Shiqi Chen

I-Chun Chern, Steffi Chern, Shiqi Chen et al. · cmu

The emergence of generative pre-trained models has facilitated the synthesis of high-quality text, but it has also posed challenges in identifying factual errors in the generated text. In particular: (1) A wider range of tasks now face an increasing risk of containing factual errors when handled by generative models. (2) Generated texts tend to be lengthy and lack a clearly defined granularity for individual facts. (3) There is a scarcity of explicit evidence available during the process of fact checking. With the above challenges in mind, in this paper, we propose FacTool, a task and domain agnostic framework for detecting factual errors of texts generated by large language models (e.g., ChatGPT). Experiments on four different tasks (knowledge-based QA, code generation, mathematical reasoning, and scientific literature review) show the efficacy of the proposed method. We release the code of FacTool associated with ChatGPT plugin interface at https://github.com/GAIR-NLP/factool .

Kangrui Wang, Linjie Li, Zhengyuan Yang et al.

Can VLMs predict how each camera move changes the view, and plan many such moves ahead? We call this capability view planning, requiring (1)understanding how a single action transforms the view, and (2)composing many such transformations across multi-turn plans to identify a target view. We probe both abilities in our proposed ViewSuite, a 3D point-cloud environment on real ScanNet scenes. Across 13 frontier VLMs, a critical planning gap emerges: they possess basic view-action knowledge but fail to compose it across multi-turn plans, with the gap widening as viewpoint distance grows. To close this gap, we propose an iterative framework that alternates self-exploration with view graph distillation. The key insight is that all exploration trajectories, regardless of their outcome, collectively form a view graph that compactly captures how viewpoints connect across a scene. Distilling this graph into diverse supervised tasks reshapes the policy distribution and overcomes the sparse rewards that stall pure RL. This improves Qwen2.5-VL-7B from 2.5% to 47.8% on interactive view planning, surpassing GPT-5.4 Pro (18.5%) and Gemini 3.1 Pro (21.4%). Self-exploration emerges as a promising path toward VLMs that can actively reason and plan in 3D space.

Zheyan Jin, Shiqi Chen, Yueting Chen et al.

The large language model and high-level vision model have achieved impressive performance improvements with large datasets and model sizes. However, low-level computer vision tasks, such as image dehaze and blur removal, still rely on a small number of datasets and small-sized models, which generally leads to overfitting and local optima. Therefore, we propose a framework to integrate large-model prior into low-level computer vision tasks. Just as with the task of image segmentation, the degradation of haze is also texture-related. So we propose to detect gray-scale coding, network channel expansion, and pre-dehaze structures to integrate large-model prior knowledge into any low-level dehazing network. We demonstrate the effectiveness and applicability of large models in guiding low-level visual tasks through different datasets and algorithms comparison experiments. Finally, we demonstrate the effect of grayscale coding, network channel expansion, and recurrent network structures through ablation experiments. Under the conditions where additional data and training resources are not required, we successfully prove that the integration of large-model prior knowledge will improve the dehaze performance and save training time for low-level visual tasks.

Jinghan Zhang, Shiqi Chen, Junteng Liu et al.

As an efficient alternative to conventional full finetuning, parameter-efficient finetuning (PEFT) is becoming the prevailing method to adapt pretrained language models. In PEFT, a lightweight module is learned on each dataset while the underlying pretrained language model remains unchanged, resulting in multiple compact modules representing diverse skills when applied to various domains and tasks. In this paper, we propose to compose these parameter-efficient modules through linear arithmetic operations in the weight space, thereby integrating different module capabilities. Specifically, we first define addition and negation operators for the module, and then further compose these two basic operators to perform flexible arithmetic. Our approach requires \emph{no additional training} and enables highly flexible module composition. We apply different arithmetic operations to compose the parameter-efficient modules for (1) distribution generalization, (2) multi-tasking, (3) unlearning, and (4) domain transfer. Additionally, we extend our approach to detoxify Alpaca-LoRA, the latest instruction-tuned large language model based on LLaMA. Empirical results demonstrate that our approach produces new and effective parameter-efficient modules that significantly outperform existing ones across all settings.

Jingwen Zhou, Shiqi Chen, Zheng Ren et al.

The joint design of the optical system and the downstream algorithm is a challenging and promising task. Due to the demand for balancing the global optimal of imaging systems and the computational cost of physical simulation, existing methods cannot achieve efficient joint design of complex systems such as smartphones and drones. In this work, starting from the perspective of the optical design, we characterize the optics with separated aberrations. Additionally, to bridge the hardware and software without gradients, an image simulation system is presented to reproduce the genuine imaging procedure of lenses with large field-of-views. As for aberration correction, we propose a network to perceive and correct the spatially varying aberrations and validate its superiority over state-of-the-art methods. Comprehensive experiments reveal that the preference for correcting separated aberrations in joint design is as follows: longitudinal chromatic aberration, lateral chromatic aberration, spherical aberration, field curvature, and coma, with astigmatism coming last. Drawing from the preference, a 10% reduction in the total track length of the consumer-level mobile phone lens module is accomplished. Moreover, this procedure spares more space for manufacturing deviations, realizing extreme-quality enhancement of computational photography. The optimization paradigm provides innovative insight into the practical joint design of sophisticated optical systems and post-processing algorithms.

Shiqi Chen, Yiran Zhao, Jinghan Zhang et al.

Assessing factuality of text generated by large language models (LLMs) is an emerging yet crucial research area, aimed at alerting users to potential errors and guiding the development of more reliable LLMs. Nonetheless, the evaluators assessing factuality necessitate suitable evaluation themselves to gauge progress and foster advancements. This direction remains under-explored, resulting in substantial impediments to the progress of factuality evaluators. To mitigate this issue, we introduce a benchmark for Factuality Evaluation of large Language Models, referred to as felm. In this benchmark, we collect responses generated from LLMs and annotate factuality labels in a fine-grained manner. Contrary to previous studies that primarily concentrate on the factuality of world knowledge (e.g.~information from Wikipedia), felm focuses on factuality across diverse domains, spanning from world knowledge to math and reasoning. Our annotation is based on text segments, which can help pinpoint specific factual errors. The factuality annotations are further supplemented by predefined error types and reference links that either support or contradict the statement. In our experiments, we investigate the performance of several LLM-based factuality evaluators on felm, including both vanilla LLMs and those augmented with retrieval mechanisms and chain-of-thought processes. Our findings reveal that while retrieval aids factuality evaluation, current LLMs are far from satisfactory to faithfully detect factual errors.

Zheyan Jin, Shiqi Chen, Huajun Feng et al.

We present an image dehazing algorithm with high quality, wide application, and no data training or prior needed. We analyze the defects of the original dehazing model, and propose a new and reliable dehazing reconstruction and dehazing model based on the combination of optical scattering model and computer graphics lighting rendering model. Based on the new haze model and the images obtained by the cameras, we can reconstruct the three-dimensional space, accurately calculate the objects and haze in the space, and use the transparency relationship of haze to perform accurate haze removal. To obtain a 3D simulation dataset we used the Unreal 5 computer graphics rendering engine. In order to obtain real shot data in different scenes, we used fog generators, array cameras, mobile phones, underwater cameras and drones to obtain haze data. We use formula derivation, simulation data set and real shot data set result experimental results to prove the feasibility of the new method. Compared with various other methods, we are far ahead in terms of calculation indicators (4 dB higher quality average scene), color remains more natural, and the algorithm is more robust in different scenarios and best in the subjective perception.

Zheyan Jin, Shiqi Chen, Huajun Feng et al.

Photographing in the under-illuminated scenes, the presence of complex light sources often leave strong flare artifacts in images, where the intensity, the spectrum, the reflection, and the aberration altogether contribute the deterioration. Besides the image quality, it also influence the performance of down-stream visual applications. Thus, removing the lens flare and ghosts is a challenge issue especially in low-light environment. However, existing methods for flare removal mainly restricted to the problems of inadequate simulation and real-world capture, where the categories of scattered flares are singular and the reflected ghosts are unavailable. Therefore, a comprehensive deterioration procedure is crucial for constructing the dataset of flare removal. Based on the theoretical analysis and real-world evaluation, we propose a well-developed methodology for generating the data-pairs with flare deterioration. The procedure is comprehensive, where the similarity of scattered flares and the symmetric effect of reflected ghosts are realized. Moreover, we also construct a real-shot pipeline that respectively processes the effects of scattering and reflective flares, aiming to directly generate the data for end-to-end methods. Experimental results show that the proposed methodology add diversity to the existing flare datasets and construct a comprehensive mapping procedure for flare data pairs. And our method facilities the data-driven model to realize better restoration in flare images and proposes a better evaluation system based on real shots, resulting promote progress in the area of real flare removal.

Junteng Liu, Shiqi Chen, Yu Cheng et al.

While large language models (LLMs) have demonstrated remarkable abilities across various fields, hallucination remains a significant challenge. Recent studies have explored hallucinations through the lens of internal representations, proposing mechanisms to decipher LLMs' adherence to facts. However, these approaches often fail to generalize to out-of-distribution data, leading to concerns about whether internal representation patterns reflect fundamental factual awareness, or only overfit spurious correlations on the specific datasets. In this work, we investigate whether a universal truthfulness hyperplane that distinguishes the model's factually correct and incorrect outputs exists within the model. To this end, we scale up the number of training datasets and conduct an extensive evaluation -- we train the truthfulness hyperplane on a diverse collection of over 40 datasets and examine its cross-task, cross-domain, and in-domain generalization. Our results indicate that increasing the diversity of the training datasets significantly enhances the performance in all scenarios, while the volume of data samples plays a less critical role. This finding supports the optimistic hypothesis that a universal truthfulness hyperplane may indeed exist within the model, offering promising directions for future research.

Jiajian He, Enjie Hu, Shiqi Chen et al.

The point spread function (PSF) serves as a fundamental descriptor linking the real-world scene to the captured signal, manifesting as camera blur. Accurate PSF estimation is crucial for both optical characterization and computational vision, yet remains challenging due to the inherent ambiguity and the ill-posed nature of intensity-based deconvolution. We introduce CircleFlow, a high-fidelity PSF estimation framework that employs flow-guided edge localization for precise blur characterization. CircleFlow begins with a structured capture that encodes locally anisotropic and spatially varying PSFs by imaging a circle grid target, while leveraging the target's binary luminance prior to decouple image and kernel estimation. The latent sharp image is then reconstructed through subpixel alignment of an initialized binary structure guided by optical flow, whereas the PSF is modeled as an energy-constrained implicit neural representation. Both components are jointly optimized within a demosaicing-aware differentiable framework, ensuring physically consistent and robust PSF estimation enabled by accurate edge localization. Extensive experiments on simulated and real-world data demonstrate that CircleFlow achieves state-of-the-art accuracy and reliability, validating its effectiveness for practical PSF calibration.

Shiqi Chen, Tongyao Zhu, Ruochen Zhou et al. · deepmind, salesforce

Large Vision Language Models (VLMs) have long struggled with spatial reasoning tasks. Surprisingly, even simple spatial reasoning tasks, such as recognizing "under" or "behind" relationships between only two objects, pose significant challenges for current VLMs. In this work, we study the spatial reasoning challenge from the lens of mechanistic interpretability, diving into the model's internal states to examine the interactions between image and text tokens. By tracing attention distribution over the image through out intermediate layers, we observe that successful spatial reasoning correlates strongly with the model's ability to align its attention distribution with actual object locations, particularly differing between familiar and unfamiliar spatial relationships. Motivated by these findings, we propose ADAPTVIS based on inference-time confidence scores to sharpen the attention on highly relevant regions when confident, while smoothing and broadening the attention window to consider a wider context when confidence is lower. This training-free decoding method shows significant improvement (e.g., up to a 50 absolute point improvement) on spatial reasoning benchmarks such as WhatsUp and VSR with negligible cost. We make code and data publicly available for research purposes at https://github.com/shiqichen17/AdaptVis.

Ruikang Li, Yujin Wang, Shiqi Chen et al.

Image denoising is a critical component in a camera's Image Signal Processing (ISP) pipeline. There are two typical ways to inject a denoiser into the ISP pipeline: applying a denoiser directly to captured raw frames (raw domain) or to the ISP's output sRGB images (sRGB domain). However, both approaches have their limitations. Residual noise from raw-domain denoising can be amplified by the subsequent ISP processing, and the sRGB domain struggles to handle spatially varying noise since it only sees noise distorted by the ISP. Consequently, most raw or sRGB domain denoising works only for specific noise distributions and ISP configurations. To address these challenges, we propose DualDn, a novel learning-based dual-domain denoising. Unlike previous single-domain denoising, DualDn consists of two denoising networks: one in the raw domain and one in the sRGB domain. The raw domain denoising adapts to sensor-specific noise as well as spatially varying noise levels, while the sRGB domain denoising adapts to ISP variations and removes residual noise amplified by the ISP. Both denoising networks are connected with a differentiable ISP, which is trained end-to-end and discarded during the inference stage. With this design, DualDn achieves greater generalizability compared to most learning-based denoising methods, as it can adapt to different unseen noises, ISP parameters, and even novel ISP pipelines. Experiments show that DualDn achieves state-of-the-art performance and can adapt to different denoising architectures. Moreover, DualDn can be used as a plug-and-play denoising module with real cameras without retraining, and still demonstrate better performance than commercial on-camera denoising. The project website is available at: https://openimaginglab.github.io/DualDn/

Xiao Wang, Yiyang Wu, Yuntian Wang et al.

Despite offering high sensitivity, a high signal-to-noise ratio, and a broad spectral range, single-pixel imaging (SPI) is limited by low measurement efficiency and long data-acquisition times. To address this, we propose a wavelength-multiplexed, spatially incoherent diffractive optical processor combined with a compact/shallow digital artificial neural network (ANN) to implement compressive SPI. Specifically, we model the bucket detection process in conventional SPI as a linear intensity transformation with spatially and spectrally varying point-spread functions. This transformation matrix is treated as a learnable parameter and jointly optimized with a shallow digital ANN composed of 2 hidden nonlinear layers. The wavelength-multiplexed diffractive processor is then configured via data-free optimization to approximate this pre-trained transformation matrix; after this optimization, the diffractive processor remains static/fixed. Upon multi-wavelength illumination and diffractive modulation, the target spatial information of the input object is spectrally encoded. A single-pixel detector captures the output spectral power at each illumination band, which is then rapidly decoded by the jointly trained digital ANN to reconstruct the input image. In addition to our numerical analyses demonstrating the feasibility of this approach, we experimentally validated its proof-of-concept using an array of light-emitting diodes (LEDs). Overall, this work demonstrates a computational imaging framework for compressive SPI that can be useful in applications such as biomedical imaging, autonomous devices, and remote sensing.

Cagatay Isil, Alexander Chen, Yuhang Li et al.

3D image display is essential for next-generation volumetric imaging; however, dense depth multiplexing for 3D image projection remains challenging because diffraction-induced cross-talk rapidly increases as the axial image planes get closer. Here, we introduce a 3D display system comprising a digital encoder and a diffractive optical decoder, which simultaneously projects different images onto multiple target axial planes with high axial resolution. By leveraging multi-layer diffractive wavefront decoding and deep learning-based end-to-end optimization, the system achieves high-fidelity depth-resolved 3D image projection in a snapshot, enabling axial plane separations on the order of a wavelength. The digital encoder leverages a Fourier encoder network to capture multi-scale spatial and frequency-domain features from input images, integrates axial position encoding, and generates a unified phase representation that simultaneously encodes all images to be axially projected in a single snapshot through a jointly-optimized diffractive decoder. We characterized the impact of diffractive decoder depth, output diffraction efficiency, spatial light modulator resolution, and axial encoding density, revealing trade-offs that govern axial separation and 3D image projection quality. We further demonstrated the capability to display volumetric images containing 28 axial slices, as well as the ability to dynamically reconfigure the axial locations of the image planes, performed on demand. Finally, we experimentally validated the presented approach, demonstrating close agreement between the measured results and the target images. These results establish the diffractive 3D display system as a compact and scalable framework for depth-resolved snapshot 3D image projection, with potential applications in holographic displays, AR/VR interfaces, and volumetric optical computing.

Parnian Ghapandar Kashani, Shiqi Chen, Aydogan Ozcan

The rapid proliferation of AI-generated visual media has created an urgent need for efficient, trustworthy deepfake detection systems. However, existing deep learning-based detection methods rely on computationally intensive and energy-demanding inference algorithms, limiting their scalability. Here, we present a hybrid digital-analog deepfake video detection framework that combines a lightweight digital front-end with a spatially multiplexed optical decoding back-end for massively parallel analog inference through a programmable spatial light modulator. By simultaneously processing 15 or more video streams within a single optical propagation pass, the system enables high-throughput and accurate video-level authenticity prediction at reduced computational cost compared with purely digital methods. We validated this hybrid deepfake video processor using different datasets spanning classical face-swapping, real-world deepfake recordings, and fully AI-generated videos. Using a spatially multiplexed experimental set-up operating in the visible spectrum, we achieved average deepfake detection accuracy, sensitivity and specificity of 97.79%, 99.86% and 95.72%, respectively, on the Celeb-DF video dataset with 15 videos tested in parallel in a single optical pass per inference. The multiplexed optical decoder also demonstrates resilience against various types of video degradation, noise, compression, experimental misalignments and black-box adversarial attacks. Our results show that integrating optical computation into AI inference enables simultaneous gains in throughput, energy efficiency, and adversarial robustness - three properties that are difficult to achieve together in purely digital systems.

Zihan Wang, Chi Gui, Xing Jin et al.

RL training of multi-turn LLM agents is inherently unstable, and reasoning quality directly determines task performance. Entropy is widely used to track reasoning stability. However, entropy only measures diversity within the same input, and cannot tell whether reasoning actually responds to different inputs. In RAGEN-2, we find that even with stable entropy, models can rely on fixed templates that look diverse but are input-agnostic. We call this template collapse, a failure mode invisible to entropy and all existing metrics. To diagnose this failure, we decompose reasoning quality into within-input diversity (Entropy) and cross-input distinguishability (Mutual Information, MI), and introduce a family of mutual information proxies for online diagnosis. Across diverse tasks, mutual information correlates with final performance much more strongly than entropy, making it a more reliable proxy for reasoning quality. We further explain template collapse with a signal-to-noise ratio (SNR) mechanism. Low reward variance weakens task gradients, letting regularization terms dominate and erase cross-input reasoning differences. To address this, we propose SNR-Aware Filtering to select high-signal prompts per iteration using reward variance as a lightweight proxy. Across planning, math reasoning, web navigation, and code execution, the method consistently improves both input dependence and task performance.

Ke Wu, Shiqi Chen, Yiheng Zhong et al.

Segment Anything Model (SAM) enable scalable medical image segmentation but suffer from inference-time instability when deployed as a frozen backbone. In practice, bounding-box prompts often contain localization errors, and fixed threshold binarization introduces additional decision uncertainty. These factors jointly cause high prediction variance, especially near object boundaries, degrading reliability. We propose the Stability-Aware Inference Framework (SAIF), a training-free and plug-and-play inference framework that improves robustness by explicitly modeling prompt and threshold uncertainty. SAIF constructs a joint uncertainty space via structured box perturbations and threshold variations, evaluates each hypothesis using decision stability and boundary consistency, and introduces a stability-consistency score to filter unstable candidates and perform stability-weighted fusion in probability space. Experiments on Synapse, CVC-ClinicDB, Kvasir-SEG, and CVC-300 demonstrate that SAIF consistently improves segmentation accuracy and robustness, achieving state-of-the-art performance without retraining or architectural modification. Our anonymous code is released at https://anonymous.4open.science/r/SAIF.

Tongyao Zhu, Qian Liu, Haonan Wang et al.

Recent advancements in LLM pretraining have featured ever-expanding context windows to process longer sequences. However, our pilot study reveals that models pretrained with shorter context windows consistently outperform their long-context counterparts under a fixed token budget. This finding motivates us to explore an optimal context window scheduling strategy to better balance long-context capability with pretraining efficiency. To this end, we propose SkyLadder, a simple yet effective approach that implements a short-to-long context window transition. SkyLadder preserves strong standard benchmark performance, while matching or exceeding baseline results on long context tasks. Through extensive experiments, we pre-train 1B-parameter models (up to 32K context) and 3B-parameter models (8K context) on 100B tokens, demonstrating that SkyLadder yields consistent gains of up to 3.7% on common benchmarks, while achieving up to 22% faster training speeds compared to baselines. The code is at https://github.com/sail-sg/SkyLadder.

Zheng Ren, Jingwen Zhou, Wenguan Zhang et al.

Recently, the joint design of optical systems and downstream algorithms is showing significant potential. However, existing rays-described methods are limited to optimizing geometric degradation, making it difficult to fully represent the optical characteristics of complex, miniaturized lenses constrained by wavefront aberration or diffraction effects. In this work, we introduce a precise optical simulation model, and every operation in pipeline is differentiable. This model employs a novel initial value strategy to enhance the reliability of intersection calculation on high aspherics. Moreover, it utilizes a differential operator to reduce memory consumption during coherent point spread function calculations. To efficiently address various degradation, we design a joint optimization procedure that leverages field information. Guided by a general restoration network, the proposed method not only enhances the image quality, but also successively improves the optical performance across multiple lenses that are already in professional level. This joint optimization pipeline offers innovative insights into the practical design of sophisticated optical systems and post-processing algorithms. The source code will be made publicly available at https://github.com/Zrr-ZJU/Successive-optimization

Junteng Liu, Yuanxiang Fan, Zhuo Jiang et al.

Recent advances such as OpenAI-o1 and DeepSeek R1 have demonstrated the potential of Reinforcement Learning (RL) to enhance reasoning abilities in Large Language Models (LLMs). While open-source replication efforts have primarily focused on mathematical and coding domains, methods and resources for developing general reasoning capabilities remain underexplored. This gap is partly due to the challenge of collecting diverse and verifiable reasoning data suitable for RL. We hypothesize that logical reasoning is critical for developing general reasoning capabilities, as logic forms a fundamental building block of reasoning. In this work, we present SynLogic, a data synthesis framework and dataset that generates diverse logical reasoning data at scale, encompassing 35 diverse logical reasoning tasks. The SynLogic approach enables controlled synthesis of data with adjustable difficulty and quantity. Importantly, all examples can be verified by simple rules, making them ideally suited for RL with verifiable rewards. In our experiments, we validate the effectiveness of RL training on the SynLogic dataset based on 7B and 32B models. SynLogic leads to state-of-the-art logical reasoning performance among open-source datasets, surpassing DeepSeek-R1-Distill-Qwen-32B by 6 points on BBEH. Furthermore, mixing SynLogic data with mathematical and coding tasks improves the training efficiency of these domains and significantly enhances reasoning generalization. Notably, our mixed training model outperforms DeepSeek-R1-Zero-Qwen-32B across multiple benchmarks. These findings position SynLogic as a valuable resource for advancing the broader reasoning capabilities of LLMs. We open-source both the data synthesis pipeline and the SynLogic dataset at https://github.com/MiniMax-AI/SynLogic.

Zida Chen, Ziran Zhang, Haoying Li et al.

Linear Array Pushbroom (LAP) imaging technology is widely used in the realm of remote sensing. However, images acquired through LAP always suffer from distortion and blur because of camera jitter. Traditional methods for restoring LAP images, such as algorithms estimating the point spread function (PSF), exhibit limited performance. To tackle this issue, we propose a Jitter-Aware Restoration Network (JARNet), to remove the distortion and blur in two stages. In the first stage, we formulate an Optical Flow Correction (OFC) block to refine the optical flow of the degraded LAP images, resulting in pre-corrected images where most of the distortions are alleviated. In the second stage, for further enhancement of the pre-corrected images, we integrate two jitter-aware techniques within the Spatial and Frequency Residual (SFRes) block: 1) introducing Coordinate Attention (CoA) to the SFRes block in order to capture the jitter state in orthogonal direction; 2) manipulating image features in both spatial and frequency domains to leverage local and global priors. Additionally, we develop a data synthesis pipeline, which applies Continue Dynamic Shooting Model (CDSM) to simulate realistic degradation in LAP images. Both the proposed JARNet and LAP image synthesis pipeline establish a foundation for addressing this intricate challenge. Extensive experiments demonstrate that the proposed two-stage method outperforms state-of-the-art image restoration models. Code is available at https://github.com/JHW2000/JARNet.

Shiqi Chen, Huajun Feng, Dexin Pan et al.

As the popularity of mobile photography continues to grow, considerable effort is being invested in the reconstruction of degraded images. Due to the spatial variation in optical aberrations, which cannot be avoided during the lens design process, recent commercial cameras have shifted some of these correction tasks from optical design to postprocessing systems. However, without engaging with the optical parameters, these systems only achieve limited correction for aberrations.In this work, we propose a practical method for recovering the degradation caused by optical aberrations. Specifically, we establish an imaging simulation system based on our proposed optical point spread function model. Given the optical parameters of the camera, it generates the imaging results of these specific devices. To perform the restoration, we design a spatial-adaptive network model on synthetic data pairs generated by the imaging simulation system, eliminating the overhead of capturing training data by a large amount of shooting and registration. Moreover, we comprehensively evaluate the proposed method in simulations and experimentally with a customized digital-single-lens-reflex (DSLR) camera lens and HUAWEI HONOR 20, respectively. The experiments demonstrate that our solution successfully removes spatially variant blur and color dispersion. When compared with the state-of-the-art deblur methods, the proposed approach achieves better results with a lower computational overhead. Moreover, the reconstruction technique does not introduce artificial texture and is convenient to transfer to current commercial cameras. Project Page: \url{https://github.com/TanGeeGo/ImagingSimulation}.

Shiqi Chen, Miao Xiong, Junteng Liu et al.

Large language models (LLMs) frequently hallucinate and produce factual errors, yet our understanding of why they make these errors remains limited. In this study, we delve into the underlying mechanisms of LLM hallucinations from the perspective of inner representations, and discover a salient pattern associated with hallucinations: correct generations tend to have sharper context activations in the hidden states of the in-context tokens, compared to the incorrect ones. Leveraging this insight, we propose an entropy-based metric to quantify the ``sharpness'' among the in-context hidden states and incorporate it into the decoding process to formulate a constrained decoding approach. Experiments on various knowledge-seeking and hallucination benchmarks demonstrate our approach's consistent effectiveness, for example, achieving up to an 8.6 point improvement on TruthfulQA. We believe this study can improve our understanding of hallucinations and serve as a practical solution for hallucination mitigation.

Shiqi Chen, Jinghan Zhang, Tongyao Zhu et al.

Vision-Language Models (VLMs) combine visual perception with the general capabilities, such as reasoning, of Large Language Models (LLMs). However, the mechanisms by which these two abilities can be combined and contribute remain poorly understood. In this work, we explore to compose perception and reasoning through model merging that connects parameters of different models. Unlike previous works that often focus on merging models of the same kind, we propose merging models across modalities, enabling the incorporation of the reasoning capabilities of LLMs into VLMs. Through extensive experiments, we demonstrate that model merging offers a successful pathway to transfer reasoning abilities from LLMs to VLMs in a training-free manner. Moreover, we utilize the merged models to understand the internal mechanism of perception and reasoning and how merging affects it. We find that perception capabilities are predominantly encoded in the early layers of the model, whereas reasoning is largely facilitated by the middle-to-late layers. After merging, we observe that all layers begin to contribute to reasoning, whereas the distribution of perception abilities across layers remains largely unchanged. These observations shed light on the potential of model merging as a tool for multimodal integration and interpretation.

Lei Yu, Shiqi Chen, Hang Yuan et al.

With the rapid development of blockchain technology, smart contract security has become a critical challenge. Existing smart contract vulnerability detection methods face three main issues: (1) Insufficient quality of datasets, lacking detailed explanations and precise vulnerability locations. (2) Limited adaptability of large language models (LLMs) to the smart contract domain, as most LLMs are pre-trained on general text data but minimal smart contract-specific data. (3) Lack of high-quality explanations for detected vulnerabilities, as existing methods focus solely on detection without clear explanations. These limitations hinder detection performance and make it harder for developers to understand and fix vulnerabilities quickly, potentially leading to severe financial losses. To address these problems, we propose Smart-LLaMA, an advanced detection method based on the LLaMA language model. First, we construct a comprehensive dataset covering four vulnerability types with labels, detailed explanations, and precise vulnerability locations. Second, we introduce Smart Contract-Specific Continual Pre-Training, using raw smart contract data to enable the LLM to learn smart contract syntax and semantics, enhancing their domain adaptability. Furthermore, we propose Explanation-Guided Fine-Tuning, which fine-tunes the LLM using paired vulnerable code and explanations, enabling both vulnerability detection and reasoned explanations. We evaluate explanation quality through LLM and human evaluation, focusing on Correctness, Completeness, and Conciseness. Experimental results show that Smart-LLaMA outperforms state-of-the-art baselines, with average improvements of 6.49% in F1 score and 3.78% in accuracy, while providing reliable explanations.

Shiqi Chen, Yuhang Li, Hanlong Chen et al.

Generative models cover various application areas, including image, video and music synthesis, natural language processing, and molecular design, among many others. As digital generative models become larger, scalable inference in a fast and energy-efficient manner becomes a challenge. Here, we present optical generative models inspired by diffusion models, where a shallow and fast digital encoder first maps random noise into phase patterns that serve as optical generative seeds for a desired data distribution; a jointly-trained free-space-based reconfigurable decoder all-optically processes these generative seeds to create novel images (never seen before) following the target data distribution. Except for the illumination power and the random seed generation through a shallow encoder, these optical generative models do not consume computing power during the synthesis of novel images. We report the optical generation of monochrome and multi-color novel images of handwritten digits, fashion products, butterflies, and human faces, following the data distributions of MNIST, Fashion MNIST, Butterflies-100, and Celeb-A datasets, respectively, achieving an overall performance comparable to digital neural network-based generative models. To experimentally demonstrate optical generative models, we used visible light to generate, in a snapshot, novel images of handwritten digits and fashion products. These optical generative models might pave the way for energy-efficient, scalable and rapid inference tasks, further exploiting the potentials of optics and photonics for artificial intelligence-generated content.

Longxu Dou, Qian Liu, Fan Zhou et al.

Sailor2 is a family of cutting-edge multilingual language models for South-East Asian (SEA) languages, available in 1B, 8B, and 20B sizes to suit diverse applications. Building on Qwen2.5, Sailor2 undergoes continuous pre-training on 500B tokens (400B SEA-specific and 100B replay tokens) to support 13 SEA languages while retaining proficiency in Chinese and English. Sailor2-20B model achieves a 50-50 win rate against GPT-4o across SEA languages. We also deliver a comprehensive cookbook on how to develop the multilingual model in an efficient manner, including five key aspects: data curation, pre-training, post-training, model customization and evaluation. We hope that Sailor2 model (Apache 2.0 license) will drive language development in the SEA region, and Sailor2 cookbook will inspire researchers to build more inclusive LLMs for other under-served languages.

Xiangyu Zhao, Zhiwang Zhou, Wenlong Zhang et al.

The Earth's weather system encompasses intricate weather data modalities and diverse weather understanding tasks, which hold significant value to human life. Existing data-driven models focus on single weather understanding tasks (e.g., weather forecasting). Although these models have achieved promising results, they fail to tackle various complex tasks within a single and unified model. Moreover, the paradigm that relies on limited real observations for a single scenario hinders the model's performance upper bound. In response to these limitations, we draw inspiration from the in-context learning paradigm employed in state-of-the-art visual foundation models and large language models. In this paper, we introduce the first generalist weather foundation model (WeatherGFM), designed to address a wide spectrum of weather understanding tasks in a unified manner. More specifically, we initially unify the representation and definition of the diverse weather understanding tasks. Subsequently, we devised weather prompt formats to manage different weather data modalities, namely single, multiple, and temporal modalities. Finally, we adopt a visual prompting question-answering paradigm for the training of unified weather understanding tasks. Extensive experiments indicate that our WeatherGFM can effectively handle up to ten weather understanding tasks, including weather forecasting, super-resolution, weather image translation, and post-processing. Our method also showcases generalization ability on unseen tasks.

Xuming He, Zhiwang Zhou, Wenlong Zhang et al.

Weather radar data synthesis can fill in data for areas where ground observations are missing. Existing methods often employ reconstruction-based approaches with MSE loss to reconstruct radar data from satellite observation. However, such methods lead to over-smoothing, which hinders the generation of high-frequency details or high-value observation areas associated with convective weather. To address this issue, we propose a two-stage diffusion-based method called DiffSR. We first pre-train a reconstruction model on global-scale data to obtain radar estimation and then synthesize radar reflectivity by combining radar estimation results with satellite data as conditions for the diffusion model. Extensive experiments show that our method achieves state-of-the-art (SOTA) results, demonstrating the ability to generate high-frequency details and high-value areas.

Ruochen Zhou, Minrui Xu, Shiqi Chen et al.

There has been a growing interest in enhancing the mathematical problem-solving (MPS) capabilities of large language models. While the majority of research efforts concentrate on creating specialized models to solve mathematical problems, it remains unknown how learning mathematical problem-solving generalizes to help develop other reasoning abilities. In this paper, we present an empirical investigation into the generalization potential of various MPS training approaches, such as continual pretraining, instruction tuning, and rule-based reinforcement learning across various data sources, including both short and long chain-of-thought (CoT) samples. Evaluation on 5 mathematical and 8 general reasoning benchmarks show that continual pretraining on math text is able to generalize to general reasoning tasks to some extent. In constrast, instruction tuning on conventional, short MPS samples provides limited benefits and, in many cases, even impairs generalization performance. Notably, training with long CoT responses for MPS samples and incorporating rule-based reinforcement learning on MPS queries exhibit distinct behavior, significantly enhancing generalization by extending the model's reasoning processes into other domains. These results suggest that traditional approaches to learning MPS with short reasoning chains largely fail to achieve robust generalization. However, the emerging paradigm of longer reasoning chains, coupled with self-reflection, offers a promising direction for improving generalized reasoning abilities through learning from specialized domains.

Shiqi Chen, Tongyao Zhu, Zian Wang et al.

Large Language Models (LLMs) as agents often struggle in out-of-distribution (OOD) scenarios. Real-world environments are complex and dynamic, governed by task-specific rules and stochasticity, which makes it difficult for LLMs to ground their internal knowledge in those dynamics. Under such OOD conditions, vanilla RL training often fails to scale; we observe Pass@k--the probability that at least one of (k) sampled trajectories succeeds--drops markedly across training steps, indicating brittle exploration and limited generalization. Inspired by model-based reinforcement learning, we hypothesize that equipping LLM agents with an internal world model can better align reasoning with environmental dynamics and improve decision-making. We show how to encode this world model by decomposing it into two components: state representation and transition modeling. Building on this, we introduce SPA, a simple reinforcement learning framework that cold-starts the policy via a Self-Play supervised finetuning (SFT) stage to learn the world model by interacting with the environment, then uses it to simulate future states prior to policy optimization. This simple initialization outperforms the online world-modeling baseline and greatly boosts the RL-based agent training performance. Experiments across diverse environments like Sokoban, FrozenLake, and Sudoku show that our approach significantly improves performance. For example, SPA boosts the Sokoban success rate from 25.6% to 59.8% and raises the FrozenLake score from 22.1% to 70.9% for the Qwen2.5-1.5B-Instruct model.

Zhengyu Chen, Siqi Wang, Teng Xiao et al. · pku

Traditional scaling laws in natural language processing suggest that increasing model size and training data enhances performance. However, recent studies reveal deviations, particularly in large language models, where performance improvements decelerate, which is a phenomenon known as sub-scaling. This paper revisits these scaling laws by examining the impact of data quality and training strategies on model performance. Through extensive empirical analysis of over 400 models, we identify high data density and non-optimal resource allocation as key factors contributing to sub-scaling. High data density leads to diminishing returns due to redundant information, while optimal resource allocation is crucial for sustained performance improvements. We propose a sub-optimal scaling law that better predicts performance in sub-scaling regimes, highlighting the importance of data quality and diversity.

Yuhang Li, Shiqi Chen, Tingyu Gong et al.

Optical computing holds promise for high-speed, energy-efficient information processing, with diffractive optical networks emerging as a flexible platform for implementing task-specific transformations. A challenge, however, is the effective optimization and alignment of the diffractive layers, which is hindered by the difficulty of accurately modeling physical systems with their inherent hardware imperfections, noise, and misalignments. While existing in situ optimization methods offer the advantage of direct training on the physical system without explicit system modeling, they are often limited by slow convergence and unstable performance due to inefficient use of limited measurement data. Here, we introduce a model-free reinforcement learning approach utilizing Proximal Policy Optimization (PPO) for the in situ training of diffractive optical processors. PPO efficiently reuses in situ measurement data and constrains policy updates to ensure more stable and faster convergence. We experimentally validated our method across a range of in situ learning tasks, including targeted energy focusing through a random diffuser, holographic image generation, aberration correction, and optical image classification, demonstrating in each task better convergence and performance. Our strategy operates directly on the physical system and naturally accounts for unknown real-world imperfections, eliminating the need for prior system knowledge or modeling. By enabling faster and more accurate training under realistic experimental constraints, this in situ reinforcement learning approach could offer a scalable framework for various optical and physical systems governed by complex, feedback-driven dynamics.

Zhengyue Zhuge, Jiahui Xu, Shiqi Chen et al.

Snapshot hyperspectral imaging systems acquire spectral data cubes through compressed sensing. Recently, diffractive snapshot spectral imaging (DSSI) methods have attracted significant attention. While various optical designs and improvements continue to emerge, research on reconstruction algorithms remains limited. Although numerous networks and deep unfolding methods have been applied on similar tasks, they are not fully compatible with DSSI systems because of their distinct optical encoding mechanism. In this paper, we propose an efficient deep unfolding framework for diffractive systems, termed diffractive deep unfolding (DDU). Specifically, we derive an analytical solution for the data fidelity term in DSSI, ensuring both the efficiency and the effectiveness during the iterative reconstruction process. Given the severely ill-posed nature of the problem, we employ a network-based initialization strategy rather than non-learning-based methods or linear layers, leading to enhanced stability and performance. Our framework demonstrates strong compatibility with existing state-of-the-art (SOTA) models, which effectively address the initialization and prior subproblem. Extensive experiments validate the superiority of the proposed DDU framework, showcasing improved performance while maintaining comparable parameter counts and computational complexity. These results suggest that DDU provides a solid foundation for future unfolding-based methods in DSSI.

Yuhang Li, Shiqi Chen, Bijie Bai et al.

We introduce an all-optical system, termed the "lying mirror", to hide input information by transforming it into misleading, ordinary-looking patterns that effectively camouflage the underlying image data and deceive the observers. This misleading transformation is achieved through passive light-matter interactions of the incident light with an optimized structured diffractive surface, enabling the optical concealment of any form of secret input data without any digital computing. These lying mirror designs were shown to camouflage different types of input image data, exhibiting robustness against a range of adversarial manipulations, including random image noise as well as unknown, random rotations, shifts, and scaling of the object features. The feasibility of the lying mirror concept was also validated experimentally using a structured micro-mirror array along with multi-wavelength illumination at 480, 550 and 600 nm, covering the blue, green and red image channels. This framework showcases the power of structured diffractive surfaces for visual information processing and might find various applications in defense, security and entertainment.

Lingyun Zhang, Bin jin, Gaojian Ge et al.

Human priors play a crucial role in efficiently utilizing data in deep learning. However, with the development of large language models (LLMs), there is an increasing emphasis on scaling both model size and data volume, which often diminishes the importance of human priors in data construction. Influenced by these trends, existing Small Language Models (SLMs) mainly rely on web-scraped large-scale training data, neglecting the proper incorporation of human priors. This oversight limits the training efficiency of language models in resource-constrained settings. In this paper, we propose a principle to leverage human priors for data construction. This principle emphasizes achieving high-performance SLMs by training on a concise dataset that accommodates both semantic diversity and data quality consistency, while avoiding benchmark data leakage. Following this principle, we train an SLM named HARE-1.1B. Extensive experiments on large-scale benchmark datasets demonstrate that HARE-1.1B performs favorably against state-of-the-art SLMs, validating the effectiveness of the proposed principle. Additionally, this provides new insights into efficient language model training in resource-constrained environments from the view of human priors.

Shiqi Chen, Siyang Gao, Junxian He

Detecting factual errors in summaries has been an important and challenging subject in summarization research. Inspired by the emergent ability of large language models (LLMs), we explore evaluating factual consistency of summaries by directly prompting LLMs. We present a comprehensive empirical study to assess the ability of LLMs as factual consistency evaluators, which consists of (1) analyzing different LLMs such as the GPT model series and Flan-T5; (2) investigating a variety of prompting methods including vanilla prompting, chain-of-thought prompting, and a sentence-by-sentence prompting method to tackle long summaries; and (3) evaluating on diverse summaries generated by multiple summarization systems, ranging from pre-transformer methods to SOTA pretrained models. Our experiments demonstrate that prompting LLMs is able to outperform the previous best factuality systems in all settings, by up to 12.2 absolute points in terms of the binary classification accuracy on inconsistency detection.

Shiqi Chen, Jinwen Zhou, Menghao Li et al.

In digital images, the performance of optical aberration is a multivariate degradation, where the spectral of the scene, the lens imperfections, and the field of view together contribute to the results. Besides eliminating it at the hardware level, the post-processing system, which utilizes various prior information, is significant for correction. However, due to the content differences among priors, the pipeline that aligns these factors shows limited efficiency and unoptimized restoration. Here, we propose a prior quantization model to correct the optical aberrations in image processing systems. To integrate these messages, we encode various priors into a latent space and quantify them by the learnable codebooks. After quantization, the prior codes are fused with the image restoration branch to realize targeted optical aberration correction. Comprehensive experiments demonstrate the flexibility of the proposed method and validate its potential to accomplish targeted restoration for a specific camera. Furthermore, our model promises to analyze the correlation between the various priors and the optical aberration of devices, which is helpful for joint soft-hardware design.

Shiqi Chen, Ting Lin, Huajun Feng et al.

Correcting the optical aberrations and the manufacturing deviations of cameras is a challenging task. Due to the limitation on volume and the demand for mass production, existing mobile terminals cannot rectify optical degradation. In this work, we systematically construct the perturbed lens system model to illustrate the relationship between the deviated system parameters and the spatial frequency response measured from photographs. To further address this issue, an optimization framework is proposed based on this model to build proxy cameras from the machining samples' SFRs. Engaging with the proxy cameras, we synthetic data pairs, which encode the optical aberrations and the random manufacturing biases, for training the learning-based algorithms. In correcting aberration, although promising results have been shown recently with convolutional neural networks, they are hard to generalize to stochastic machining biases. Therefore, we propose a dilated Omni-dimensional dynamic convolution and implement it in post-processing to account for the manufacturing degradation. Extensive experiments which evaluate multiple samples of two representative devices demonstrate that the proposed optimization framework accurately constructs the proxy camera. And the dynamic processing model is well-adapted to manufacturing deviations of different cameras, realizing perfect computational photography. The evaluation shows that the proposed method bridges the gap between optical design, system machining, and post-processing pipeline, shedding light on the joint of image signal reception (lens and sensor) and image signal processing.

Shiqi Chen, Zhengyu Chen, Donglin Wang

Meta Reinforcement Learning (MRL) enables an agent to learn from a limited number of past trajectories and extrapolate to a new task. In this paper, we attempt to improve the robustness of MRL. We build upon model-agnostic meta-learning (MAML) and propose a novel method to generate adversarial samples for MRL by using Generative Adversarial Network (GAN). That allows us to enhance the robustness of MRL to adversal attacks by leveraging these attacks during meta training process.