with two stages of product inspection. My application illustrates that regulations on drip pricing lower equilibrium price and raise consumer surplus.

econ.TH↗

arXiv2026

Entropy-Adaptive Fine-Tuning: Resolving Confident Conflicts to Mitigate Forgetting

Supervised Fine-Tuning (SFT) is the standard paradigm for domain adaptation, yet it frequently incurs the cost of catastrophic forgetting. In sharp contrast, on-policy Reinforcement Learning (RL) effectively preserves general capabilities. We investigate this discrepancy and identify a fundamental distributional gap: while RL aligns with the model's internal belief, SFT forces the model to fit external supervision. This mismatch often manifests as "Confident Conflicts" tokens characterized by low probability but low entropy. In these instances, the model is highly confident in its own prediction but is forced to learn a divergent ground truth, triggering destructive gradient updates. To address this, we propose Entropy-Adaptive Fine-Tuning (EAFT). Unlike methods relying solely on prediction probability, EAFT utilizes token-level entropy as a gating mechanism to distinguish between epistemic uncertainty and knowledge conflict. This allows the model to learn from uncertain samples while suppressing gradients on conflicting data. Extensive experiments on Qwen and GLM series (ranging from 4B to 32B parameters) across mathematical, medical, and agentic domains confirm our hypothesis. EAFT consistently matches the downstream performance of standard SFT while significantly mitigating the degradation of general capabilities.

cs.LG↗

arXiv2025

Robust Tensor Principal Component Analysis: Exact Recovery via Deterministic Model

Tensor, also known as multi-dimensional array, arises from many applications in signal processing, manufacturing processes, healthcare, among others. As one of the most popular methods in tensor literature, Robust tensor principal component analysis (RTPCA) is a very effective tool to extract the low rank and sparse components in tensors. In this paper, a new method to analyze RTPCA is proposed based on the recently developed tensor-tensor product and tensor singular value decomposition (t-SVD). Specifically, it aims to solve a convex optimization problem whose objective function is a weighted combination of the tensor nuclear norm and the l1-norm. In most of literature of RTPCA, the exact recovery is built on the tensor incoherence conditions and the assumption of a uniform model on the sparse support. Unlike this conventional way, in this paper, without any assumption of randomness, the exact recovery can be achieved in a completely deterministic fashion by characterizing the tensor rank-sparsity incoherence, which is an uncertainty principle between the low-rank tensor spaces and the pattern of sparse tensor.

stat.ML↗

arXiv2025

Competition for being visited first and ordered search deterrence

When customers must visit a seller to learn the valuation of its product, sellers potentially benefit from charging a lower price on the first visit and a higher price when a buyer returns. Armstrong and Zhou (2016) show that such price discrimination can arise in equilibrium when buyers learn a seller's pricing policy only upon visiting. We depart from this assumption by supposing that sellers commit to observable pricing policies that guide consumer search and buyers can choose whom to visit first. We show that no seller engages in price discrimination in equilibrium.

econ.TH↗

arXiv2025

Evaluating Large Language Models for Radiology Natural Language Processing

The rise of large language models (LLMs) has marked a pivotal shift in the field of natural language processing (NLP). LLMs have revolutionized a multitude of domains, and they have made a significant impact in the medical field. Large language models are now more abundant than ever, and many of these models exhibit bilingual capabilities, proficient in both English and Chinese. However, a comprehensive evaluation of these models remains to be conducted. This lack of assessment is especially apparent within the context of radiology NLP. This study seeks to bridge this gap by critically evaluating thirty two LLMs in interpreting radiology reports, a crucial component of radiology NLP. Specifically, the ability to derive impressions from radiologic findings is assessed. The outcomes of this evaluation provide key insights into the performance, strengths, and weaknesses of these LLMs, informing their practical applications within the medical domain.

cs.CL↗

arXiv2025

C-RAN Advanced: From a Network Cooperation Perspective

Future mobile networks in the sixth generation (6G) are poised for a paradigm shift from conventional communication services toward comprehensive information services, driving the evolution of radio access network (RAN) architectures toward enhanced cooperation, intelligence, and service orientation. Building upon the concept of centralized, collaborative, cloud, and clean RAN (C-RAN), this article proposes a novel cooperative, intelligent, and service-based RAN (CIS-RAN) architecture. Focusing on cooperation, CIS-RAN extends the traditional cooperative communication paradigm by further integrating cooperative sensing and cooperative artificial intelligence (AI). To improve both performance and effectiveness across diverse application scenarios, CIS-RAN enhances network cooperation throughout the entire process of acquisition, transmission, and processing, thereby enabling efficient information acquisition, diverse cooperative interactions, and intelligent fusion decision-making. Key technologies are discussed, with network cooperative multiple-input multiple-output (MIMO) examined as a case study, demonstrating superior performance over traditional architectures, as demonstrated by numerical results. Future research directions are outlined, emphasizing the continued exploration and advancement of the CIS-RAN architecture, particularly in enhancing network cooperation.

eess.SP↗

arXiv2025

FairReason: Balancing Reasoning and Social Bias in MLLMs

Multimodal Large Language Models (MLLMs) already achieve state-of-the-art results across a wide range of tasks and modalities. To push their reasoning ability further, recent studies explore advanced prompting schemes and post-training fine-tuning. Although these techniques improve logical accuracy, they frequently leave the models' outputs burdened with pronounced social biases. Clarifying how reasoning gains interact with bias mitigation-and whether the two objectives inherently trade off-therefore remains an open and pressing research problem. Our study begins by benchmarking three bias-mitigation strategies-supervised fine-uning (SFT), knowledge distillation (KD), and rule-based reinforcement learning (RL)-under identical conditions, establishing their baseline strengths and weaknesses. Building on these results, we vary the proportion of debias-focused and reasoning-centric samples within each paradigm to chart the reasoning-versus-bias trade-off. Our sweeps reveal a consistent sweet spot: a roughly 1:4 mix trained with reinforcement learning cuts stereotype scores by 10% while retaining 88% of the model's original reasoning accuracy, offering concrete guidance for balancing fairness and capability in MLLMs.

cs.AI↗

arXiv2025

Evo-MARL: Co-Evolutionary Multi-Agent Reinforcement Learning for Internalized Safety

Multi-agent systems (MAS) built on multimodal large language models exhibit strong collaboration and performance. However, their growing openness and interaction complexity pose serious risks, notably jailbreak and adversarial attacks. Existing defenses typically rely on external guard modules, such as dedicated safety agents, to handle unsafe behaviors. Unfortunately, this paradigm faces two challenges: (1) standalone agents offer limited protection, and (2) their independence leads to single-point failure-if compromised, system-wide safety collapses. Naively increasing the number of guard agents further raises cost and complexity. To address these challenges, we propose Evo-MARL, a novel multi-agent reinforcement learning (MARL) framework that enables all task agents to jointly acquire defensive capabilities. Rather than relying on external safety modules, Evo-MARL trains each agent to simultaneously perform its primary function and resist adversarial threats, ensuring robustness without increasing system overhead or single-node failure. Furthermore, Evo-MARL integrates evolutionary search with parameter-sharing reinforcement learning to co-evolve attackers and defenders. This adversarial training paradigm internalizes safety mechanisms and continually enhances MAS performance under co-evolving threats. Experiments show that Evo-MARL reduces attack success rates by up to 22% while boosting accuracy by up to 5% on reasoning tasks-demonstrating that safety and utility can be jointly improved.

cs.AI↗

arXiv2025

Burst: Collaborative Curation in Connected Social Media Communities

Positive social interactions can occur in groups of many shapes and sizes, spanning from small and private to large and open. However, social media tends to binarize our experiences into either isolated small groups or into large public squares. In this paper, we introduce Burst, a social media design that allows users to share and curate content between many spaces of varied size and composition. Users initially post content to small trusted groups, who can then burst that content, routing it to the groups that would be the best audience. We instantiate this approach into a mobile phone application, and demonstrate through a ten-day field study (N=36) that Burst enabled a participatory curation culture. With this work, we aim to articulate potential new design directions for social media sharing.

cs.HC↗

arXiv2025

TrajFlow: Multi-modal Motion Prediction via Flow Matching

Efficient and accurate motion prediction is crucial for ensuring safety and informed decision-making in autonomous driving, particularly under dynamic real-world conditions that necessitate multi-modal forecasts. We introduce TrajFlow, a novel flow matching-based motion prediction framework that addresses the scalability and efficiency challenges of existing generative trajectory prediction methods. Unlike conventional generative approaches that employ i.i.d. sampling and require multiple inference passes to capture diverse outcomes, TrajFlow predicts multiple plausible future trajectories in a single pass, significantly reducing computational overhead while maintaining coherence across predictions. Moreover, we propose a ranking loss based on the Plackett-Luce distribution to improve uncertainty estimation of predicted trajectories. Additionally, we design a self-conditioning training technique that reuses the model's own predictions to construct noisy inputs during a second forward pass, thereby improving generalization and accelerating inference. Extensive experiments on the large-scale Waymo Open Motion Dataset (WOMD) demonstrate that TrajFlow achieves state-of-the-art performance across various key metrics, underscoring its effectiveness for safety-critical autonomous driving applications. The code and other details are available on the project website https://traj-flow.github.io/.

cs.CV↗

arXiv2025

Organize, Then Vote: Exploring Cognitive Load in Quadratic Survey Interfaces

Quadratic Surveys (QSs) elicit more accurate preferences than traditional methods like Likert-scale surveys. However, the cognitive load associated with QSs has hindered their adoption in digital surveys for collective decision-making. We introduce a two-phase "organize-then-vote" QS to reduce cognitive load. As interface design significantly impacts survey results and accuracy, our design scaffolds survey takers' decision-making while managing the cognitive load imposed by QS. In a 2x2 between-subject in-lab study on public resource allotment, we compared our interface with a traditional text interface across a QS with 6 (short) and 24 (long) options. Two-phase interface participants spent more time per option and exhibited shorter voting edit distances. We qualitatively observed shifts in cognitive effort from mechanical operations to constructing more comprehensive preferences. We conclude that this interface promoted deeper engagement, potentially reducing satisficing behaviors caused by cognitive overload in longer QSs. This research clarifies how human-centered design improves preference elicitation tools for collective decision-making.

cs.HC↗

arXiv2025

Unfettered Forceful Skill Acquisition with Physical Reasoning and Coordinate Frame Labeling

Vision language models (VLMs) exhibit vast knowledge of the physical world, including intuition of physical and spatial properties, affordances, and motion. With fine-tuning, VLMs can also natively produce robot trajectories. We demonstrate that eliciting wrenches, not trajectories, allows VLMs to explicitly reason about forces and leads to zero-shot generalization in a series of manipulation tasks without pretraining. We achieve this by overlaying a consistent visual representation of relevant coordinate frames on robot-attached camera images to augment our query. First, we show how this addition enables a versatile motion control framework evaluated across four tasks (opening and closing a lid, pushing a cup or chair) spanning prismatic and rotational motion, an order of force and position magnitude, different camera perspectives, annotation schemes, and two robot platforms over 220 experiments, resulting in 51% success across the four tasks. Then, we demonstrate that the proposed framework enables VLMs to continually reason about interaction feedback to recover from task failure or incompletion, with and without human supervision. Finally, we observe that prompting schemes with visual annotation and embodied reasoning can bypass VLM safeguards. We characterize prompt component contribution to harmful behavior elicitation and discuss its implications for developing embodied reasoning. Our code, videos, and data are available at: https://scalingforce.github.io/.

cs.RO↗

arXiv2025

Mantodea phylogenomics provides new insights into X-chromosome progression and evolutionary radiation

Background: Praying mantises, members of the order Mantodea, play important roles in agriculture, medicine, bionics, and entertainment. However, the scarcity of genomic resources has hindered extensive studies on mantis evolution and behaviour. Results: Here, we present the chromosome-scale reference genomes of five mantis species: the European mantis (Mantis religiosa), Chinese mantis (Tenodera sinensis), triangle dead leaf mantis (Deroplatys truncata), orchid mantis (Hymenopus coronatus), and metallic mantis (Metallyticus violaceus). We found that transposable element expansion is the major force governing genome size in Mantodea. Based on whole-alignments, we deduced that the Mantodea ancestor may have had only one X chromosome and that translocations between the X chromosome and an autosome may have occurred in the lineage of the superfamily Mantoidea. Furthermore, we found a lower evolutionary rate for the metallic mantis than for the other mantises. We also found that Mantodea underwent rapid radiation after the K-Pg mass extinction event, which could have contributed to the confusion in species classification. Conclusions: We present the chromosome-scale reference genomes of five mantis species to reveal the X-chromosome evolution, clarify the phylogeny relationship, and transposable element expansion.

q-bio.PE↗

arXiv2025

Kimi-Audio Technical Report

We present Kimi-Audio, an open-source audio foundation model that excels in audio understanding, generation, and conversation. We detail the practices in building Kimi-Audio, including model architecture, data curation, training recipe, inference deployment, and evaluation. Specifically, we leverage a 12.5Hz audio tokenizer, design a novel LLM-based architecture with continuous features as input and discrete tokens as output, and develop a chunk-wise streaming detokenizer based on flow matching. We curate a pre-training dataset that consists of more than 13 million hours of audio data covering a wide range of modalities including speech, sound, and music, and build a pipeline to construct high-quality and diverse post-training data. Initialized from a pre-trained LLM, Kimi-Audio is continual pre-trained on both audio and text data with several carefully designed tasks, and then fine-tuned to support a diverse of audio-related tasks. Extensive evaluation shows that Kimi-Audio achieves state-of-the-art performance on a range of audio benchmarks including speech recognition, audio understanding, audio question answering, and speech conversation. We release the codes, model checkpoints, as well as the evaluation toolkits in https://github.com/MoonshotAI/Kimi-Audio.

eess.AS↗