arXiv subjects

Jianfeng Cai

Publications and source records attributed to Jianfeng Cai.

16 recordsLinked to original sources

arXiv2026

Kimi K3: Open Frontier Intelligence

We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is built on Kimi Delta Attention and Attention Residuals, which improve information flow across sequence length and model depth. Together with Stable LatentMoE, which effectively activates 16 of 896 routed experts per token, and refined training and data recipes, these advances yield an approximately 2.5x improvement in overall scaling efficiency over Kimi K2. Post-training highlights reinforcement learning across general, agentic, and coding domains and multiple reasoning-effort levels, enabling compositional generalization and robust long-horizon execution. At 2.8T scale, Kimi K3 is supported by infrastructure advances in multiple areas: algorithm-system co-design for KDA, perfectly balanced expert-parallel training with efficient memory management, million-token agentic RL with persistent rollout and sandbox states, and deployment innovations. Extensive evaluations show that Kimi K3 achieves frontier-level performance across long-horizon coding, agentic, knowledge, reasoning, and vision tasks. While its overall performance still trails the most powerful proprietary models, namely Claude Fable 5 and GPT-5.6 Sol, Kimi K3 consistently outperforms other open and proprietary models evaluated in our suite. We release the full Kimi K3 model weights to facilitate future research and accelerate the broader deployment and adoption of frontier intelligence.

cs.CL↗

arXiv2026

Thermodynamics of stacking faults and phase stability in cobalt alloys: A combined computational and experimental study

Stacking fault energy dictates phase stability and deformation behavior in Co alloys and WC-Co cemented carbides, yet a quantitative assessment of alloying effects at finite temperatures remains poorly established. By integrating first-principles thermodynamics with microstructural characterization, we provide a rigorous evaluation of these influences across atomic and macroscopic scales. We show that stacking fault energetics at 0K for transition metal solutes are primarily governed by atomic misfit volume. While 4d and 5d elements follow a consistent linear trend, specific 3d solutes exhibit significant deviations due to non-negligible magnetic contributions. By incorporating phonon, electronic, longitudinal spin-fluctuation, and magnetic free-energy contributions, the model accurately captures the fcc-hcp transformation and quantifies how diverse solutes modulate the phase landscape. We demonstrate that V, Ni, Fe, Mo, and W lower the transformation temperature by stabilizing fcc phase, while Cr and C exhibit the opposite effect, consistent with experimental phase diagrams. Furthermore, microscopic analysis confirms that higher W content dissolved in the Co suppresses stacking-fault formation by elevating the stacking fault energy at finite temperatures. This work clarifies the physical mechanisms by which alloying regulates stacking fault energy and phase stability in Co-based systems, providing guidance for the design of Co-based alloys and WC-Co cemented carbides.

cond-mat.mtrl-sci↗

arXiv2026

Weighted Riemannian Optimization for Solving Quadratic Equations from Gaussian Magnitude Measurements

This paper explores the problem of generalized phase retrieval, which involves reconstructing a length-$n$ signal $\bm{x}$ from its $m$ phaseless samples $y_k = \left|\langle \bm{a}_k,\bm{x}\rangle\right|^2$, where $k = 1,2,...,m$, and $\bm{a}_k$ are the measurement vectors. This problem can be reformulated into recovering a positive semidefinite rank-$1$ matrix $\bm{X}=\bm{x}\bm{x}^*$ from linear samples $\bm{y}=\mathcal{A}(\bm{X})\in\mathbb{R}^m$, thereby requiring us to find a rank-$1$ solution of the linear equations. We demonstrate that several existing phase retrieval algorithms, including Wirtinger Flow (WF) and the canonical Riemannian gradient descent (RGD), actually solve the least-squares fitting of this linear equation on the Riemannian manifold of rank-$1$ matrices, but utilize different metrics on this manifold. Nevertheless, these metrics only allow for a stable and far-apart-from-isometric embedding of rank-$1$ matrices to $\mathbb{R}^m$ by $\mathcal{A}$, resulting in a linear convergence with a considerably large convergence factor. To expedite the convergence, we establish a new metric on the rank-$1$ matrix manifold that facilitates the nearly isometric embedding of rank-$1$ matrices into $\mathbb{R}^m$ through $\mathcal{A}$. A RGD algorithm under this new metric, termed Weighted RGD (WRGD), is proposed to tackle the phase retrieval problem. Owing to the near isometry, we prove that our WRGD algorithm, initialized by spectral methods, can linearly converge to the underlying signal $\bm{x}$ with a small convergence factor. Empirical experiments strongly validate the efficiency and resilience of our algorithms compared to the truncated Wirtinger Flow (TWF) algorithm and the canonical RGD algorithm.

cs.IT↗

arXiv2026

CodeContests-O: Powering LLMs via Feedback-Driven Iterative Test Case Generation

The rise of reasoning models necessitates large-scale verifiable data, for which programming tasks serve as an ideal source. However, while competitive programming platforms provide abundant problems and solutions, high-quality test cases for verification remain scarce. Existing approaches attempt to synthesize test cases using Large Language Models (LLMs), but rely solely on the model's intrinsic generation capabilities without external feedback, frequently resulting in insufficiently diverse cases. To address this limitation, we propose a $\textbf{Feedback-Driven Iterative Framework}$ for comprehensive test case construction. Specifically, our method leverages the LLM to generate initial test cases, executes them against known correct and incorrect solutions, and utilizes the failed results as feedback to guide the LLM in refining the test cases toward high fidelity and discriminability. We then apply this method to the CodeContests dataset to construct an optimized high-quality derivative, $\textbf{CodeContests-O}$. Evaluating against the entire pool of solutions ($1.1 \times 10^7$ in total), our dataset achieves an average True Positive Rate (TPR) of $89.37\%$ and True Negative Rate (TNR) of $90.89\%$, significantly outperforming the CodeContests and CodeContests+ by margins of $4.32\%$ and $9.37\%$, respectively. Furthermore, fine-tuning the Qwen2.5-7B model on CodeContests-O results in a $9.52\%$ improvement on LiveCodeBench (Pass@1). Experiments demonstrate the effectiveness of our framework and the quality of CodeContests-O. To support reproducibility and facilitate future research, we release the $\href{https://github.com/cai-jianfeng/CodeContests-O}{code}$ and $\href{https://huggingface.co/datasets/caijanfeng/CodeContests-O}{dataset}$.

cs.SE↗

arXiv2025

Remote Sensing Image Classification with Decoupled Knowledge Distillation

To address the challenges posed by the large number of parameters in existing remote sensing image classification models, which hinder deployment on resource-constrained devices, this paper proposes a lightweight classification method based on knowledge distillation. Specifically, G-GhostNet is adopted as the backbone network, leveraging feature reuse to reduce redundant parameters and significantly improve inference efficiency. In addition, a decoupled knowledge distillation strategy is employed, which separates target and non-target classes to effectively enhance classification accuracy. Experimental results on the RSOD and AID datasets demonstrate that, compared with the high-parameter VGG-16 model, the proposed method achieves nearly equivalent Top-1 accuracy while reducing the number of parameters by 6.24 times. This approach strikes an excellent balance between model size and classification performance, offering an efficient solution for deployment on resource-limited devices.

cs.CV↗

arXiv2025

Disentangling Length Bias In Preference Learning Via Response-Conditioned Modeling

Reinforcement Learning from Human Feedback (RLHF) has achieved considerable success in aligning large language models (LLMs) by modeling human preferences with a learnable reward model and employing a reinforcement learning algorithm to maximize the reward model's scores. However, these reward models are susceptible to exploitation through various superficial confounding factors, with length bias emerging as a particularly significant concern. Moreover, while the pronounced impact of length bias on preference modeling suggests that LLMs possess an inherent sensitivity to length perception, our preliminary investigations reveal that fine-tuned LLMs consistently struggle to adhere to explicit length instructions. To address these two limitations, we propose a novel framework wherein the reward model explicitly differentiates between human semantic preferences and response length requirements. Specifically, we introduce a $\textbf{R}$esponse-$\textbf{c}$onditioned $\textbf{B}$radley-$\textbf{T}$erry (Rc-BT) model that enhances the model's capability in length bias mitigating and length instruction following, through training on our augmented dataset. Furthermore, we propose the Rc-RM and Rc-DPO algorithm to leverage the Rc-BT model for reward modeling and direct policy optimization (DPO) of LLMs, simultaneously mitigating length bias and promoting adherence to length instructions. Extensive experiments across various foundational models and datasets demonstrate the effectiveness and generalizability of our approach.

cs.LG↗

arXiv2025

Mitigating Hallucination in VideoLLMs via Temporal-Aware Activation Engineering

Multimodal large language models (MLLMs) have achieved remarkable progress in video understanding.However, hallucination, where the model generates plausible yet incorrect outputs, persists as a significant and under-addressed challenge in the video domain. Among existing solutions, activation engineering has proven successful in mitigating hallucinations in LLMs and ImageLLMs, yet its applicability to VideoLLMs remains largely unexplored. In this work, we are the first to systematically investigate the effectiveness and underlying mechanisms of activation engineering for mitigating hallucinations in VideoLLMs. We initially conduct an investigation of the key factors affecting the performance of activation engineering and find that a model's sensitivity to hallucination depends on $\textbf{temporal variation}$ rather than task type. Moreover, selecting appropriate internal modules and dataset for activation engineering is critical for reducing hallucination. Guided by these findings, we propose a temporal-aware activation engineering framework for VideoLLMs, which adaptively identifies and manipulates hallucination-sensitive modules based on the temporal variation characteristic, substantially mitigating hallucinations without additional LLM fine-tuning. Experiments across multiple models and benchmarks demonstrate that our method markedly reduces hallucination in VideoLLMs, thereby validating the robustness of our findings.

cs.CV↗

arXiv2025

Bias Fitting to Mitigate Length Bias of Reward Model in RLHF

Reinforcement Learning from Human Feedback (RLHF) relies on reward models to align large language models with human preferences. However, RLHF often suffers from reward hacking, wherein policy learning exploits flaws in the trained reward model to maximize reward scores without genuinely aligning with human preferences. A significant example of such reward hacking is length bias, where reward models usually favor longer responses irrespective of actual response quality. Previous works on tackling length bias have notable limitations, these approaches either mitigate bias without characterizing the bias form, or simply assume a linear length-reward relation. To accurately model the intricate nature of length bias and facilitate more effective bias mitigation, we propose FiMi-RM (Bias Fitting to Mitigate Length Bias of Reward Model), a framework that autonomously learns and corrects underlying bias patterns. Our approach consists of three stages: First, we warm up by training a standard reward model which inherently contains length bias. Next, we deploy a lightweight fitting model to capture the non-linear relation between length and reward. Finally, we incorporate this learned relation into the reward model, effectively decoupling length from reward while preserving preference modeling capabilities. Experimental results demonstrate that FiMi-RM achieves a more balanced length-reward distribution. Furthermore, when applied to alignment algorithms such as Direct Preference Optimization (DPO) and Best-of-N (BoN), our debiased reward model improves length-controlled win rate and reduces verbosity without compromising its performance.

cs.LG↗

arXiv2025

Multi-Level Aware Preference Learning: Enhancing RLHF for Complex Multi-Instruction Tasks

RLHF has emerged as a predominant approach for aligning artificial intelligence systems with human preferences, demonstrating exceptional and measurable efficacy in instruction following tasks; however, it exhibits insufficient compliance capabilities when confronted with complex multi-instruction tasks. Conventional approaches rely heavily on human annotation or more sophisticated large language models, thereby introducing substantial resource expenditure or potential bias concerns. Meanwhile, alternative synthetic methods that augment standard preference datasets often compromise the model's semantic quality. Our research identifies a critical oversight in existing techniques, which predominantly focus on comparing responses while neglecting valuable latent signals embedded within prompt inputs, and which only focus on preference disparities at the intra-sample level, while neglecting to account for the inter-sample level preference differentials that exist among preference data. To leverage these previously neglected indicators, we propose a novel Multi-level Aware Preference Learning (MAPL) framework, capable of enhancing multi-instruction capabilities. Specifically, for any given response in original preference data pairs, we construct varied prompts with a preference relation under different conditions, in order to learn intra-sample level preference disparities. Furthermore, for any given original preference pair, we synthesize multi-instruction preference pairs to capture preference discrepancies at the inter-sample level. Building on the two datasets constructed above, we consequently devise two sophisticated training objective functions. Subsequently, our framework integrates seamlessly into both Reward Modeling and Direct Preference Optimization paradigms. Through rigorous evaluation across multiple benchmarks, we empirically validate the efficacy of our framework.

cs.AI↗

arXiv2025

Phase stabilization of In2Se3 by disordered Ni intercalation and its enhanced thermoelectrical performance

Van der Waals (vdW) layered materials have gained significant attention owing to their distinctive structure and unique properties. The weak interlayer bonding in vdW layered materials enables guest atom intercalation, allowing precise tuning of their physical and chemical properties. In this work, a ternary compound, NixIn2Se3 (x = 0-0.3), with Ni randomly occupying the interlayers of In2Se3, was synthesized via an intercalation route driven by electron injection. The intercalated Ni atoms act as anchor points within the interlayer of In2Se3, which effectively suppresses the phase transition of In2Se3 at elevated temperatures. Furthermore, the disordered Ni intercalation significantly enhanced the electrical conductivity of In2Se3 through electron injection, while reducing the thermal conductivity due to the interlayer phonon scattering, leading to an improved thermoelectric performance. For instance, the thermoelectric figure of merit (ZT) of Ni0.3In2Se3 increased by 86% (in-plane) and 222% (out-of-plane) compared to In2Se3 at 500 oC. These findings not only provide an effective strategy to enhance the performance of layered thermoelectric materials, but also demonstrate the potential of intercalation chemistry for expanding the application scope of van der Waals (vdW) layered materials.

cond-mat.mtrl-sci↗

arXiv2024

Heterogeneous Network Based Contrastive Learning Method for PolSAR Land Cover Classification

Polarimetric synthetic aperture radar (PolSAR) image interpretation is widely used in various fields. Recently, deep learning has made significant progress in PolSAR image classification. Supervised learning (SL) requires a large amount of labeled PolSAR data with high quality to achieve better performance, however, manually labeled data is insufficient. This causes the SL to fail into overfitting and degrades its generalization performance. Furthermore, the scattering confusion problem is also a significant challenge that attracts more attention. To solve these problems, this article proposes a Heterogeneous Network based Contrastive Learning method(HCLNet). It aims to learn high-level representation from unlabeled PolSAR data for few-shot classification according to multi-features and superpixels. Beyond the conventional CL, HCLNet introduces the heterogeneous architecture for the first time to utilize heterogeneous PolSAR features better. And it develops two easy-to-use plugins to narrow the domain gap between optics and PolSAR, including feature filter and superpixel-based instance discrimination, which the former is used to enhance the complementarity of multi-features, and the latter is used to increase the diversity of negative samples. Experiments demonstrate the superiority of HCLNet on three widely used PolSAR benchmark datasets compared with state-of-the-art methods. Ablation studies also verify the importance of each component. Besides, this work has implications for how to efficiently utilize the multi-features of PolSAR data to learn better high-level representation in CL and how to construct networks suitable for PolSAR data better.

cs.CV↗

arXiv2024

Flowmind2Digital: The First Comprehensive Flowmind Recognition and Conversion Approach

Flowcharts and mind maps, collectively known as flowmind, are vital in daily activities, with hand-drawn versions facilitating real-time collaboration. However, there's a growing need to digitize them for efficient processing. Automated conversion methods are essential to overcome manual conversion challenges. Existing sketch recognition methods face limitations in practical situations, being field-specific and lacking digital conversion steps. Our paper introduces the Flowmind2digital method and hdFlowmind dataset to address these challenges. Flowmind2digital, utilizing neural networks and keypoint detection, achieves a record 87.3% accuracy on our dataset, surpassing previous methods by 11.9%. The hdFlowmind dataset, comprising 1,776 annotated flowminds across 22 scenarios, outperforms existing datasets. Additionally, our experiments emphasize the importance of simple graphics, enhancing accuracy by 9.3%.

cs.CV↗

arXiv2021

Solving phase retrieval with random initial guess is nearly as good as by spectral initialization

The problem of recovering a signal $\mathbf{x}\in \mathbb{R}^n$ from a set of magnitude measurements $y_i=|\langle \mathbf{a}_i, \mathbf{x} \rangle |, \; i=1,\ldots,m$ is referred as phase retrieval, which has many applications in fields of physical sciences and engineering. In this paper we show that the smoothed amplitude flow model for phase retrieval has benign geometric structure under the optimal sampling complexity. In particular, we show that when the measurements $\mathbf{a}_i\in \mathbb{R}^n$ are Gaussian random vectors and the number of measurements $m\ge Cn$, our smoothed amplitude flow model has no spurious local minimizers with high probability, ie., the target solution $\mathbf{x}$ is the unique global minimizer (up to a global phase) and the loss function has a negative directional curvature around each saddle point. Due to this benign geometric landscape, the phase retrieval problem can be solved by the gradient descent algorithms without spectral initialization. Numerical experiments show that the gradient descent algorithm with random initialization performs well even comparing with state-of-the-art algorithms with spectral initialization in empirical success rate and convergence speed.

cs.IT↗

arXiv2019

Data-driven Method for 3D Axis-symmetric Object Reconstruction from Single Cone-beam Projection Data

In this paper we consider 3D axis-symmetric (AS) object reconstruction from single cone-beam x-ray projection data. Traditional x-ray CT fails to capture fleeting state of material due to the long time for data acquisition at all angles. Therefore, AS object is devised to investigate the instant deformation of material under pulse change of environment because single projection data is enough to reconstruct its inner structure. Previous reconstruction methods are layer by layer, and ignore the longitudinal tilt of x-ray paths. We propose a regularization method using adaptive tight frame to reconstruct the 3D AS object structure simultaneously. Alternating direction method is adopted to solve the proposed model. More importantly, a numerical algorithm is developed to compute imaging matrix. Experiments on simulation data verify the effectiveness of our method

math.NA↗

arXiv2019

A stochastic alternating minimizing method for sparse phase retrieval

Sparse phase retrieval plays an important role in many fields of applied science and thus attracts lots of attention. In this paper, we propose a \underline{sto}chastic alte\underline{r}nating \underline{m}inimizing method for \underline{sp}arse ph\underline{a}se \underline{r}etrieval (\textit{StormSpar}) algorithm which {emprically} is able to recover $n$-dimensional $s$-sparse signals from only $O(s\,\mathrm{log}\, n)$ number of measurements without a desired initial value required by many existing methods. In \textit{StormSpar}, the hard-thresholding pursuit (HTP) algorithm is employed to solve the sparse constraint least square sub-problems. The main competitive feature of \textit{StormSpar} is that it converges globally requiring optimal order of number of samples with random initialization. Extensive numerical experiments are given to validate the proposed algorithm.

stat.ML↗

arXiv2011

Sparse Recovery from Nonlinear Measurements with Applications in Bad Data Detection for Power Networks

In this paper, we consider the problem of sparse recovery from nonlinear measurements, which has applications in state estimation and bad data detection for power networks. An iterative mixed $\ell_1$ and $\ell_2$ convex program is used to estimate the true state by locally linearizing the nonlinear measurements. When the measurements are linear, through using the almost Euclidean property for a linear subspace, we derive a new performance bound for the state estimation error under sparse bad data and additive observation noise. As a byproduct, in this paper we provide sharp bounds on the almost Euclidean property of a linear subspace, using the "escape-through-the-mesh" theorem from geometric functional analysis. When the measurements are nonlinear, we give conditions under which the solution of the iterative algorithm converges to the true state even though the locally linearized measurements may not be the actual nonlinear measurements. We numerically evaluate our iterative convex programming approach to perform bad data detections in nonlinear electrical power networks problems. We are able to use semidefinite programming to verify the conditions for convergence of the proposed iterative sparse recovery algorithms from nonlinear measurements.

cs.IT↗