arXiv subjects

Xiang Huang

Publications and source records attributed to Xiang Huang.

At least 19 recordsLinked to original sources

arXiv2026

Ripple: An Open, AI-Formalized Lean 4 Framework for Computing with CRNs

We present Ripple, an open, AI-formalized Lean 4 framework for the mathematics of computing with chemical reaction networks (CRNs) -- one extensible, machine-checked development that gathers several strands of the field into a single setting, and is built to grow. It formalizes: the theory of which real numbers a CRN can compute -- a single Lean definition of real-time CRN computation, the class of reals it captures, and the compilation pipeline (a GPAC / polynomial-ODE layer, a dual-rail compiler, and four stages down to large-population protocols) that realizes them, built so that adding a new number is a plug-in; three landmark population-protocol majority algorithms -- approximate, exact, and self-stabilizing exact majority; the stochastic-to-deterministic bridge, through three machine-checked versions of Kurtz's mean-field theorem; and two classical Turing-completeness results -- Bournez--Graca--Pouly for the deterministic GPAC and Soloveichik--Cook--Winfree--Bruck for stochastic CRNs. Each of these is verified to depend on exactly the three Mathlib foundational axioms, with no sorry. Along the way the formalization repaired genuine, previously unnoticed gaps in published proofs -- a compilation step that can transiently leave the unit interval, and an algebraic-number construction that silently rests on Catalan's conjecture -- and surfaced a sharp open problem about when a holonomic series admits an exact, rational-data polynomial-ODE encoding. The whole development is open and every proof is kernel-checked, so the results can be independently re-verified; and because it was written predominantly by AI agents using only publicly available models, the workflow that produced it can be reproduced with the same public toolchain.

cs.LO↗

arXiv2026

Bounded Analog Complexity

Current analog complexity theory, built on the General-Purpose Analog Computer (GPAC) model and polynomial ODEs, allows unbounded state variables -- an assumption that is physically unrealistic for chemical reaction networks and other laboratory-scale analog computers. We develop a bounded analog complexity theory in which all state variables remain in compact intervals and physical time (wall-clock time) is the only diverging resource. Our main technical contribution is bounded surrogate compilation, a compilation framework that transforms unbounded polynomial ODE systems into bounded ones while preserving computational limits and time-to-precision guarantees. We prove that if a system is compiled into a bounded system through our algorithm, the wall-clock time of the compiled system is polynomial in the arc length and physical time of the original system. We exhibit concrete constructions demonstrating fine-grained bounded time complexity -- a tunable polynomial-degree family, a Lambert-$W$-based system achieving $\Theta(r\log r)$ time-to-precision (where $r$ is the desired precision parameter, in nats: $|x(t)-\alpha|<e^{-r}$), and an iterated-logarithm tower realizing arbitrarily high complexity classes -- all for the task of computing the constant 1. We show that bounded GPACs are closed under exponentiation ($\alpha^\beta$) with time complexity equal to the harder input, and that the full GPAC-to-CRN compilation pipeline preserves time complexity class via a low-pass filter analysis of readout modules.

cs.CC↗

arXiv2026

EIBench: A Simulator-Based Benchmark and Turn-Credit RL for Emotion Management

Emotional intelligence (EI) in Large Language Models (LLMs) is often evaluated through static understanding tasks or single-response dialogue generation. However, emotion management is interactive: a good model should not only recognize a user's emotion, but also improve the user's emotional and relational state over several turns. We introduce EIBench, a simulator-based benchmark for interactive emotion management. EIBench contains 2,222 scenarios, with 2,009 for training and 213 for held-out testing. The scenarios are organized by a 2x2 taxonomy covering Support, Defense, Repair, and Charm, which together capture different forms of support, boundary maintenance, trust repair, and rapport building. In each scenario, an LLM simulator plays the user, updates an emotion-relation state after each turn, and maps the final state to an anchor-based score. This design makes EIBench both an evaluation benchmark and a training environment: the final state gives the outcome reward, while the per-turn state updates provide dense feedback for RL. We evaluate 15 open- and closed-source LLMs. Current models perform well on support and rapport-building scenes, but struggle with boundary maintenance under user pressure. To improve the EI ability of LLMs, we propose Centered Turn-Credit GRPO (CTC-GRPO), a GRPO extension that reuses the simulator's per-turn state updates as dense turn-level feedback while preserving the final outcome reward. CTC-GRPO improves Qwen3-8B from -22.4 to +22.4 on EIBench and also improves on out-of-distribution evaluations including SAGE (+12.4) and EQBench3 (+20.9%). Our results show that simulator-tracked user states can support both evaluation and training for multi-turn emotion management.

cs.CL↗

arXiv2026

Scaling Self-Evolving Agents via Parametric Memory

Existing memory-augmented LLM agents store past experience exclusively in prompt space, as textual summaries or retrieved passages, while keeping model parameters frozen throughout a rollout. Such agents can \emph{look up} what they have seen but cannot \emph{learn from} it: their policy is unchanged by experience, and any information dropped from the context is permanently lost. We introduce \texttt{TMEM}, a self-evolving parametric memory framework in which the agent not only compresses history into explicit memory but also absorbs distilled supervision into fast LoRA weights $\Delta_t$ via lightweight online updates, genuinely altering its future behavior within a single episode. We formalize this as an agentic decision process with fast-weight rollout dynamics: actions are sampled from $\pi_{\theta_0+\Delta_t}$, while extraction actions produce supervision that updates $\Delta_t$ for subsequent decisions. This view makes the extraction policy directly optimizable by RL: training $\theta_0$ improves not only task actions but also the quality of the data used for online LoRA adaptation. We further propose SVD-based initialization of the LoRA subspace to accelerate online convergence. Experiments on LoCoMo, LongMemEval-S, multi-objective search, and CL-Bench show that \texttt{TMEM} consistently outperforms summary-based and retrieval-based baselines across different model scales.

cs.AI↗

arXiv2026

Magnetic order and excitations in the magnetically intercalated van der Waals material Cr$_{\frac{1}{4}}$NbSe$_2$

Cr$_{\frac{1}{4}}$NbSe$_2$ is a triangular lattice magnet in which magnetic Cr$^{3+}$ ions are intercalated to form triangular lattices between NbSe$_2$ van der Waals layers stacked along the c axis. By unpolarized and polarized neutron scattering experiments, we have revealed that the magnetic ground state of this system is a 120$^{\circ}$-type antiferromagnetic order characterized by the magnetic propagation wave vector of $q=(\frac{1}{3}, \frac{1}{3}, 0)$. We also performed inelastic neutron scattering measurements using co-aligned single crystals, and determined dispersion relations of magnetic excitations at low temperatures. Comparing the observed spectra with calculations based on the linear spin-wave theory, we revealed that the out-of-plane ferromagnetic interaction is fairly strong as compared to the in-plane nearest neighbor antiferromagnetic interaction. Although the crystal structure of this system is composed of two-dimensional van der Waals layers, the magnetic order has a three dimensional character, which would be attributed to long-range magnetic interactions mediated by conduction electrons.

cond-mat.str-el↗

arXiv2025

Multi-Head Finite-State Dimension

We introduce multi-head finite-state dimension, a generalization of finite-state dimension in which a group of finite-state agents (the heads) with oblivious, one-way movement rules, each reporting only one symbol at a time, enable their leader to bet on subsequent symbols in an infinite data stream. In aggregate, such a scheme constitutes an $h$-head finite state gambler whose maximum achievable growth rate of capital in this task, quantified using betting strategies called gales, determines the multi-head finite-state dimension of the sequence. The 1-head case is equivalent to finite-state dimension as defined by Dai, Lathrop, Lutz and Mayordomo (2004). In our main theorem, we prove a strict hierarchy as the number of heads increases, giving an explicit sequence family that separates, for each positive integer $h$, the earning power of $h$-head finite-state gamblers from that of $(h+1)$-head finite-state gamblers. We prove that multi-head finite-state dimension is stable under finite unions but that the corresponding quantity for any fixed number $h>1$ of heads--the $h$-head finite-state predimension--lacks this stability property.

cs.IT↗

arXiv2025

PDEformer-2: A Versatile Foundation Model for Two-Dimensional Partial Differential Equations

Partial differential equations (PDEs) play a central role in describing many physical phenomena. Various scientific and engineering applications demand a versatile and differentiable PDE solver that can quickly generate solutions with adequate accuracy, and limitations of the traditional solvers and specialized neural operators motivate the development of foundation models for solving PDEs. This paper introduces PDEformer-2, a versatile foundation model for two-dimensional PDEs. Based on our previous one-dimensional PDEformer-1 model, PDEformer-2 receives the PDE form as network input via computational graph representation, which has the flexibility to encode most common PDEs. The mesh-free predicted solutions can be directly queried at arbitrary spatio-temporal coordinates. A large (40TB) diverse dataset is employed to pretrain the current model, making it capable of simultaneously addressing PDEs with different symbolic forms, domain shapes, boundary conditions, number of variables, and time-dependency. Accurate zero-shot prediction is allowed for PDEs that resemble the pretraining ones. When adapted to new unseen PDEs, PDEformer-2 demonstrates faster learning than many specialized models, and has smaller errors given limited (less than 100) samples. Additionally, PDEformer-2 can be employed in the inverse problems thanks to its fast and differentiable nature and produces reasonable results in our experiments to recover coefficient scalars and fields of a PDE.

math.NA↗

arXiv2025

TimeHC-RL: Temporal-aware Hierarchical Cognitive Reinforcement Learning for Enhancing LLMs' Social Intelligence

Recently, Large Language Models (LLMs) have made significant progress in IQ-related domains that require careful thinking, such as mathematics and coding. However, enhancing LLMs' cognitive development in social domains, particularly from a post-training perspective, remains underexplored. Recognizing that the social world follows a distinct timeline and requires a richer blend of cognitive modes (from intuitive reactions (System 1) and surface-level thinking to deliberate thinking (System 2)) than mathematics, which primarily relies on System 2 cognition (careful, step-by-step reasoning), we introduce Temporal-aware Hierarchical Cognitive Reinforcement Learning (TimeHC-RL) for enhancing LLMs' social intelligence. In our experiments, we systematically explore improving LLMs' social intelligence and validate the effectiveness of the TimeHC-RL method, through five other post-training paradigms and two test-time intervention paradigms on eight datasets with diverse data patterns. Experimental results reveal the superiority of our proposed TimeHC-RL method compared to the widely adopted System 2 RL method. It gives the 7B backbone model wings, enabling it to rival the performance of advanced models like DeepSeek-R1 and OpenAI-O3. Additionally, the systematic exploration from post-training and test-time interventions perspectives to improve LLMs' social intelligence has uncovered several valuable insights.

cs.CL↗

arXiv2025

Act-Adaptive Margin: Dynamically Calibrating Reward Models for Subjective Ambiguity

Currently, most reinforcement learning tasks focus on domains like mathematics and programming, where verification is relatively straightforward. However, in subjective tasks such as role-playing, alignment techniques struggle to make progress, primarily because subjective reward modeling using the Bradley-Terry model faces significant challenges when dealing with ambiguous preferences. To improve reward modeling in subjective tasks, this paper proposes AAM (\textbf{\underline{A}}ct-\textbf{\underline{A}}daptive \textbf{\underline{M}}argin), which enhances reward modeling by dynamically calibrating preference margins using the model's internal parameter knowledge. We design two versions of AAM that efficiently generate contextually-appropriate preference gaps without additional human annotation. This approach fundamentally improves how reward models handle subjective rewards by better integrating generative understanding with preference scoring. To validate AAM's effectiveness in subjective reward modeling, we conduct evaluations on RewardBench, JudgeBench, and challenging role-playing tasks. Results show that AAM significantly improves subjective reward modeling performance, enhancing Bradley-Terry reward models by 2.95\% in general tasks and 4.85\% in subjective role-playing tasks. Furthermore, reward models trained with AAM can help downstream alignment tasks achieve better results. Our test results show that applying rewards generated by AAM-Augmented RM to preference learning techniques (e.g., GRPO) achieves state-of-the-art results on CharacterEval and Charm. Code and dataset are available at https://github.com/calubkk/AAM.

cs.CL↗

arXiv2025

Reverse Preference Optimization for Complex Instruction Following

Instruction following (IF) is a critical capability for large language models (LLMs). However, handling complex instructions with multiple constraints remains challenging. Previous methods typically select preference pairs based on the number of constraints they satisfy, introducing noise where chosen examples may fail to follow some constraints and rejected examples may excel in certain respects over the chosen ones. To address the challenge of aligning with multiple preferences, we propose a simple yet effective method called Reverse Preference Optimization (RPO). It mitigates noise in preference pairs by dynamically reversing the constraints within the instruction to ensure the chosen response is perfect, alleviating the burden of extensive sampling and filtering to collect perfect responses. Besides, reversal also enlarges the gap between chosen and rejected responses, thereby clarifying the optimization direction and making it more robust to noise. We evaluate RPO on two multi-turn IF benchmarks, Sysbench and Multi-IF, demonstrating average improvements over the DPO baseline of 4.6 and 2.5 points (on Llama-3.1 8B), respectively. Moreover, RPO scales effectively across model sizes (8B to 70B parameters), with the 70B RPO model surpassing GPT-4o.

cs.CL↗

arXiv2025

Adaptive and Robust DBSCAN with Multi-agent Reinforcement Learning

DBSCAN, a well-known density-based clustering algorithm, has gained widespread popularity and usage due to its effectiveness in identifying clusters of arbitrary shapes and handling noisy data. However, it encounters challenges in producing satisfactory cluster results when confronted with datasets of varying density scales, a common scenario in real-world applications. In this paper, we propose a novel Adaptive and Robust DBSCAN with Multi-agent Reinforcement Learning cluster framework, namely AR-DBSCAN. First, we model the initial dataset as a two-level encoding tree and categorize the data vertices into distinct density partitions according to the information uncertainty determined in the encoding tree. Each partition is then assigned to an agent to find the best clustering parameters without manual assistance. The allocation is density-adaptive, enabling AR-DBSCAN to effectively handle diverse density distributions within the dataset by utilizing distinct agents for different partitions. Second, a multi-agent deep reinforcement learning guided automatic parameter searching process is designed. The process of adjusting the parameter search direction by perceiving the clustering environment is modeled as a Markov decision process. Using a weakly-supervised reward training policy network, each agent adaptively learns the optimal clustering parameters by interacting with the clusters. Third, a recursive search mechanism adaptable to the data's scale is presented, enabling efficient and controlled exploration of large parameter spaces. Extensive experiments are conducted on nine artificial datasets and a real-world dataset. The results of offline and online tasks show that AR-DBSCAN not only improves clustering accuracy by up to 144.1% and 175.3% in the NMI and ARI metrics, respectively, but also is capable of robustly finding dominant parameters.

cs.LG↗

arXiv2025

Exploring Dual-Iron Atomic Catalysts for Efficient Nitrogen Reduction: A Comprehensive Study on Structural and Electronic Optimization

The nitrogen reduction reaction (NRR), as an efficient and green pathway for ammonia synthesis, plays a crucial role in achieving on-demand ammonia production. This study proposes a novel design concept based on dual-iron atomic sites and nitrogen-boron co-doped graphene catalysts, exploring their high efficiency in NRR. By modulating the N and B co-doped ratios, we found that Fe2N3B@G catalyst exhibited significant activity in the adsorption and hydrogenation of N2 molecules, especially with the lowest free energy (0.32 eV) on NRR distal pathway, showing its excellent nitrogen activation capability and NRR performance. The computed electron localization function, crystal orbital Hamiltonian population, electrostatic potential map revealed that the improved NRR kinetics of Fe2N3B@G catalyst derived by N3B co-doping induced optimization of Fe-Fe electronic environment, regulation of Fe-N bond strength, and the continuous electronic support during the N2 breakage and hydrogenation. In particular, machine learning molecular dynamics (MLMD) simulations were employed to verify the high activity of Fe2N3B@G catalyst in NRR, which reveal that Fe2N3B@G effectively regulates the electron density of Fe-N bond, ensuring the smooth generation and desorption of NH3 molecules and avoiding the competition with hydrogen evolution reaction (HER). Furthermore, the determined higher HER overpotential of Fe2N3B@G catalyst can effectively inhibit the HER and enhance the selectivity toward NRR. In addition, Fe2N3B@G catalyst also showed good thermal stability by MD simulations up to 500 K, offering its feasibility in practical applications. This study demonstrates the superior performance of Fe2N3B@G in nitrogen reduction catalysis, and provides theoretical guidance for atomic catalyst design by the co-doping strategy and in-deep electronic environment modulation.

cond-mat.mtrl-sci↗

arXiv2025

High-Pressure Tuning of Electrical Transport in Freestanding Oxide Films

Electrical transport in oxide thin films under high pressure remains largely unexplored due to the lack of a universal experimental strategy. Here we develop an approach that enables high-pressure transport measurements in freestanding oxide films by enhancing their mechanical robustness and integrating them with nanoscale high-pressure devices. As a demonstration, we investigate the resistivity of perovskite SrIrO3 films under hydrostatic pressure and uncover a pressure-driven semimetal-insulator transition near 2.5 GPa, followed by an insulator-metal transition around 9 GPa. In the monolayer limit, SrIrO3 remains insulating and robust against pressure up to 5.5 GPa. The contrasting pressure-dependent phase diagrams of three- and two-dimensional iridates reveal a strong interplay between dimensionality and hydrostatic pressure in correlated oxides. Our work establishes a general platform for exploring pressure-driven phenomena in low-dimensional quantum materials.

cond-mat.mtrl-sci↗

arXiv2025

Multi-View Attention Syntactic Enhanced Graph Convolutional Network for Aspect-based Sentiment Analysis

Aspect-based Sentiment Analysis (ABSA) is the task aimed at predicting the sentiment polarity of aspect words within sentences. Recently, incorporating graph neural networks (GNNs) to capture additional syntactic structure information in the dependency tree derived from syntactic dependency parsing has been proven to be an effective paradigm for boosting ABSA. Despite GNNs enhancing model capability by fusing more types of information, most works only utilize a single topology view of the dependency tree or simply conflate different perspectives of information without distinction, which limits the model performance. To address these challenges, in this paper, we propose a new multi-view attention syntactic enhanced graph convolutional network (MASGCN) that weighs different syntactic information of views using attention mechanisms. Specifically, we first construct distance mask matrices from the dependency tree to obtain multiple subgraph views for GNNs. To aggregate features from different views, we propose a multi-view attention mechanism to calculate the attention weights of views. Furthermore, to incorporate more syntactic information, we fuse the dependency type information matrix into the adjacency matrices and present a structural entropy loss to learn the dependency type adjacency matrix. Comprehensive experiments on four benchmark datasets demonstrate that our model outperforms state-of-the-art methods. The codes and datasets are available at https://github.com/SELGroup/MASGCN.

cs.CL↗

arXiv2024

TARGA: Targeted Synthetic Data Generation for Practical Reasoning over Structured Data

Semantic parsing, which converts natural language questions into logic forms, plays a crucial role in reasoning within structured environments. However, existing methods encounter two significant challenges: reliance on extensive manually annotated datasets and limited generalization capability to unseen examples. To tackle these issues, we propose Targeted Synthetic Data Generation (TARGA), a practical framework that dynamically generates high-relevance synthetic data without manual annotation. Starting from the pertinent entities and relations of a given question, we probe for the potential relevant queries through layer-wise expansion and cross-layer combination. Then we generate corresponding natural language questions for these constructed queries to jointly serve as the synthetic demonstrations for in-context learning. Experiments on multiple knowledge base question answering (KBQA) datasets demonstrate that TARGA, using only a 7B-parameter model, substantially outperforms existing non-fine-tuned methods that utilize close-sourced model, achieving notable improvements in F1 scores on GrailQA(+7.7) and KBQA-Agent(+12.2). Furthermore, TARGA also exhibits superior sample efficiency, robustness, and generalization capabilities under non-I.I.D. settings.

cs.CL↗

arXiv2024

Structural Entropy Guided Probabilistic Coding

Probabilistic embeddings have several advantages over deterministic embeddings as they map each data point to a distribution, which better describes the uncertainty and complexity of data. Many works focus on adjusting the distribution constraint under the Information Bottleneck (IB) principle to enhance representation learning. However, these proposed regularization terms only consider the constraint of each latent variable, omitting the structural information between latent variables. In this paper, we propose a novel structural entropy-guided probabilistic coding model, named SEPC. Specifically, we incorporate the relationship between latent variables into the optimization by proposing a structural entropy regularization loss. Besides, as traditional structural information theory is not well-suited for regression tasks, we propose a probabilistic encoding tree, transferring regression tasks to classification tasks while diminishing the influence of the transformation. Experimental results across 12 natural language understanding tasks, including both classification and regression tasks, demonstrate the superior performance of SEPC compared to other state-of-the-art models in terms of effectiveness, generalization capability, and robustness to label noise. The codes and datasets are available at https://github.com/SELGroup/SEPC.

cs.AI↗

arXiv2024

Windowed Dictionary Design for Delay-Aware OMP Channel Estimation under Fractional Doppler

Delay-Doppler (DD) signal processing has emerged as a powerful tool for analyzing multipath and time-varying channel effects. Due to the inherent sparsity of the wireless channel in the DD domain, compressed sensing (CS) based techniques, such as orthogonal matching pursuit (OMP), are commonly used for channel estimation. However, many of these methods assume integer Doppler shifts, which can lead to performance degradation in the presence of fractional Doppler. In this paper, we propose a windowed dictionary design technique while we develop a delay-aware orthogonal matching pursuit (DA-OMP) algorithm that mitigates the impact of fractional Doppler shifts on DD domain channel estimation. First, we apply receiver windowing to reduce the correlation between the columns of our proposed dictionary matrix. Second, we introduce a delay-aware interference block to quantify the interference caused by fractional Doppler. This approach removes the need for a pre-determined stopping criterion, which is typically based on the number of propagation paths, in conventional OMP algorithm. Our simulation results confirm the effective performance of our proposed DA-OMP algorithm using the proposed windowed dictionary in terms of normalized mean square error (NMSE) of the channel estimate. In particular, our proposed DA-OMP algorithm demonstrates substantial gains compared to standard OMP algorithm in terms of channel estimation NMSE with and without windowed dictionary.

eess.SP↗

arXiv2024

Solving Integrated Process Planning and Scheduling Problem via Graph Neural Network Based Deep Reinforcement Learning

The Integrated Process Planning and Scheduling (IPPS) problem combines process route planning and shop scheduling to achieve high efficiency in manufacturing and maximize resource utilization, which is crucial for modern manufacturing systems. Traditional methods using Mixed Integer Linear Programming (MILP) and heuristic algorithms can not well balance solution quality and speed when solving IPPS. In this paper, we propose a novel end-to-end Deep Reinforcement Learning (DRL) method. We model the IPPS problem as a Markov Decision Process (MDP) and employ a Heterogeneous Graph Neural Network (GNN) to capture the complex relationships among operations, machines, and jobs. To optimize the scheduling strategy, we use Proximal Policy Optimization (PPO). Experimental results show that, compared to traditional methods, our approach significantly improves solution efficiency and quality in large-scale IPPS instances, providing superior scheduling strategies for modern intelligent manufacturing systems.

math.OC↗