arXiv subjects

Haotian Xu

Publications and source records attributed to Haotian Xu.

At least 55 records · Page 3Linked to original sources

arXiv2025

Concept-Guided Backdoor Attack on Vision Language Models

Vision-Language Models (VLMs) have achieved impressive progress in multimodal text generation, yet their rapid adoption raises increasing concerns about security vulnerabilities. Existing backdoor attacks against VLMs primarily rely on explicit pixel-level triggers or imperceptible perturbations injected into images. While effective, these approaches reduce stealthiness and remain vulnerable to image-based defenses. We introduce concept-guided backdoor attacks, a new paradigm that operates at the semantic concept level rather than on raw pixels. We propose two different attacks. The first, Concept-Thresholding Poisoning (CTP), uses explicit concepts in natural images as triggers: only samples containing the target concept are poisoned, causing the model to behave normally in all other cases but consistently inject malicious outputs whenever the concept appears. The second, CBL-Guided Unseen Backdoor (CGUB), leverages a Concept Bottleneck Model (CBM) during training to intervene on internal concept activations, while discarding the CBM branch at inference time to keep the VLM unchanged. This design enables systematic replacement of a targeted label in generated text (for example, replacing "cat" with "dog"), even when the replacement behavior never appears in the training data. Experiments across multiple VLM architectures and datasets show that both CTP and CGUB achieve high attack success rates while maintaining moderate impact on clean-task performance. These findings highlight concept-level vulnerabilities as a critical new attack surface for VLMs.

cs.CR↗

arXiv2025

REINFORCE++: Stabilizing Critic-Free Policy Optimization with Global Advantage Normalization

Reinforcement Learning from Human Feedback~(RLHF) plays a crucial role in aligning Large Language Models~(LLMs). The dominant algorithm, Proximal Policy Optimization~(PPO), employs a critic network to estimate advantages, which introduces significant computational and memory overhead. To address this, a family of critic-free algorithms (e.g., GRPO, RLOO) has emerged. However, these methods typically rely on \textit{prompt-level (local)} advantage normalization, which suffers from inaccurate advantage estimation, a tendency to overfit, and, as we show, is a theoretically biased estimator. To solve these challenges, we introduce REINFORCE++, a critic-free framework centered on \textbf{Global Advantage Normalization}. By normalizing advantages across the entire global batch rather than small, prompt-specific groups, our method provides a more stable and theoretically sound, \textit{effectively unbiased} estimate (whose bias vanishes as batch size increases). We introduce two variants: REINFORCE++, a highly efficient and general algorithm ($k \ge 1$) for general-domain RLHF, and REINFORCE++ /w baseline, a robust group-sampling variant ($k > 1$) for complex reasoning tasks. Our empirical evaluation demonstrates that each variant shows superior stability and performance in its respective domain, outperforming existing methods and even PPO in complex agentic settings.

cs.CL↗

arXiv2025

High Performance Distributed Control for Large-Scale Linear Systems: A Cover-Based Distributed Observer Approach

In recent years, the distributed-observer-based distributed control law has shown powerful ability to arbitrarily approximate the centralized control performance. However, the traditional distributed observer requires each local observer to reconstruct the state information of the whole system, which is unrealistic for large-scale scenarios. To fill this gap, This paper presents a coverage solution algorithm for large-scale systems that accounts for both physical and communication network characteristics, which can significantly reduce the dimension of local observers. Then, the cover-based distributed observer for large-scale systems is proposed to overcome the problem that the system dynamics are difficult to estimate due to the coupling between cover sets. Furthermore, the two-layer Lyapunov analysis method is adopted and the dynamic transformation lemma of compact errors is proved, which solves the problem of analyzing stability of the error dynamic of the cover-based distributed observer. Finally, it is proved that the distributed control law based on the cover-based distributed observer can also arbitrarily approximate the control performance of the centralized control law, and the dimension of the local observer is greatly reduced compared with the traditional method. The simulation results show the validity of the developed theories.

eess.SY↗

arXiv2025

A 95 GeV Higgs boson and spontaneous CP-violation at the finite temperature

The ATLAS and CMS collaborations reported a diphoton excess in the invariant mass distribution around the 95.4 GeV with a local significance of $3.1σ$. Moreover, there is another $2.3σ$ local excess in the $b\bar{b}$ final state at LEP in the same mass region. A plausible solution is that the Higgs sector is extended to include an additional Higgs boson with a mass of $95.4$ GeV. We study a complex singlet scalar extension of the two-Higgs-doublet model in which the 95.4 GeV Higgs is from the mixing of three CP-even Higgs fields. In addition, the extended Higgs potential can achieve spontaneous CP-violation at the finite temperature and restore CP symmetry at the present temperature of the Universe. We find that the model can simultaneously explain the baryon asymmetry of the Universe, the diphoton and $b\bar{b}$ excesses around the 95.4 GeV while satisfying various relevant constraints including the experiments of collider and electric dipole moment.

hep-ph↗

arXiv2025

95 GeV Higgs boson and nano-Hertz gravitational waves from domain walls in the N2HDM

We explore the diphoton and $b\bar{b}$ excesses at 95.4 GeV, as well as nano-Hertz gravitational waves originating from domain walls, within the framework of the next-to-two-Higgs-doublet model (N2HDM), which extends the two-Higgs-doublet model by introducing a real singlet scalar subject to a discrete $Z_2$ symmetry. The $Z_2$ symmetry is spontaneously broken by the non-zero vacuum expectation value of the singlet scalar, $v_s$, which leads to the formation of domain walls. We discuss two different scenarios: in scenario A, the 95.4 GeV Higgs boson predominantly originates from the singlet field, while in scenario B, it arises mainly from the CP-even components of the Higgs doublets. Taking into account relevant theoretical and experimental constraints, we find that scenario A can fully account for both the diphoton and $b\bar{b}$ excesses at 95.4 GeV within the $1σ$ range. Moreover, the peak amplitude of the gravitational wave spectrum at a peak frequency of $10^{-9}$ Hz can reach $2 \times 10^{-12}$ for $v_s = 100$ TeV. Scenario B only marginally accounts for the diphoton and $b\bar{b}$ excesses at the $1σ$ level, but the peak amplitude of the gravitational wave spectrum at the peak frequency of $10^{-9}$ Hz can reach $6\times 10^{-8}$ for $v_s=100$ TeV. The nano-Hertz gravitational wave signals predicted in both scenarios can be tested by the current and future pulsar timing array projects.

hep-ph↗

arXiv2025

OpenRLHF: An Easy-to-use, Scalable and High-performance RLHF Framework

Large Language Models (LLMs) fine-tuned via Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning with Verifiable Rewards (RLVR) significantly improve the alignment of human-AI values, further raising the upper bound of AI capabilities, particularly in reasoning-intensive, long-context Chain-of-Thought (CoT) tasks. However, existing frameworks commonly face challenges such as inference bottlenecks and complexity barriers, which restrict their accessibility to newcomers. To bridge this gap, we introduce \textbf{OpenRLHF}, a user-friendly, scalable, and easy-to-learn open-source RLHF framework built upon Ray, vLLM, DeepSpeed, and HuggingFace Transformers, featuring a simplified design, clear code structure, and comprehensive documentation to facilitate entry for researchers and practitioners. Experimental results show that OpenRLHF achieves superior training efficiency, with speedups ranging from 1.22x to 1.68x across different model sizes, compared to state-of-the-art frameworks. Additionally, it requires significantly fewer lines of code for implementation. OpenRLHF is publicly available at https://github.com/OpenRLHF/OpenRLHF, and has already been adopted by leading institutions to accelerate RLHF research and learning.

cs.AI↗

arXiv2025

SilentStriker:Toward Stealthy Bit-Flip Attacks on Large Language Models

The rapid adoption of large language models (LLMs) in critical domains has spurred extensive research into their security issues. While input manipulation attacks (e.g., prompt injection) have been well studied, Bit-Flip Attacks (BFAs) -- which exploit hardware vulnerabilities to corrupt model parameters and cause severe performance degradation -- have received far less attention. Existing BFA methods suffer from key limitations: they fail to balance performance degradation and output naturalness, making them prone to discovery. In this paper, we introduce SilentStriker, the first stealthy bit-flip attack against LLMs that effectively degrades task performance while maintaining output naturalness. Our core contribution lies in addressing the challenge of designing effective loss functions for LLMs with variable output length and the vast output space. Unlike prior approaches that rely on output perplexity for attack loss formulation, which inevitably degrade output naturalness, we reformulate the attack objective by leveraging key output tokens as targets for suppression, enabling effective joint optimization of attack effectiveness and stealthiness. Additionally, we employ an iterative, progressive search strategy to maximize attack efficacy. Experiments show that SilentStriker significantly outperforms existing baselines, achieving successful attacks without compromising the naturalness of generated text.

cs.CR↗

arXiv2025

LiteLong: Resource-Efficient Long-Context Data Synthesis for LLMs

High-quality long-context data is essential for training large language models (LLMs) capable of processing extensive documents, yet existing synthesis approaches using relevance-based aggregation face challenges of computational efficiency. We present LiteLong, a resource-efficient method for synthesizing long-context data through structured topic organization and multi-agent debate. Our approach leverages the BISAC book classification system to provide a comprehensive hierarchical topic organization, and then employs a debate mechanism with multiple LLMs to generate diverse, high-quality topics within this structure. For each topic, we use lightweight BM25 retrieval to obtain relevant documents and concatenate them into 128K-token training samples. Experiments on HELMET and Ruler benchmarks demonstrate that LiteLong achieves competitive long-context performance and can seamlessly integrate with other long-dependency enhancement methods. LiteLong makes high-quality long-context data synthesis more accessible by reducing both computational and data engineering costs, facilitating further research in long-context language training.

cs.CL↗

arXiv2025

Agent RL Scaling Law: Agent RL with Spontaneous Code Execution for Mathematical Problem Solving

Large Language Models (LLMs) often struggle with mathematical reasoning tasks requiring precise, verifiable computation. While Reinforcement Learning (RL) from outcome-based rewards enhances text-based reasoning, understanding how agents autonomously learn to leverage external tools like code execution remains crucial. We investigate RL from outcome-based rewards for Tool-Integrated Reasoning, ZeroTIR, training base LLMs to spontaneously generate and execute Python code for mathematical problems without supervised tool-use examples. Our central contribution is we demonstrate that as RL training progresses, key metrics scale predictably. Specifically, we observe strong positive correlations where increased training steps lead to increases in the spontaneous code execution frequency, the average response length, and, critically, the final task accuracy. This suggests a quantifiable relationship between computational effort invested in training and the emergence of effective, tool-augmented reasoning strategies. We implement a robust framework featuring a decoupled code execution environment and validate our findings across standard RL algorithms and frameworks. Experiments show ZeroTIR significantly surpasses non-tool ZeroRL baselines on challenging math benchmarks. Our findings provide a foundational understanding of how autonomous tool use is acquired and scales within Agent RL, offering a reproducible benchmark for future studies. Code is released at \href{https://github.com/yyht/openrlhf_async_pipline}{https://github.com/yyht/openrlhf\_async\_pipline}.

cs.AI↗

arXiv2025

SFedKD: Sequential Federated Learning with Discrepancy-Aware Multi-Teacher Knowledge Distillation

Federated Learning (FL) is a distributed machine learning paradigm which coordinates multiple clients to collaboratively train a global model via a central server. Sequential Federated Learning (SFL) is a newly-emerging FL training framework where the global model is trained in a sequential manner across clients. Since SFL can provide strong convergence guarantees under data heterogeneity, it has attracted significant research attention in recent years. However, experiments show that SFL suffers from severe catastrophic forgetting in heterogeneous environments, meaning that the model tends to forget knowledge learned from previous clients. To address this issue, we propose an SFL framework with discrepancy-aware multi-teacher knowledge distillation, called SFedKD, which selects multiple models from the previous round to guide the current round of training. In SFedKD, we extend the single-teacher Decoupled Knowledge Distillation approach to our multi-teacher setting and assign distinct weights to teachers' target-class and non-target-class knowledge based on the class distributional discrepancy between teacher and student data. Through this fine-grained weighting strategy, SFedKD can enhance model training efficacy while mitigating catastrophic forgetting. Additionally, to prevent knowledge dilution, we eliminate redundant teachers for the knowledge distillation and formalize it as a variant of the maximum coverage problem. Based on the greedy strategy, we design a complementary-based teacher selection mechanism to ensure that the selected teachers achieve comprehensive knowledge space coverage while reducing communication and computational costs. Extensive experiments show that SFedKD effectively overcomes catastrophic forgetting in SFL and outperforms state-of-the-art FL methods.

cs.LG↗

arXiv2025

From System 1 to System 2: A Survey of Reasoning Large Language Models

Achieving human-level intelligence requires refining the transition from the fast, intuitive System 1 to the slower, more deliberate System 2 reasoning. While System 1 excels in quick, heuristic decisions, System 2 relies on logical reasoning for more accurate judgments and reduced biases. Foundational Large Language Models (LLMs) excel at fast decision-making but lack the depth for complex reasoning, as they have not yet fully embraced the step-by-step analysis characteristic of true System 2 thinking. Recently, reasoning LLMs like OpenAI's o1/o3 and DeepSeek's R1 have demonstrated expert-level performance in fields such as mathematics and coding, closely mimicking the deliberate reasoning of System 2 and showcasing human-like cognitive abilities. This survey begins with a brief overview of the progress in foundational LLMs and the early development of System 2 technologies, exploring how their combination has paved the way for reasoning LLMs. Next, we discuss how to construct reasoning LLMs, analyzing their features, the core methods enabling advanced reasoning, and the evolution of various reasoning LLMs. Additionally, we provide an overview of reasoning benchmarks, offering an in-depth comparison of the performance of representative reasoning LLMs. Finally, we explore promising directions for advancing reasoning LLMs and maintain a real-time \href{https://github.com/zzli2022/Awesome-Slow-Reason-System}{GitHub Repository} to track the latest developments. We hope this survey will serve as a valuable resource to inspire innovation and drive progress in this rapidly evolving field.

cs.AI↗

arXiv2025

TL;DR: Too Long, Do Re-weighting for Efficient LLM Reasoning Compression

Large Language Models (LLMs) have recently achieved remarkable progress by leveraging Reinforcement Learning and extended Chain-of-Thought (CoT) techniques. However, the challenge of performing efficient language reasoning--especially during inference with extremely long outputs--has drawn increasing attention from the research community. In this work, we propose a dynamic ratio-based training pipeline that does not rely on sophisticated data annotations or interpolation between multiple models. We continuously balance the weights between the model's System-1 and System-2 data to eliminate redundant reasoning processes while preserving the model's reasoning capability. We validate our approach across models on DeepSeek-R1-Distill-7B and DeepSeek-R1-Distill-14B and on a diverse set of benchmarks with varying difficulty levels. Our method significantly reduces the number of output tokens by nearly 40% while maintaining the accuracy of the reasoning. Our code and data will be available soon.

cs.CL↗

arXiv2025

CodePMP: Scalable Preference Model Pretraining for Large Language Model Reasoning

Large language models (LLMs) have made significant progress in natural language understanding and generation, driven by scalable pretraining and advanced finetuning. However, enhancing reasoning abilities in LLMs, particularly via reinforcement learning from human feedback (RLHF), remains challenging due to the scarcity of high-quality preference data, which is labor-intensive to annotate and crucial for reward model (RM) finetuning. To alleviate this issue, we introduce CodePMP, a scalable preference model pretraining (PMP) pipeline that utilizes a large corpus of synthesized code-preference pairs from publicly available high-quality source code. CodePMP improves RM finetuning efficiency by pretraining preference models on large-scale synthesized code-preference pairs. We evaluate CodePMP on mathematical reasoning tasks (GSM8K, MATH) and logical reasoning tasks (ReClor, LogiQA2.0), consistently showing significant improvements in reasoning performance of LLMs and highlighting the importance of scalable preference model pretraining for efficient reward modeling.

cs.AI↗

arXiv2025

Beyond Single-Task: Robust Multi-Task Length Generalization for LLMs

Length generalization, the ability to solve problems longer than those seen during training, remains a critical challenge for large language models (LLMs). Previous work modifies positional encodings (PEs) and data formats to improve length generalization on specific symbolic tasks such as addition and sorting. However, these approaches are fundamentally limited to special tasks, often degrading general language performance. Furthermore, they are typically evaluated on small transformers trained from scratch on single tasks and can cause performance drop when applied during post-training stage of practical LLMs with general capabilities. Hu et al., (2024) proposed Rule-Following Fine-Tuning (RFFT) to improve length generalization in the post-training stage of LLMs. Despite its compatibility with practical models and strong performance, RFFT is proposed for single tasks too, requiring re-training for each individual task with extensive examples. In this paper, we study length generalization in multi-task settings and propose Meta Rule-Following Fine-Tuning (Meta-RFFT), the first framework enabling robust cross-task length generalization. As our first contribution, we construct a large length generalization dataset containing 86 tasks spanning code execution, number processing, symbolic and logical reasoning tasks, beyond the common addition or multiplication tasks. Secondly, we show that cross-task length generalization is possible with Meta-RFFT. After training on a large number of tasks and instances, the models achieve remarkable length generalization ability on unseen tasks with minimal fine-tuning or one-shot prompting. For example, after fine-tuning on 1 to 5 digit addition, our 32B model achieves 95% accuracy on 30 digit addition, significantly outperforming the state-of-the-art reasoning models (DeepSeek-R1-671B: 72%), despite never seeing this task during RF-pretraining.

cs.CL↗

arXiv2025

Shuttle Between the Instructions and the Parameters of Large Language Models

The interaction with Large Language Models (LLMs) through instructions has been extensively investigated in the research community. While instructions have been widely used as the guidelines for task solving, this paper further notices that both instructions and parameters are the compression of task data. Therefore, they could be strongly correlated and can be learned to predict one from the other. This paper proposes a novel neural network framework, SHIP (\textbf{Sh}uttle between the \textbf{I}nstructions and the \textbf{P}arameters), to model and learn the mutual mappings between the instructions and the parameters of LLMs. We verify that SHIP can effectively map one of the instructions/parameters to the other by evaluating it on the tasks of instruction deduction and induction. The results show that SHIP performs better than existing baseline methods in terms of deductive capabilities while significantly surpassing them in inductive capabilities. Moreover, SHIP can effectively combine the two mapping processes to perform excellent inductive reasoning. The code and data for this paper are released at https://anonymous.4open.science/r/Shuttle-Between-Instructions-Parameters/.

cs.LG↗

arXiv2025

Probabilistic Uncertain Reward Model

Reinforcement learning from human feedback (RLHF) is a critical technique for training large language models. However, conventional reward models based on the Bradley-Terry model (BTRM) often suffer from overconfidence when faced with inconsistent labels or out-of-distribution samples, leading to reward hacking, where the policy model blindly optimizes for proxy rewards while degrading true performance. This paper proposes the Probabilistic Uncertain Reward Model (PURM), which generalizes the Bradley-Terry model to learn the reward distributions that emerged from the preference data. We theoretically derive the loss function of PURM and introduce a novel method that uses the overlap between distributions to quantify uncertainty. Empirical results show that PURM outperforms existing methods with more accurate reward and sound uncertainty estimations, and sustains effective learning for more optimization steps and obtain higher maximum win rate in RLHF. The data and code of this paper are released at https://anonymous.4open.science/r/Probabilistic-Uncertain-Reward-Model/

cs.LG↗

arXiv2025

JPDS-NN: Reinforcement Learning-Based Dynamic Task Allocation for Agricultural Vehicle Routing Optimization

The Entrance Dependent Vehicle Routing Problem (EDVRP) is a variant of the Vehicle Routing Problem (VRP) where the scale of cities influences routing outcomes, necessitating consideration of their entrances. This paper addresses EDVRP in agriculture, focusing on multi-parameter vehicle planning for irregularly shaped fields. To address the limitations of traditional methods, such as heuristic approaches, which often overlook field geometry and entrance constraints, we propose a Joint Probability Distribution Sampling Neural Network (JPDS-NN) to effectively solve the EDVRP. The network uses an encoder-decoder architecture with graph transformers and attention mechanisms to model routing as a Markov Decision Process, and is trained via reinforcement learning for efficient and rapid end-to-end planning. Experimental results indicate that JPDS-NN reduces travel distances by 48.4-65.4%, lowers fuel consumption by 14.0-17.6%, and computes two orders of magnitude faster than baseline methods, while demonstrating 15-25% superior performance in dynamic arrangement scenarios. Ablation studies validate the necessity of cross-attention and pre-training. The framework enables scalable, intelligent routing for large-scale farming under dynamic constraints.

cs.RO↗

arXiv2025

Ordered Genetic Algorithm for Entrance Dependent Vehicle Routing Problem in Farms

Vehicle Routing Problems (VRP) are widely studied issues that play important roles in many production scenarios. We have noticed that in some practical scenarios of VRP, the size of cities and their entrances can significantly influence the optimization process. To address this, we have constructed the Entrance Dependent VRP (EDVRP) to describe such problems. We provide a mathematical formulation for the EDVRP in farms and propose an Ordered Genetic Algorithm (OGA) to solve it. The effectiveness of OGA is demonstrated through our experiments, which involve a multitude of randomly generated cases. The results indicate that OGA offers certain advantages compared to a random strategy baseline and a genetic algorithm without ordering. Furthermore, the novel operators introduced in this paper have been validated through ablation experiments, proving their effectiveness in enhancing the performance of the algorithm.

cs.RO↗