Tiancheng Li

Tiancheng Li, Juan M Corchado, Shudong Sun

We propose a novel consensus notion, called "partial consensus", for distributed GM-PHD (Gaussian mixture probability hypothesis density) fusion based on a peer-to-peer (P2P) sensor network, in which only highly-weighted posterior Gaussian components (GCs) are disseminated in the P2P communication for fusion while the insignificant GCs are not involved. The partial consensus does not only enjoy high efficiency in both network communication and local fusion computation, but also significantly reduces the affect of potential false data (clutter) to the filter, leading to increased signal-to-noise ratio at local sensors. Two "conservative" mixture reduction schemes are advocated for fusing the shared GCs in a fully distributed manner. One is given by pairwise averaging GCs between sensors based on Hungarian assignment and the other is merging close GCs based a new GM merging scheme. The proposed approaches have a close connection to the conservative fusion approaches known as covariance union and arithmetic mean density. In parallel, average consensus is sought on the cardinality distribution (namely the GM weight sum) among sensors. Simulations for tracking either a single target or multiple targets that simultaneously appear are presented based on a sensor network where each sensor operates a GM-PHD filter, in order to compare our approaches with the benchmark generalized covariance intersection approach. The results demonstrate that the partial, arithmetic average, consensus outperforms the complete, geometric average, consensus.

Tiancheng Li, Franz Hlawatsch

We propose a particle-based distributed PHD filter for tracking an unknown, time-varying number of targets. To reduce communication, the local PHD filters at neighboring sensors communicate Gaussian mixture (GM) parameters. In contrast to most existing distributed PHD filters, our filter employs an `arithmetic average' fusion. For particles--GM conversion, we use a method that avoids particle clustering and enables a significance-based pruning of the GM components. For GM--particles conversion, we develop an importance sampling based method that enables a parallelization of filtering and dissemination/fusion operations. The proposed distributed particle-PHD filter is able to integrate GM-based local PHD filters. Simulations demonstrate the excellent performance and small communication and computation requirements of our filter.

Tiancheng Li, Weijian Luo, Zhiyang Chen et al.

Proper guidance strategies are essential to achieve high-quality generation results without retraining diffusion and flow-based text-to-image models. Existing guidance either requires specific training or strong inductive biases of diffusion model networks, which potentially limits their ability and application scope. Motivated by the observation that artifact outliers can be detected by a significant decline in the density from a noisier to a cleaner noise level, we propose Self-Guidance (SG), which can significantly improve the quality of the generated image by suppressing the generation of low-quality samples. The biggest difference from existing guidance is that SG only relies on the sampling score function of the original diffusion or flow model at different noise levels, with no need for any tricky and expensive guidance-specific training. This makes SG highly flexible to be used in a plug-and-play manner by any diffusion or flow models. We also introduce an efficient variant of SG, named SG-prev, which reuses the output from the immediately previous diffusion step to avoid additional forward passes of the diffusion network.We conduct extensive experiments on text-to-image and text-to-video generation with different architectures, including UNet and transformer models. With open-sourced diffusion models such as Stable Diffusion 3.5 and FLUX, SG exceeds existing algorithms on multiple metrics, including both FID and Human Preference Score. SG-prev also achieves strong results over both the baseline and the SG, with 50 percent more efficiency. Moreover, we find that SG and SG-prev both have a surprisingly positive effect on the generation of physiologically correct human body structures such as hands, faces, and arms, showing their ability to eliminate human body artifacts with minimal efforts. We have released our code at https://github.com/maple-research-lab/Self-Guidance.

Zemin Huang, Zhiyang Chen, Zijun Wang et al.

We introduce the Diffusion Chain of Lateral Thought (DCoLT), a reasoning framework for diffusion language models. DCoLT treats each intermediate step in the reverse diffusion process as a latent "thinking" action and optimizes the entire reasoning trajectory to maximize the reward on the correctness of the final answer with outcome-based Reinforcement Learning (RL). Unlike traditional Chain-of-Thought (CoT) methods that follow a causal, linear thinking process, DCoLT allows bidirectional, non-linear reasoning with no strict rule on grammatical correctness amid its intermediate steps of thought. We implement DCoLT on two representative Diffusion Language Models (DLMs). First, we choose SEDD as a representative continuous-time discrete diffusion model, where its concrete score derives a probabilistic policy to maximize the RL reward over the entire sequence of intermediate diffusion steps. We further consider the discrete-time masked diffusion language model -- LLaDA, and find that the order to predict and unmask tokens plays an essential role to optimize its RL action resulting from the ranking-based Unmasking Policy Module (UPM) defined by the Plackett-Luce model. Experiments on both math and code generation tasks show that using only public data and 16 H800 GPUs, DCoLT-reinforced DLMs outperform other DLMs trained by SFT or RL or even both. Notably, DCoLT-reinforced LLaDA boosts its reasoning accuracy by +9.8%, +5.7%, +11.4%, +19.5% on GSM8K, MATH, MBPP, and HumanEval.

Zilyu Ye, Zhiyang Chen, Tiancheng Li et al.

Diffusion and flow matching models have achieved remarkable success in text-to-image generation. However, these models typically rely on the predetermined denoising schedules for all prompts. The multi-step reverse diffusion process can be regarded as a kind of chain-of-thought for generating high-quality images step by step. Therefore, diffusion models should reason for each instance to adaptively determine the optimal noise schedule, achieving high generation quality with sampling efficiency. In this paper, we introduce the Time Prediction Diffusion Model (TPDM) for this. TPDM employs a plug-and-play Time Prediction Module (TPM) that predicts the next noise level based on current latent features at each denoising step. We train the TPM using reinforcement learning to maximize a reward that encourages high final image quality while penalizing excessive denoising steps. With such an adaptive scheduler, TPDM not only generates high-quality images that are aligned closely with human preferences but also adjusts diffusion time and the number of denoising steps on the fly, enhancing both performance and efficiency. With Stable Diffusion 3 Medium architecture, TPDM achieves an aesthetic score of 5.44 and a human preference score (HPS) of 29.59, while using around 50% fewer denoising steps to achieve better performance.

Tiancheng Li, Yan Song, Hongqi Fan et al.

In the realm of target tracking, performance evaluation plays a pivotal role in the design, comparison, and analytics of trackers. Compared with the traditional trajectory composed of a set of point-estimates obtained by a tracker in the measurement time-series, the trajectory that our series of studies including this paper pursued is given by a curve function of time (FoT). The trajectory FoT provides complete information of the movement of the target over time and can be used to infer the state corresponding to arbitrary time, not only at the measurement time. However, there are no metrics available for comparing and evaluating the trajectory FoT. To address this lacuna, we propose a metric denominated as the spatiotemporal-aligned trajectory integral distance (Star-ID). The StarID associates and aligns the estimated and actual trajectories in the spatio-temporal domain and distinguishes between the time-aligned and unaligned segments in calculating the spatial divergence including false alarm, miss-detection and localization errors. The effectiveness of the proposed distance metric and the time-averaged version is validated through theoretical analysis and numerical examples of a single target or multiple targets.

Tiancheng Li, Yuhang Wang, Zhiyang Chen et al.

Prompt evolution algorithms offer a powerful paradigm for enhancing AI systems based on closed-source models, while few work explores whether aggregating results from multiple prompts to reach a consensus can further advance the system capability boundary. In this paper, we introduce Consensus-Evolve (C-Evolve), an evolutionary algorithm that discovers a group of prompts whose aggregated outputs after majority voting achieve optimal performance. More specifically, C-Evolve employs an island-based evolutionary algorithm to maintain population diversity, and prompts from distinct islands are selected to form groups to aggregate their outputs. The key difference from single individual evolution is a voting score, which evaluates each individual prompt's contribution within groups. We take this as the fitness score for evolution instead of individual performance. Consequently, C-Evolve is more likely to produce and maintain prompts with higher potential to form a high-performing group and eliminate low-performing ones, gradually improving the group performance after reaching consensus. Our method achieves state-of-the-art performance across a wide range of tasks, including both open-ended tasks like HotpotQA and closed-ended tasks like MATH. On Qwen3-8B, C-Evolve achieves 70.67% on HotpotQA and 43.88% on IFBench, which are 4.95% and 2.73% higher than GEPA, respectively. For GPT-4.1-mini, the accuracy on IFBench is further improved to 47.96% and reaches 95.33% in the MATH benchmark. These results demonstrate the C-Evolve's competitive performance.

Tiancheng Li, Jinxiu Liu, Huajun Chen et al.

Instruction-based image editing has made a great process in using natural human language to manipulate the visual content of images. However, existing models are limited by the quality of the dataset and cannot accurately localize editing regions in images with complex object relationships. In this paper, we propose Reinforcement Learning Guided Image Editing Method(InstructRL4Pix) to train a diffusion model to generate images that are guided by the attention maps of the target object. Our method maximizes the output of the reward model by calculating the distance between attention maps as a reward function and fine-tuning the diffusion model using proximal policy optimization (PPO). We evaluate our model in object insertion, removal, replacement, and transformation. Experimental results show that InstructRL4Pix breaks through the limitations of traditional datasets and uses unsupervised learning to optimize editing goals and achieve accurate image editing based on natural human commands.

Tiancheng Li, Huimin Chen, Shudong Sun et al.

We present a continuous time state estimation framework that unifies traditionally individual tasks of smoothing, tracking, and forecasting (STF), for a class of targets subject to smooth motion processes, e.g., the target moves with nearly constant acceleration or affected by insignificant noises. Fundamentally different from the conventional Markov transition formulation, the state process is modeled by a continuous trajectory function of time (FoT) and the STF problem is formulated as an online data fitting problem with the goal of finding the trajectory FoT that best fits the observations in a sliding time-window. Then, the state of the target, whether the past (namely, smoothing), the current (filtering) or the near-future (forecasting), can be inferred from the FoT. Our framework releases stringent statistical modeling of the target motion in real time, and is applicable to a broad range of real world targets of significance such as passenger aircraft and ships which move on scheduled, (segmented) smooth paths but little statistical knowledge is given about their real time movement and even about the sensors. In addition, the proposed STF framework inherits the advantages of data fitting for accommodating arbitrary sensor revisit time, target maneuvering and missed detection. The proposed method is compared with state of the art estimators in scenarios of either maneuvering or non-maneuvering target.

Tiancheng Li, Shudong Sun, Tariq P. Sattar et al.

During the last two decades there has been a growing interest in Particle Filtering (PF). However, PF suffers from two long-standing problems that are referred to as sample degeneracy and impoverishment. We are investigating methods that are particularly efficient at Particle Distribution Optimization (PDO) to fight sample degeneracy and impoverishment, with an emphasis on intelligence choices. These methods benefit from such methods as Markov Chain Monte Carlo methods, Mean-shift algorithms, artificial intelligence algorithms (e.g., Particle Swarm Optimization, Genetic Algorithm and Ant Colony Optimization), machine learning approaches (e.g., clustering, splitting and merging) and their hybrids, forming a coherent standpoint to enhance the particle filter. The working mechanism, interrelationship, pros and cons of these approaches are provided. In addition, Approaches that are effective for dealing with high-dimensionality are reviewed. While improving the filter performance in terms of accuracy, robustness and convergence, it is noted that advanced techniques employed in PF often causes additional computational requirement that will in turn sacrifice improvement obtained in real life filtering. This fact, hidden in pure simulations, deserves the attention of the users and designers of new filters.

Tiancheng Li, Shudong Sun, Tariq Pervez Sattar

This letter provides an adaptive resampling method. It determines the number of particles to resample so that the Kullback-Leibler distance (KLD) between distribution of particles before resampling and after resampling does not exceed a pre-specified error bound. The basis of the method is the same as Fox's KLD-sampling but implemented differently. The KLD-sampling assumes that samples are coming from the true posterior distribution and ignores any mismatch between the true and the proposal distribution. In contrast, we incorporate the KLD measure into the resampling in which the distribution of interest is just the posterior distribution. That is to say, for sample size adjustment, it is more theoretically rigorous and practically flexible to measure the fit of the distribution represented by weighted particles based on KLD during resampling than in sampling. Simulations of target tracking demonstrate the efficiency of our method.