Zhengyi Zhao

Zhengyi Zhao, Shubo Zhang, Huimin Wang et al.

Large Reasoning Models (LRMs) have demonstrated impressive capabilities in many tasks, yet they struggle with reliably following multiple instructions, either by failing to satisfy individual constraints or by struggling to balance competing constraints simultaneously. We formalize this challenge as the Constraint Adherence Problem (CAP). This paper introduces a novel framework that addresses CAP by representing instructions as a structured knowledge graph of constraints. Our approach, Constraint Relationship Graph Completion (CRGC), explicitly models relationships between constraints, identifies adherence challenges, and discovers ``bridge constraints'' that help the model better focus on and reconcile requirements. Bridge constraints act as auxiliary instructions that make primary constraints more salient and compatible. Unlike existing approaches that enhance instruction following through general training methods, CRGC specifically improves constraint satisfaction by leveraging the model's own knowledge to create better pathways for generation. Experiments across three popular instruction following datasets demonstrate that our approach reduces constraint violations by 39% compared to standard prompting while maintaining reasoning abilities of large reasoning models.

Zhengyi Zhao, Shubo Zhang, Zezhong Wang et al.

When asked what a meme or sarcastic post means, Large Vision Language Models (LVLMs) tend to describe what the image shows rather than what the author is trying to communicate. Standard instruction tuning entangles a post's literal content with its pragmatic meaning, letting surface-level details contaminate the final response. We reframe meme understanding as a problem of literal-pragmatic decomposition and propose \textbf{Intent Projection}, a framework that separates the two signals at the representation, output, and objective levels within a single LVLM backbone. At the representation level, an orthogonal projection module removes dominant unimodal directions from the fused image-text representation, retaining only the pragmatic residual, while a surface-real affect classifier anchors the decoder with a discrete tag that names the polarity gap. At the output level, the model externalizes a structured reasoning chain, and at the objective level a contrastive reward explicitly penalizes answers that restate the literal description. Across six multimodal benchmarks, Intent Projection consistently outperforms open-source baselines and narrows the gap to proprietary models, with the largest gains on high-divergence posts where literal collapse is most damaging.

Zhu Yu, Zhengyi Zhao, Runmin Zhang et al.

This work presents the Large Depth Completion Model (LDCM), a simple, effective, and robust framework for single-view metric depth estimation with sparse observations. Without relying on complex architectural designs, LDCM generates metric-accurate dense depth maps using a transformer. It outperforms existing approaches across diverse datasets and sparse observations. We achieve this from two key perspectives: (1) leveraging existing monocular foundation models to improve the quality of sparse depth inputs, and (2) reformulating training objectives to better capture geometric structure and metric consistency. Specifically, a Poisson-based depth initialization strategy is first introduced to generate a uniform coarse dense depth map from diverse sparse observations, providing a strong structural prior for the network. Regarding the training objective, we replace the conventional depth head with a point map head that regresses per-pixel 3D coordinates in camera space, enabling the model to directly learn the underlying 3D scene structure instead of performing pixel-wise depth map restoration. Moreover, this design eliminates the need for camera intrinsic parameters, allowing LDCM to naturally produce metric-scaled 3D point maps. Extensive experiments demonstrate that LDCM consistently outperforms state-of-the-art methods across multiple benchmarks and varying sparsity levels in both depth completion and point map estimation, showcasing its effectiveness and strong generalization to unseen data distributions.

Zhu Yu, Jingnan Gao, Runmin Zhang et al.

This work presents ViGeo, a feed-forward foundation model for recovering spatially dense and temporally consistent geometry from video sequences. Built upon a plain transformer architecture without task-specific architectural modifications, ViGeo supports streaming, full-sequence, and long-video inference within a unified model. The key design is dynamic chunking attention, which exposes the model to both bidirectional and causal temporal contexts during training and allows it to adapt its attention pattern at test time without retraining. To improve supervision quality, we further introduce a completion-based data refinement framework. This framework trains a video depth completion teacher that conditions on sparse and noisy annotations and exploits video/multi-view context to produce dense, temporally coherent, and geometrically reliable training targets. Beyond depth and point maps, ViGeo also predicts surface normals within the same framework. Trained solely on public datasets, ViGeo achieves state-of-the-art performance across online, offline, and long-video depth estimation, surface normal estimation, and video point map estimation.

Zhengyi Zhao, Shubo Zhang, Yuxi Zhang et al.

Large Language Models (LLMs) have shown impressive capabilities in complex reasoning tasks. However, current approaches employ uniform language density for both intermediate reasoning and final answers, leading to computational inefficiency. Our observation found that reasoning process serves a computational function for the model itself, while answering serves a communicative function for human understanding. This distinction enables the use of compressed, symbol-rich language for intermediate computations while maintaining human-readable final explanations. To address this inefficiency, we present Denser: \underline{D}ual-d\underline{ens}ity inf\underline{er}ence, a novel framework that optimizes information density separately for reasoning and answering phases. Our framework implements this through three components: a query processing module that analyzes input problems, a high-density compressed reasoning mechanism for efficient intermediate computations, and an answer generation component that translates compressed reasoning into human-readable solutions. Experimental evaluation across multiple reasoning question answering benchmarks demonstrates that Denser reduces token consumption by up to 62\% compared to standard Chain-of-Thought methods while preserving or improving accuracy. These efficiency gains are particularly significant for complex multi-step reasoning problems where traditional methods generate extensive explanations.

Yisheng He, Xiaodong Gu, Xiaodan Ye et al.

We present LAM, an innovative Large Avatar Model for animatable Gaussian head reconstruction from a single image. Unlike previous methods that require extensive training on captured video sequences or rely on auxiliary neural networks for animation and rendering during inference, our approach generates Gaussian heads that are immediately animatable and renderable. Specifically, LAM creates an animatable Gaussian head in a single forward pass, enabling reenactment and rendering without additional networks or post-processing steps. This capability allows for seamless integration into existing rendering pipelines, ensuring real-time animation and rendering across a wide range of platforms, including mobile phones. The centerpiece of our framework is the canonical Gaussian attributes generator, which utilizes FLAME canonical points as queries. These points interact with multi-scale image features through a Transformer to accurately predict Gaussian attributes in the canonical space. The reconstructed canonical Gaussian avatar can then be animated utilizing standard linear blend skinning (LBS) with corrective blendshapes as the FLAME model did and rendered in real-time on various platforms. Our experimental results demonstrate that LAM outperforms state-of-the-art methods on existing benchmarks. Our code and video are available at https://aigc3d.github.io/projects/LAM/

Weichao Shen, Yuan Dong, Zonghao Chen et al.

In this paper, we propose PanoViT, a panorama vision transformer to estimate the room layout from a single panoramic image. Compared to CNN models, our PanoViT is more proficient in learning global information from the panoramic image for the estimation of complex room layouts. Considering the difference between a perspective image and an equirectangular image, we design a novel recurrent position embedding and a patch sampling method for the processing of panoramic images. In addition to extracting global information, PanoViT also includes a frequency-domain edge enhancement module and a 3D loss to extract local geometric features in a panoramic image. Experimental results on several datasets demonstrate that our method outperforms state-of-the-art solutions in room layout prediction accuracy.

Qi Zuo, Xiaodong Gu, Lingteng Qiu et al.

Generating multi-view images based on text or single-image prompts is a critical capability for the creation of 3D content. Two fundamental questions on this topic are what data we use for training and how to ensure multi-view consistency. This paper introduces a novel framework that makes fundamental contributions to both questions. Unlike leveraging images from 2D diffusion models for training, we propose a dense consistent multi-view generation model that is fine-tuned from off-the-shelf video generative models. Images from video generative models are more suitable for multi-view generation because the underlying network architecture that generates them employs a temporal module to enforce frame consistency. Moreover, the video data sets used to train these models are abundant and diverse, leading to a reduced train-finetuning domain gap. To enhance multi-view consistency, we introduce a 3D-Aware Denoising Sampling, which first employs a feed-forward reconstruction module to get an explicit global 3D model, and then adopts a sampling strategy that effectively involves images rendered from the global 3D model into the denoising sampling loop to improve the multi-view consistency of the final images. As a by-product, this module also provides a fast way to create 3D assets represented by 3D Gaussians within a few seconds. Our approach can generate 24 dense views and converges much faster in training than state-of-the-art approaches (4 GPU hours versus many thousand GPU hours) with comparable visual quality and consistency. By further fine-tuning, our approach outperforms existing state-of-the-art methods in both quantitative metrics and visual effects. Our project page is aigc3d.github.io/VideoMV.

Yiming Du, Hongru Wang, Zhengyi Zhao et al.

Long-term memory plays a critical role in personal interaction, considering long-term memory can better leverage world knowledge, historical information, and preferences in dialogues. Our research introduces PerLTQA, an innovative QA dataset that combines semantic and episodic memories, including world knowledge, profiles, social relationships, events, and dialogues. This dataset is collected to investigate the use of personalized memories, focusing on social interactions and events in the QA task. PerLTQA features two types of memory and a comprehensive benchmark of 8,593 questions for 30 characters, facilitating the exploration and application of personalized memories in Large Language Models (LLMs). Based on PerLTQA, we propose a novel framework for memory integration and generation, consisting of three main components: Memory Classification, Memory Retrieval, and Memory Synthesis. We evaluate this framework using five LLMs and three retrievers. Experimental results demonstrate that BERT-based classification models significantly outperform LLMs such as ChatGLM3 and ChatGPT in the memory classification task. Furthermore, our study highlights the importance of effective memory integration in the QA task.

Shenhao Zhu, Lingteng Qiu, Xiaodong Gu et al.

Existing 2D methods utilize UNet-based diffusion models to generate multi-view physically-based rendering (PBR) maps but struggle with multi-view inconsistency, while some 3D methods directly generate UV maps, encountering generalization issues due to the limited 3D data. To address these problems, we propose a two-stage approach, including multi-view generation and UV materials refinement. In the generation stage, we adopt a Diffusion Transformer (DiT) model to generate PBR materials, where both the specially designed multi-branch DiT and reference-based DiT blocks adopt a global attention mechanism to promote feature interaction and fusion between different views, thereby improving multi-view consistency. In addition, we adopt a PBR-based diffusion loss to ensure that the generated materials align with realistic physical principles. In the refinement stage, we propose a material-refined DiT that performs inpainting in empty areas and enhances details in UV space. Except for the normal condition, this refinement also takes the material map from the generation stage as an additional condition to reduce the learning difficulty and improve generalization. Extensive experiments show that our method achieves state-of-the-art performance in texturing 3D objects with PBR materials and provides significant advantages for graphics relighting applications. Project Page: https://lingtengqiu.github.io/2024/MCMat/

Zhengyi Zhao, Shubo Zhang, Yiming Du et al.

Existing large language models (LLMs) have shown remarkable progress in dialogue systems. However, many approaches still overlook the fundamental role of events throughout multi-turn interactions, leading to \textbf{incomplete context tracking}. Without tracking these events, dialogue systems often lose coherence and miss subtle shifts in user intent, causing disjointed responses. To bridge this gap, we present \textbf{EventWeave}, an event-centric framework that identifies and updates both core and supporting events as the conversation unfolds. Specifically, we organize these events into a dynamic event graph, which represents the interplay between \textbf{core events} that shape the primary idea and \textbf{supporting events} that provide critical context during the whole dialogue. By leveraging this dynamic graph, EventWeave helps models focus on the most relevant events when generating responses, thus avoiding repeated visits of the entire dialogue history. Experimental results on two benchmark datasets show that EventWeave improves response quality and event relevance without fine-tuning.

Zhengyi Zhao, Shubo Zhang, Zezhong Wang et al.

Retrieval-Augmented Generation (RAG) significantly enhances Large Language Models (LLMs) by providing access to external knowledge. However, current research primarily focuses on retrieval quality, often overlooking the critical ''integration bottleneck'': even when relevant documents are retrieved, LLMs frequently fail to utilize them effectively due to conflicts with their internal parametric knowledge. In this paper, we argue that implicitly resolving this conflict in a single generation pass is suboptimal. We introduce GuarantRAG, a framework that explicitly decouples reasoning from evidence integration. First, we generate an ''Inner-Answer'' based solely on parametric knowledge to capture the model's reasoning flow. Second, to guarantee faithful evidence extraction, we generate a ''Refer-Answer'' using a novel Contrastive DPO objective. This objective treats the parametric Inner-Answer as a negative constraint and the retrieved documents as positive ground truth, forcing the model to suppress internal hallucinations in favor of external evidence during this phase. Finally, rather than naive concatenation or using the DPO trained model directly, we propose a joint decoding mechanism that dynamically fuses the logical coherence of the Inner-Answer with the factual precision of the Refer-Answer at the token level. Experiments on five QA benchmarks demonstrate that GuarantRAG improves accuracy by up to 12.1% and reduces hallucinations by 16.3% compared to standard and dynamic RAG baselines.

Zhengyi Zhao, Shubo Zhang, Zezhong Wang et al.

Recent advances in Large Language Models (LLMs) have demonstrated remarkable performance in Contextual Question Answering (CQA). However, prior approaches typically employ elaborate reasoning strategies regardless of question complexity, leading to low adaptability. Recent efficient test-time scaling methods introduce budget constraints or early stop mechanisms to avoid overthinking for straightforward questions. But they add human bias to the reasoning process and fail to leverage models' inherent reasoning capabilities. To address these limitations, we present T$^2$: Think-to-Think, a novel framework that dynamically adapts reasoning depth based on question complexity. T$^2$ leverages the insight that if an LLM can effectively solve similar questions using specific reasoning strategies, it can apply the same strategy to the original question. This insight enables to adoption of concise reasoning for straightforward questions while maintaining detailed analysis for complex problems. T$^2$ works through four key steps: decomposing questions into structural elements, generating similar examples with candidate reasoning strategies, evaluating these strategies against multiple criteria, and applying the most appropriate strategy to the original question. Experimental evaluation across seven diverse CQA benchmarks demonstrates that T$^2$ not only achieves higher accuracy than baseline methods but also reduces computational overhead by up to 25.2\%.

Zhengyi Zhao, Shubo Zhang, Bin Liang et al.

In Biomedical Natural Language Processing (BioNLP) tasks, such as Relation Extraction, Named Entity Recognition, and Text Classification, the scarcity of high-quality data remains a significant challenge. This limitation poisons large language models to correctly understand relationships between biological entities, such as molecules and diseases, or drug interactions, and further results in potential misinterpretation of biomedical documents. To address this issue, current approaches generally adopt the Synthetic Data Augmentation method which involves similarity computation followed by word replacement, but counterfactual data are usually generated. As a result, these methods disrupt meaningful word sets or produce sentences with meanings that deviate substantially from the original context, rendering them ineffective in improving model performance. To this end, this paper proposes a biomedical-dedicated rationale-based synthetic data augmentation method. Beyond the naive lexicon similarity, specific bio-relation similarity is measured to hold the augmented instance having a strong correlation with bio-relation instead of simply increasing the diversity of augmented data. Moreover, a multi-agents-involved reflection mechanism helps the model iteratively distinguish different usage of similar entities to escape falling into the mis-replace trap. We evaluate our method on the BLURB and BigBIO benchmark, which includes 9 common datasets spanning four major BioNLP tasks. Our experimental results demonstrate consistent performance improvements across all tasks, highlighting the effectiveness of our approach in addressing the challenges associated with data scarcity and enhancing the overall performance of biomedical NLP models.

Zhengyi Zhao, Chen Song, Xiaodong Gu et al.

A fundamental problem in the texturing of 3D meshes using pre-trained text-to-image models is to ensure multi-view consistency. State-of-the-art approaches typically use diffusion models to aggregate multi-view inputs, where common issues are the blurriness caused by the averaging operation in the aggregation step or inconsistencies in local features. This paper introduces an optimization framework that proceeds in four stages to achieve multi-view consistency. Specifically, the first stage generates an over-complete set of 2D textures from a predefined set of viewpoints using an MV-consistent diffusion process. The second stage selects a subset of views that are mutually consistent while covering the underlying 3D model. We show how to achieve this goal by solving semi-definite programs. The third stage performs non-rigid alignment to align the selected views across overlapping regions. The fourth stage solves an MRF problem to associate each mesh face with a selected view. In particular, the third and fourth stages are iterated, with the cuts obtained in the fourth stage encouraging non-rigid alignment in the third stage to focus on regions close to the cuts. Experimental results show that our approach significantly outperforms baseline approaches both qualitatively and quantitatively. Project page: https://aigc3d.github.io/ConsistenTex.

Zhengyi Zhao, Shubo Zhang, Yuxi Zhang et al.

Memes have emerged as a popular form of multimodal online communication, where their interpretation heavily depends on the specific context in which they appear. Current approaches predominantly focus on isolated meme analysis, either for harmful content detection or standalone interpretation, overlooking a fundamental challenge: the same meme can express different intents depending on its conversational context. This oversight creates an evaluation gap: although humans intuitively recognize how context shapes meme interpretation, Large Vision Language Models (LVLMs) can hardly understand context-dependent meme intent. To address this critical limitation, we introduce MemeReaCon, a novel benchmark specifically designed to evaluate how LVLMs understand memes in their original context. We collected memes from five different Reddit communities, keeping each meme's image, the post text, and user comments together. We carefully labeled how the text and meme work together, what the poster intended, how the meme is structured, and how the community responded. Our tests with leading LVLMs show a clear weakness: models either fail to interpret critical information in the contexts, or overly focus on visual details while overlooking communicative purpose. MemeReaCon thus serves both as a diagnostic tool exposing current limitations and as a challenging benchmark to drive development toward more sophisticated LVLMs of the context-aware understanding.

Zhengyi Zhao, Shubo Zhang, Zezhong Wang et al.

Long-context question-answering (LCQA) systems have greatly benefited from the powerful reasoning capabilities of large language models (LLMs), which can be categorized into slow and quick reasoning modes. However, both modes have their limitations. Slow thinking generally leans to explore every possible reasoning path, which leads to heavy overthinking and wastes time. Quick thinking usually relies on pattern matching rather than truly understanding the query logic, which misses proper understanding. To address these issues, we propose FReM: Flexible Reasoning Mechanism, a method that adjusts reasoning depth according to the complexity of each question. Specifically, FReM leverages synthetic reference QA examples to provide an explicit chain of thought, enabling efficient handling of simple queries while allowing deeper reasoning for more complex ones. By doing so, FReM helps quick-thinking models move beyond superficial pattern matching and narrows the reasoning space for slow-thinking models to avoid unnecessary exploration. Experiments on seven QA datasets show that FReM improves reasoning accuracy and scalability, particularly for complex multihop questions, indicating its potential to advance LCQA methodologies.