arXiv subjects

Yanjing Li

Publications and source records attributed to Yanjing Li.

At least 19 recordsLinked to original sources

arXiv2026

Kalman Meets Curriculum: Efficient Dynamic Prompt Selection for Adaptive RL Finetuning

Reinforcement learning (RL) finetuning significantly enhances the reasoning capabilities of large language models (LLMs), yet its effectiveness critically depends on selecting prompts of appropriate difficulty for the current policy. This is challenging because prompt difficulty evolves throughout training. Existing online methods therefore face a trade-off: evaluation-based approaches are accurate but expensive, while prediction-based approaches are efficient but typically assume stationary difficulty, making them ill-suited to RL's non-stationary training dynamics. To address these issues, we propose a Kalman-Guided Prompt Selection method (KGPS), which reformulates prompt selection as a dynamic state estimation problem rather than static difficulty prediction. KGPS models each prompt's latent success rate in logit space using a linear-Gaussian state-space model, with process noise coupled to the magnitude of policy updates so that uncertainty increases when the policy changes more substantially. A Kalman filter then maintains a calibrated Gaussian posterior over prompt difficulty, and prompts are selected by maximizing a posterior-expected training utility that favors intermediate-difficulty prompts while naturally revisiting uncertain ones. The resulting procedure is adaptive to policy drift and requires no additional rollouts beyond standard policy training. Extensive experiments across mathematics, planning, and geometry reasoning benchmarks, as well as multiple RL algorithms, show that KGPS consistently improves both final accuracy and rollout efficiency over strong baselines, establishing state-of-the-art performance among online prompt selection methods. For example, on DeepSeek-R1-Distill-7B, KGPS uses 83% fewer rollouts than DS while even improving the average performance by 0.12 point across six math reasoning benchmarks.

cs.LG↗

arXiv2026

Context Over Compute Human-in-the-Loop Outperforms Iterative Chain-of-Thought Prompting in Interview Answer Quality

Behavioral interview evaluation using large language models presents unique challenges that require structured assessment, realistic interviewer behavior simulation, and pedagogical value for candidate training. We investigate chain of thought prompting for interview answer evaluation and improvement through two controlled experiments with 50 behavioral interview question and answer pairs. Our contributions are threefold. First, we provide a quantitative comparison between human in the loop and automated chain of thought improvement. Using a within subject paired design with n equals 50, both approaches show positive rating improvements. The human in the loop approach provides significant training benefits. Confidence improves from 3.16 to 4.16 (p less than 0.001) and authenticity improves from 2.94 to 4.53 (p less than 0.001, Cohen's d is 3.21). The human in the loop method also requires five times fewer iterations (1.0 versus 5.0, p less than 0.001) and achieves full personal detail integration. Second, we analyze convergence behavior. Both methods converge rapidly with mean iterations below one, with the human in the loop approach achieving a 100 percent success rate compared to 84 percent for automated approaches among initially weak answers (Cohen's h is 0.82, large effect). Additional iterations provide diminishing returns, indicating that the primary limitation is context availability rather than computational resources. Third, we propose an adversarial challenging mechanism based on a negativity bias model, named bar raiser, to simulate realistic interviewer behavior, although quantitative validation remains future work. Our findings demonstrate that while chain of thought prompting provides a useful foundation for interview evaluation, domain specific enhancements and context aware approach selection are essential for realistic and pedagogically valuable results.

cs.CL↗

arXiv2026

SURGE: Surrogate Gradient Adaptation in Binary Neural Networks

The training of Binary Neural Networks (BNNs) is fundamentally based on gradient approximation for non-differentiable binarization operations (e.g., sign function). However, prevailing methods including the Straight-Through Estimator (STE) and its improved variants, rely on hand-crafted designs that suffer from gradient mismatch problem and information loss induced by fixed-range gradient clipping. To address this, we propose SURrogate GradiEnt Adaptation (SURGE), a novel learnable gradient compensation framework with theoretical grounding. SURGE mitigates gradient mismatch through auxiliary backpropagation. Specifically, we design a Dual-Path Gradient Compensator (DPGC) that constructs a parallel full-precision auxiliary branch for each binarized layer, decoupling gradient flow via output decomposition during backpropagation. DPGC enables bias-reduced gradient estimation by leveraging the full-precision branch to estimate components beyond STE's first-order approximation. To further enhance training stability, we introduce an Adaptive Gradient Scaler (AGS) based on an optimal scale factor to dynamically balance inter-branch gradient contributions via norm-based scaling. Experiments on image classification, object detection, and language understanding tasks demonstrate that SURGE performs best over state-of-the-art methods.

cs.LG↗

arXiv2026

Design and Implementation of BNN-Based Object Detection on FPGA

This paper implements a Binary Neural Network (BNN) based YOLOv3-tiny-like object detector on a low-cost FPGA. The network takes 320*320*3 RGB images as input. Its main convolution layers use 1-bit weights and 8-bit activations, while Conv1 and the final detection head use fixed-point standard convolutions. From the trained ONNX model, weights, biases, and quantization parameters are extracted, converted to fixed point, packed into COE files, and stored in Vivado BRAM ROMs. The hardware is written fully in Verilog RTL and includes padding, line buffering, binary convolution, quantization post-processing, max pooling, and detection-head computation. For layers where Mul_prev is indexed by input channel and Div_current by output channel, Mul_prev is fused in-to the BNN PE so that channel-wise compensation is applied during accumulation. On VOC, the model obtains 39.6% mAP50 with 0.098 GFLOPs and 0.74 M parameters. RTL simulation shows that the final raw detection output reaches a correlation coefficient of 0.999964 and a mean absolute error of 0.020027 against the corresponding ONNX node.

cs.AR↗

arXiv2026

GenOpticalFlow: A Generative Approach to Unsupervised Optical Flow Learning

Optical flow estimation is a fundamental problem in computer vision, yet the reliance on expensive ground-truth annotations limits the scalability of supervised approaches. Although unsupervised and semi-supervised methods alleviate this issue, they often suffer from unreliable supervision signals based on brightness constancy and smoothness assumptions, leading to inaccurate motion estimation in complex real-world scenarios. To overcome these limitations, we introduce \textbf{\modelname}, a novel framework that synthesizes large-scale, perfectly aligned frame--flow data pairs for supervised optical flow training without human annotations. Specifically, our method leverages a pre-trained depth estimation network to generate pseudo optical flows, which serve as conditioning inputs for a next-frame generation model trained to produce high-fidelity, pixel-aligned subsequent frames. This process enables the creation of abundant, high-quality synthetic data with precise motion correspondence. Furthermore, we propose an \textit{inconsistent pixel filtering} strategy that identifies and removes unreliable pixels in generated frames, effectively enhancing fine-tuning performance on real-world datasets. Extensive experiments on KITTI2012, KITTI2015, and Sintel demonstrate that \textbf{\modelname} achieves competitive or superior results compared to existing unsupervised and semi-supervised approaches, highlighting its potential as a scalable and annotation-free solution for optical flow learning. We will release our code upon acceptance.

cs.CV↗

arXiv2026

Rigidity-Aware Geometric Pretraining for Protein Design and Conformational Ensembles

Generative models have recently advanced $\textit{de novo}$ protein design by learning the statistical regularities of natural structures. However, current approaches face three key limitations: (1) Existing methods cannot jointly learn protein geometry and design tasks, where pretraining can be a solution; (2) Current pretraining methods mostly rely on local, non-rigid atomic representations for property prediction downstream tasks, limiting global geometric understanding for protein generation tasks; and (3) Existing approaches have yet to effectively model the rich dynamic and conformational information of protein structures. To overcome these issues, we introduce $\textbf{RigidSSL}$ ($\textit{Rigidity-Aware Self-Supervised Learning}$), a geometric pretraining framework that front-loads geometry learning prior to generative finetuning. Phase I (RigidSSL-Perturb) learns geometric priors from 432K structures from the AlphaFold Protein Structure Database with simulated perturbations. Phase II (RigidSSL-MD) refines these representations on 1.3K molecular dynamics trajectories to capture physically realistic transitions. Underpinning both phases is a bi-directional, rigidity-aware flow matching objective that jointly optimizes translational and rotational dynamics to maximize mutual information between conformations. Empirically, RigidSSL variants improve designability by up to 43% while enhancing novelty and diversity in unconditional generation. Furthermore, RigidSSL-Perturb improves the success rate by 5.8% in zero-shot motif scaffolding and RigidSSL-MD captures more biophysically realistic conformational ensembles in G protein-coupled receptor modeling.

cs.LG↗

arXiv2026

Unbiased Dynamic Pruning for Efficient Group-Based Policy Optimization

Group Relative Policy Optimization (GRPO) effectively scales LLM reasoning but incurs prohibitive computational costs due to its extensive group-based sampling requirement. While recent selective data utilization methods can mitigate this overhead, they could induce estimation bias by altering the underlying sampling distribution, compromising theoretical rigor and convergence behavior. To address this limitation, we propose Dynamic Pruning Policy Optimization (DPPO), a framework that enables dynamic pruning while preserving unbiased gradient estimation through importance sampling-based correction. By incorporating mathematically derived rescaling factors, DPPO significantly accelerates GRPO training without altering the optimization objective of the full-batch baseline. Furthermore, to mitigate the data sparsity induced by pruning, we introduce Dense Prompt Packing, a window-based greedy strategy that maximizes valid token density and hardware utilization. Extensive experiments demonstrate that DPPO consistently accelerates training across diverse models and benchmarks. For instance, on Qwen3-4B trained on MATH, DPPO achieves 2.37$\times$ training speedup and outperforms GRPO by 3.36% in average accuracy across six mathematical reasoning benchmarks.

cs.LG↗

arXiv2025

SONIQ: System-Optimized Noise-Injected Ultra-Low-Precision Quantization with Full-Precision Parity

Ultra-low-precision inference can sharply reduce memory and latency but often degrades accuracy and relies on specialized hardware. We present SONIQ, a system-optimized, noise-injected quantization framework that learns per-channel mixed precision for both weights and activations while training under the same rules used at inference. By injecting hardware-calibrated quantization noise during training, SONIQ steers models toward the discrete arithmetic used at deployment -- without bespoke runtimes. Across CNNs and Transformers, SONIQ achieves up to 16x and 7x compression, respectively, while matching or exceeding full-precision accuracy. Measured end-to-end, SONIQ delivers up to 7.3x CPU speedup over strong INT8 baselines and up to 6.3x (vector units) / 2.8x (tensor cores) GPU speedup relative to FP16. A practical outcome is that two per-channel precision levels -- one in the 1--4-bit range and one in the 4--8-bit range -- suffice in practice; at inference, each channel selects one of the two, keeping kernels simple and fast. To our knowledge, SONIQ is the first framework to reach or surpass full-precision accuracy under ultra-low (1--4 bits per parameter) regimes while remaining deployable on commodity hardware, narrowing the gap between quantization theory and practical, high-throughput inference.

cs.AR↗

arXiv2025

Squeeze10-LLM: Squeezing LLMs' Weights by 10 Times via a Staged Mixed-Precision Quantization Method

Deploying large language models (LLMs) is challenging due to their massive parameters and high computational costs. Ultra low-bit quantization can significantly reduce storage and accelerate inference, but extreme compression (i.e., mean bit-width <= 2) often leads to severe performance degradation. To address this, we propose Squeeze10-LLM, effectively "squeezing" 16-bit LLMs' weights by 10 times. Specifically, Squeeze10-LLM is a staged mixed-precision post-training quantization (PTQ) framework and achieves an average of 1.6 bits per weight by quantizing 80% of the weights to 1 bit and 20% to 4 bits. We introduce Squeeze10LLM with two key innovations: Post-Binarization Activation Robustness (PBAR) and Full Information Activation Supervision (FIAS). PBAR is a refined weight significance metric that accounts for the impact of quantization on activations, improving accuracy in low-bit settings. FIAS is a strategy that preserves full activation information during quantization to mitigate cumulative error propagation across layers. Experiments on LLaMA and LLaMA2 show that Squeeze10-LLM achieves state-of-the-art performance for sub-2bit weight-only quantization, improving average accuracy from 43% to 56% on six zero-shot classification tasks--a significant boost over existing PTQ methods.

cs.LG↗

arXiv2025

Uncertainty-Aware Gradient Stabilization for Small Object Detection

Despite advances in generic object detection, there remains a performance gap in detecting small objects compared to normal-scale objects. We reveal that conventional object localization methods suffer from gradient instability in small objects due to sharper loss curvature, leading to a convergence challenge. To address the issue, we propose Uncertainty-Aware Gradient Stabilization (UGS), a framework that reformulates object localization as a classification task to stabilize gradients. UGS quantizes continuous labels into interval non-uniform discrete representations. Under a classification-based objective, the localization branch generates bounded and confidence-driven gradients, mitigating instability. Furthermore, UGS integrates an uncertainty minimization (UM) loss that reduces prediction variance and an uncertainty-guided refinement (UR) module that identifies and refines high-uncertainty regions via perturbations. Evaluated on four benchmarks, UGS consistently improves anchor-based, anchor-free, and leading small object detectors. Especially, UGS enhances DINO-5scale by 2.6 AP on VisDrone, surpassing previous state-of-the-art results.

cs.CV↗

arXiv2025

A Text-guided Protein Design Framework

Current AI-assisted protein design mainly utilizes protein sequential and structural information. Meanwhile, there exists tremendous knowledge curated by humans in the text format describing proteins' high-level functionalities. Yet, whether the incorporation of such text data can help protein design tasks has not been explored. To bridge this gap, we propose ProteinDT, a multi-modal framework that leverages textual descriptions for protein design. ProteinDT consists of three subsequent steps: ProteinCLAP which aligns the representation of two modalities, a facilitator that generates the protein representation from the text modality, and a decoder that creates the protein sequences from the representation. To train ProteinDT, we construct a large dataset, SwissProtCLAP, with 441K text and protein pairs. We quantitatively verify the effectiveness of ProteinDT on three challenging tasks: (1) over 90% accuracy for text-guided protein generation; (2) best hit ratio on 12 zero-shot text-guided protein editing tasks; (3) superior performance on four out of six protein property prediction benchmarks.

cs.LG↗

arXiv2024

Predictive Window Decoding for Fault-Tolerant Quantum Programs

Real-time decoding is a key ingredient in future fault-tolerant quantum systems, yet many decoders are too slow to run in real time. Prior work has shown that parallel window decoding schemes can scalably meet throughput requirements in the presence of increasing decoding times, given enough classical resources. However, windowed decoding schemes require that some decoding tasks be delayed until others have completed, which can be problematic during time-sensitive operations such as T gate teleportation, leading to suboptimal program runtimes. To alleviate this, we introduce a speculative window decoding scheme. Taking inspiration from branch prediction in classical computer architecture our decoder utilizes a light-weight speculation step to predict data dependencies between adjacent decoding windows, allowing multiple layers of decoding tasks to be resolved simultaneously. Through a state-of-the-art compilation pipeline and a detailed simulator, we find that speculation reduces application runtimes by 40% on average compared to prior parallel window decoders.

quant-ph↗

arXiv2024

A Multi-Grained Symmetric Differential Equation Model for Learning Protein-Ligand Binding Dynamics

In drug discovery, molecular dynamics (MD) simulation for protein-ligand binding provides a powerful tool for predicting binding affinities, estimating transport properties, and exploring pocket sites. There has been a long history of improving the efficiency of MD simulations through better numerical methods and, more recently, by utilizing machine learning (ML) methods. Yet, challenges remain, such as accurate modeling of extended-timescale simulations. To address this issue, we propose NeuralMD, the first ML surrogate that can facilitate numerical MD and provide accurate simulations in protein-ligand binding dynamics. We propose a principled approach that incorporates a novel physics-informed multi-grained group symmetric framework. Specifically, we propose (1) the BindingNet model that satisfies group symmetry using vector frames and captures the multi-level protein-ligand interactions, and (2) an augmented neural differential equation solver that learns the trajectory under Newtonian mechanics. For the experiment, we design ten single-trajectory and three multi-trajectory binding simulation tasks. We demonstrate the efficiency and effectiveness of NeuralMD, achieving over 1K$\times$ speedup compared to standard numerical MD simulations. NeuralMD also outperforms all other ML approaches, achieving up to 15$\times$ reduction in reconstruction error and 70% increase in validity. Additionally, we qualitatively illustrate that the oscillations in the predicted trajectories align more closely with ground-truth dynamics than those of other machine-learning methods. We believe NeuralMD paves the foundation for a new research paradigm in simulating protein-ligand dynamics.

cs.LG↗

arXiv2024

CCExpert: Advancing MLLM Capability in Remote Sensing Change Captioning with Difference-Aware Integration and a Foundational Dataset

Remote Sensing Image Change Captioning (RSICC) aims to generate natural language descriptions of surface changes between multi-temporal remote sensing images, detailing the categories, locations, and dynamics of changed objects (e.g., additions or disappearances). Many current methods attempt to leverage the long-sequence understanding and reasoning capabilities of multimodal large language models (MLLMs) for this task. However, without comprehensive data support, these approaches often alter the essential feature transmission pathways of MLLMs, disrupting the intrinsic knowledge within the models and limiting their potential in RSICC. In this paper, we propose a novel model, CCExpert, based on a new, advanced multimodal large model framework. Firstly, we design a difference-aware integration module to capture multi-scale differences between bi-temporal images and incorporate them into the original image context, thereby enhancing the signal-to-noise ratio of differential features. Secondly, we constructed a high-quality, diversified dataset called CC-Foundation, containing 200,000 image pairs and 1.2 million captions, to provide substantial data support for continue pretraining in this domain. Lastly, we employed a three-stage progressive training process to ensure the deep integration of the difference-aware integration module with the pretrained MLLM. CCExpert achieved a notable performance of $S^*_m=81.80$ on the LEVIR-CC benchmark, significantly surpassing previous state-of-the-art methods. The code and part of the dataset will soon be open-sourced at https://github.com/Meize0729/CCExpert.

cs.CV↗

arXiv2024

P4Q: Learning to Prompt for Quantization in Visual-language Models

Large-scale pre-trained Vision-Language Models (VLMs) have gained prominence in various visual and multimodal tasks, yet the deployment of VLMs on downstream application platforms remains challenging due to their prohibitive requirements of training samples and computing resources. Fine-tuning and quantization of VLMs can substantially reduce the sample and computation costs, which are in urgent need. There are two prevailing paradigms in quantization, Quantization-Aware Training (QAT) can effectively quantize large-scale VLMs but incur a huge training cost, while low-bit Post-Training Quantization (PTQ) suffers from a notable performance drop. We propose a method that balances fine-tuning and quantization named ``Prompt for Quantization'' (P4Q), in which we design a lightweight architecture to leverage contrastive loss supervision to enhance the recognition performance of a PTQ model. Our method can effectively reduce the gap between image features and text features caused by low-bit quantization, based on learnable prompts to reorganize textual representations and a low-bit adapter to realign the distributions of image and text features. We also introduce a distillation loss based on cosine similarity predictions to distill the quantized model using a full-precision teacher. Extensive experimental results demonstrate that our P4Q method outperforms prior arts, even achieving comparable results to its full-precision counterparts. For instance, our 8-bit P4Q can theoretically compress the CLIP-ViT/B-32 by 4 $\times$ while achieving 66.94\% Top-1 accuracy, outperforming the learnable prompt fine-tuned full-precision model by 2.24\% with negligible additional parameters on the ImageNet dataset.

cs.CV↗

arXiv2024

UpDown: Programmable fine-grained Events for Scalable Performance on Irregular Applications

Applications with irregular data structures, data-dependent control flows and fine-grained data transfers (e.g., real-world graph computations) perform poorly on cache-based systems. We propose the UpDown accelerator that supports fine-grained execution with novel architecture mechanisms - lightweight threading, event-driven scheduling, efficient ultra-short threads, and split-transaction DRAM access with software-controlled synchronization. These hardware primitives support software programmable events, enabling high performance on diverse data structures and algorithms. UpDown also supports scalable performance; hardware replication enables programs to scale up performance. Evaluation results show UpDown's flexibility and scalability enable it to outperform CPUs on graph mining and analytics computations by up to 116-195x geomean speedup and more than 4x speedup over prior accelerators. We show that UpDown generates high memory parallelism (~4.6x over CPU) required for memory intensive graph computations. We present measurements that attribute the performance of UpDown (23x architectural advantage) to its individual architectural mechanisms. Finally, we also analyze the area and power cost of UpDown's mechanisms for software programmability.

cs.AR↗

arXiv2024

Drop-Connect as a Fault-Tolerance Approach for RRAM-based Deep Neural Network Accelerators

Resistive random-access memory (RRAM) is widely recognized as a promising emerging hardware platform for deep neural networks (DNNs). Yet, due to manufacturing limitations, current RRAM devices are highly susceptible to hardware defects, which poses a significant challenge to their practical applicability. In this paper, we present a machine learning technique that enables the deployment of defect-prone RRAM accelerators for DNN applications, without necessitating modifying the hardware, retraining of the neural network, or implementing additional detection circuitry/logic. The key idea involves incorporating a drop-connect inspired approach during the training phase of a DNN, where random subsets of weights are selected to emulate fault effects (e.g., set to zero to mimic stuck-at-1 faults), thereby equipping the DNN with the ability to learn and adapt to RRAM defects with the corresponding fault rates. Our results demonstrate the viability of the drop-connect approach, coupled with various algorithm and system-level design and trade-off considerations. We show that, even in the presence of high defect rates (e.g., up to 30%), the degradation of DNN accuracy can be as low as less than 1% compared to that of the fault-free version, while incurring minimal system-level runtime/energy costs.

cs.ET↗

arXiv2023

YFlows: Systematic Dataflow Exploration and Code Generation for Efficient Neural Network Inference using SIMD Architectures on CPUs

We address the challenges associated with deploying neural networks on CPUs, with a particular focus on minimizing inference time while maintaining accuracy. Our novel approach is to use the dataflow (i.e., computation order) of a neural network to explore data reuse opportunities using heuristic-guided analysis and a code generation framework, which enables exploration of various Single Instruction, Multiple Data (SIMD) implementations to achieve optimized neural network execution. Our results demonstrate that the dataflow that keeps outputs in SIMD registers while also maximizing both input and weight reuse consistently yields the best performance for a wide variety of inference workloads, achieving up to 3x speedup for 8-bit neural networks, and up to 4.8x speedup for binary neural networks, respectively, over the optimized implementations of neural networks today.

cs.AR↗