arXiv subjects

Xiaoyuan Li

Publications and source records attributed to Xiaoyuan Li.

At least 19 recordsLinked to original sources

arXiv2026

Adaptive Multi-Head Finite-State Gamblers

Multi-head finite-state dimensions and predimensions quantify the predictability of a sequence by a gambler with trailing heads acting as "probes to the past." These additional heads allow the gambler to exploit patterns that are simple but non-local, such as in a sequence $S$ with $S[n]=S[2n]$ for all $n$. In the original definitions of Huang, Li, Lutz, and Lutz (2025), the head movements were required to be oblivious (i.e., data-independent). Here, we introduce a model in which head movements are adaptive (i.e., data-dependent) and compare it to the oblivious model. We establish that for each $h\geq 2$, adaptivity enhances the predictive power of $h$-head finite-state gamblers, in the sense that there are sequences whose oblivious $h$-head finite-state predimensions strictly exceed their adaptive $h$-head finite-state predimensions. We further prove that adaptive finite-state predimensions admit a strict hierarchy as the number of heads increases, and in fact that for all $h\geq 1$ there is a sequence whose adaptive $(h+1)$-head finite-state predimension is strictly less than its adaptive $h$-head predimension.

cs.IT↗

arXiv2026

ARES: Automated Rubric Synthesis for Scalable LLM Reinforcement Learning

Rubric-based rewards offer a promising way to extend reinforcement learning (RL) for large language models beyond tasks with automatically verifiable answers. However, scaling rubric-based RL remains challenging: existing approaches often rely on expert-written rubrics and manually constructed question sets, while fixed task-level rubrics may fail to capture the evaluation requirements of individual questions. We propose ARES (Automated Rubric synthEsis for Scalable RL), a framework for automatically constructing rubric-based RL data at scale. Starting from raw pretraining documents, ARES converts source knowledge into self-contained question-answer pairs and co-generates question-specific weighted rubrics, enabling instance-level reward supervision for open-ended responses. To improve diversity and quality, ARES conditions generation on domain labels and persona information, and applies validation filters for question self-containment, answer faithfulness, and rubric validity. Using ARES, we construct 100K rubric-annotated instances across ten domains. Experiments on seven benchmarks show that rubric-based RL trained with ARES, outperforms continual pretraining, supervised fine-tuning, and binary-reward RL, with the largest gains on multi-dimensional open-ended tasks such as healthcare and instruction following.

cs.CL↗

arXiv2026

MTR-Bench: A Comprehensive Benchmark for Multi-Turn Reasoning Evaluation

Recent advances in Large Language Models (LLMs) have shown promising results in complex reasoning tasks. However, current evaluations predominantly focus on single-turn reasoning scenarios, leaving interactive tasks largely unexplored. We attribute it to the absence of comprehensive datasets and scalable automatic evaluation protocols. To fill these gaps, we present MTR-Bench for LLMs' Multi-Turn Reasoning evaluation. Comprising 4 classes, 40 tasks, and 3600 instances, MTR-Bench covers diverse reasoning capabilities, fine-grained difficulty granularity, and necessitates multi-turn interactions with the environments. Moreover, MTR-Bench features fully-automated framework spanning both dataset constructions and model evaluations, which enables scalable assessment without human interventions. Extensive experiments reveal that even the cutting-edge reasoning models fall short of multi-turn, interactive reasoning tasks. And the further analysis upon these results brings valuable insights for future research in interactive AI systems.

cs.CL↗

arXiv2026

Unified Data Selection for LLM Reasoning

Effectively training Large Language Models (LLMs) for complex, long-CoT reasoning is often bottlenecked by the need for massive high-quality reasoning data. Existing methods are either computationally expensive or fail to reliably distinguish high- from low-quality reasoning samples. To address this, we propose High-Entropy Sum (HES), a training-free metric that quantifies reasoning quality by summing only the entropy of the top (e.g., 0.5\%) highest-entropy tokens in each reasoning sample. We validate HES across three mainstream training paradigms: Supervised Fine-tuning (SFT), Rejection Fine-tuning (RFT), and Reinforcement Learning (RL), with extensive results demonstrating its consistent effectiveness and significantly reduced computational overhead. In SFT, training on the top 20\% HES-ranked data matches full-dataset performance, while using the lowest-HES data degrades it. In RFT, our HES-based training approach significantly outperforms baseline methods. In RL, HES-selected successful trajectories enable the model to learn strong reasoning patterns, significantly surpassing other compared methods. Our findings establish HES as a robust, training-free metric that enables a unified, effective, and efficient method for developing advanced reasoning in LLMs.

cs.CL↗

arXiv2026

On Predicting the Post-training Potential of Pre-trained LLMs

The performance of Large Language Models (LLMs) on downstream tasks is fundamentally constrained by the capabilities acquired during pre-training. However, traditional benchmarks like MMLU often fail to reflect a base model's plasticity in complex open-ended scenarios, leading to inefficient model selection. We address this by introducing a new task of predicting post-training potential - forecasting a base model's performance before post-training. We propose RuDE (Rubric-based Discriminative Evaluation), a unified framework that bypasses the generation gap of base models by leveraging response discrimination. Guided by our systematic 4C Taxonomy, RuDE constructs controlled contrastive pairs across diverse domains by fine-grained rubric violations. Extensive experiments demonstrate a correlation greater than 90% with post-training performance. Crucially, validation via Reinforcement Learning (RL) confirms that RuDE effectively identifies high-potential smaller models that outperform larger counterparts, offering a compute-efficient mechanism for foundation model development.

cs.CL↗

arXiv2026

SAGE: Scalable Automated Robustness Augmentation for LLM Knowledge Evaluation

Large Language Models (LLMs) achieve strong performance on standard knowledge evaluation benchmarks, yet recent work shows that their knowledge capabilities remain brittle under question variants that test the same knowledge in different forms. Robustness augmentation of existing knowledge evaluation benchmarks is therefore necessary, but current LLM-assisted generate-then-verify pipelines are costly and difficult to scale due to low-yield variant generation and unreliable variant verification. We propose SAGE (Scalable Automated Generation of Robustness BEnchmarks), a framework for scalable robustness augmentation of knowledge evaluation benchmarks using fine-tuned smaller models. SAGE consists of VariantQual, a rubric-based verifier trained on human-labeled seed data, and VariantGen, a variant generator initialized with supervised fine-tuning and further optimized with reinforcement learning using VariantQual as the reward model. Experiments on HellaSwag show that SAGE constructs a large-scale robustness-augmented benchmark with quality comparable to the human-annotated HellaSwag-Pro at substantially lower cost, while the fine-tuned models further generalize to MMLU without benchmark-specific fine-tuning.

cs.CL↗

arXiv2026

SkillGraph: Skill-Augmented Reinforcement Learning for Agents via Evolving Skill Graphs

Skill libraries enable large language model agents to reuse experience from past interactions, but most existing libraries store skills as isolated entries and retrieve them only by semantic similarity. This leads to two key challenges for compositional tasks. Firstly, an agent must identify not only relevant skills but also how they depend on and build upon each other. Secondly, it also makes library maintenance difficult, since the system lacks structural cues for deciding when skills should be merged, split, or removed. We propose SKILLGRAPH, a framework that represents reusable skills as nodes in a directed graph, with typed edges encoding prerequisite, enhancement, and co-occurrence relations. Given a new task, SKILLGRAPH retrieves not just individual skills, but an ordered skill subgraph that can guide multi-step decision making. The graph is continuously updated from agent trajectories and reinforcement learning feedback, allowing both the skill library and the agent policy to improve together. Experiments on ALFWorld, WebShop, and seven search-augmented QA tasks show that SKILLGRAPH achieves state-of-the-art performance against memory-augmented RL methods, with especially large gains on complex tasks that require composing multiple skills.

cs.CL↗

arXiv2026

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

Brain-Muscle Atlas: A novel framework for Motor Brain-Computer Interfaces

Motor brain-computer interfaces (BCIs) enable the control of external devices by decoding neural signals. However, most existing systems rely on a direct "brain-machine" mapping, overlooking the hierarchical physiological pathway of natural movement, namely the "brain-muscle-joint" cascade. Due to the lack of explicit modeling and enhancement of this pathway, current systems are often constrained by the low amplitude and high noise of EEG signals, resulting in motor outputs that are unstable, discontinuous, and insufficiently natural.To address these limitations, this study introduces the concept of a brain-muscle atlas, designed to systematically characterize the mapping between motor cortical activity and corresponding muscle activation, thereby establishing a movement decoding framework that better aligns with neuromuscular physiology. Using synchronously recorded EEG-EMG data, we constructed the first brain-muscle atlas for elbow flexion-extension, achieving a structured mapping from cortical activity to muscle activation.Offline experiments demonstrate that the proposed atlas accurately reconstructs the temporal activation patterns of primary elbow agonists, achieving a maximum correlation coefficient of 0.8314, thereby validating its ability to capture cortical-muscular mapping. Furthermore, by leveraging atlas-derived muscle activation representations, we enabled continuous real-time control of a virtual elbow joint. All ten participants successfully completed the online flexion-extension task, indicating that the system robustly extracts motor intent even under low-SNR EEG conditions.

cs.HC↗

arXiv2025

System Filter-Based Common Components Modeling for Cross-Subject EEG Decoding

Brain-computer interface (BCI) technology enables direct communication between the brain and external devices through electroencephalography (EEG) signals. However, existing decoding models often mix common and personalized components, leading to interference from individual variability that limits cross-subject decoding performance. To address this issue, this paper proposes a system filter that extends the concept of signal filtering to the system level. The method expands a system into its spectral representation, selectively removes unnecessary components, and reconstructs the system from the retained target components, thereby achieving explicit system-level decomposition and filtering. We further integrate the system filter into a Cross-Subject Decoding framework based on the System Filter (CSD-SF) and evaluate it on the four-class motor imagery (MI) task of the BCIC IV 2a dataset. Personalized models are transformed into relation spectrums, and statistical testing across subjects is used to remove personalized components. The remaining stable relations, representing common components across subjects, are then used to construct a common model for cross-subject decoding. Experimental results show an average improvement of 3.28% in decoding accuracy over baseline methods, demonstrating that the proposed system filter effectively isolates stable common components and enhances model robustness and generalizability in cross-subject EEG decoding.

q-bio.NC↗

arXiv2025

Memory-DD: A Low-Complexity Dendrite-Inspired Neuron for Temporal Prediction Tasks

Dendrite-inspired neurons have been widely used in tasks such as image classification due to low computational complexity and fast inference speed. Temporal data prediction, as a key machine learning task, plays a key role in real-time scenarios such as sensor data analysis, financial forecasting, and urban traffic management. However, existing dendrite-inspired neurons are mainly designed for static data. Studies on capturing dynamic features and modeling long-term dependencies in temporal sequences remain limited. Efficient architectures specifically designed for temporal sequence prediction are still lacking. In this paper, we propose Memory-DD, a low-complexity dendrite-inspired neuron model. Memory-DD consists of two dendrite-inspired neuron groups that contain no nonlinear activation functions but can still realize nonlinear mappings. Compared with traditional neurons without dendritic functions, Memory-DD requires only two neuron groups to extract logical relationships between features in input sequences. This design effectively captures temporal dependencies and is suitable for both classification and regression tasks on sequence data. Experimental results show that Memory-DD achieves an average accuracy of 89.41% on 18 temporal classification benchmark datasets, outperforming LSTM by 4.25%. On 9 temporal regression datasets, it reaches comparable performance to LSTM, while using only 50% of the parameters and reducing computational complexity (FLOPs) by 27.7%. These results demonstrate that Memory-DD successfully extends the low-complexity advantages of dendrite-inspired neurons to temporal prediction, providing a low-complexity and efficient solution for time-series data processing.

cs.NE↗

arXiv2025

MathOPEval: A Fine-grained Evaluation Benchmark for Visual Operations of MLLMs in Mathematical Reasoning

Recent progress in Multi-modal Large Language Models (MLLMs) has enabled step-by-step multi-modal mathematical reasoning by performing visual operations based on the textual instructions. A promising approach uses code as an intermediate representation to precisely express and manipulate the images in the reasoning steps. However, existing evaluations focus mainly on text-only reasoning outputs, leaving the MLLM's ability to perform accurate visual operations via code largely unexplored. This work takes a first step toward addressing that gap by evaluating MLLM's code-based capabilities in multi-modal mathematical reasoning.Specifically, our framework focuses on two key evaluation aspects: (1) Multi-modal Code Generation (MCG) evaluates the model's ability to accurately understand and construct visualizations from scratch. (2) Multi-modal Code Editing (MCE) assesses the model's capacity for fine-grained operations, which include three types: Deletion, Modification and Annotation. To evaluate the above tasks, we incorporate a dataset that covers the five most popular types of mathematical figures, including geometric diagrams, function plots, and three types of statistical charts, to provide a comprehensive and effective measurement of existing MLLMs. Our experimental evaluation involves nine mainstream MLLMs, and the results reveal that existing models still lag significantly behind human performance in performing fine-grained visual operations.

cs.CL↗

arXiv2025

Teaching LLM to Reason: Reinforcement Learning from Algorithmic Problems without Code

Enhancing reasoning capabilities remains a central focus in the LLM reasearch community. A promising direction involves requiring models to simulate code execution step-by-step to derive outputs for given inputs. However, as code is often designed for large-scale systems, direct application leads to over-reliance on complex data structures and algorithms, even for simple cases, resulting in overfitting to algorithmic patterns rather than core reasoning structures. To address this, we propose TeaR, which aims at teaching LLMs to reason better. TeaR leverages careful data curation and reinforcement learning to guide models in discovering optimal reasoning paths through code-related tasks, thereby improving general reasoning abilities. We conduct extensive experiments using two base models and three long-CoT distillation models, with model sizes ranging from 1.5 billion to 32 billion parameters, and across 17 benchmarks spanning Math, Knowledge, Code, and Logical Reasoning. The results consistently show significant performance improvements. Notably, TeaR achieves a 35.9% improvement on Qwen2.5-7B and 5.9% on R1-Distilled-7B.

cs.CL↗

arXiv2025

HellaSwag-Pro: A Large-Scale Bilingual Benchmark for Evaluating the Robustness of LLMs in Commonsense Reasoning

Large language models (LLMs) have shown remarkable capabilities in commonsense reasoning; however, some variations in questions can trigger incorrect responses. Do these models truly understand commonsense knowledge, or just memorize expression patterns? To investigate this question, we present the first extensive robustness evaluation of LLMs in commonsense reasoning. We introduce HellaSwag-Pro, a large-scale bilingual benchmark consisting of 11,200 cases, by designing and compiling seven types of question variants. To construct this benchmark, we propose a two-stage method to develop Chinese HellaSwag, a finely annotated dataset comprising 12,000 instances across 56 categories. We conduct extensive experiments on 41 representative LLMs, revealing that these LLMs are far from robust in commonsense reasoning. Furthermore, this robustness varies depending on the language in which the LLM is tested. This work establishes a high-quality evaluation benchmark, with extensive experiments offering valuable insights to the community in commonsense reasoning for LLMs.

cs.CL↗

arXiv2024

Evaluating Mathematical Reasoning of Large Language Models: A Focus on Error Identification and Correction

The rapid advancement of Large Language Models (LLMs) in the realm of mathematical reasoning necessitates comprehensive evaluations to gauge progress and inspire future directions. Existing assessments predominantly focus on problem-solving from the examinee perspective, overlooking a dual perspective of examiner regarding error identification and correction. From the examiner perspective, we define four evaluation tasks for error identification and correction along with a new dataset with annotated error types and steps. We also design diverse prompts to thoroughly evaluate eleven representative LLMs. Our principal findings indicate that GPT-4 outperforms all models, while open-source model LLaMA-2-7B demonstrates comparable abilities to closed-source models GPT-3.5 and Gemini Pro. Notably, calculation error proves the most challenging error type. Moreover, prompting LLMs with the error types can improve the average correction accuracy by 47.9\%. These results reveal potential directions for developing the mathematical reasoning abilities of LLMs. Our code and dataset is available on https://github.com/LittleCirc1e/EIC.

cs.CL↗

arXiv2023

Self-Paced Neutral Expression-Disentangled Learning for Facial Expression Recognition

The accuracy of facial expression recognition is typically affected by the following factors: high similarities across different expressions, disturbing factors, and micro-facial movement of rapid and subtle changes. One potentially viable solution for addressing these barriers is to exploit the neutral information concealed in neutral expression images. To this end, in this paper we propose a self-Paced Neutral Expression-Disentangled Learning (SPNDL) model. SPNDL disentangles neutral information from facial expressions, making it easier to extract key and deviation features. Specifically, it allows to capture discriminative information among similar expressions and perceive micro-facial movements. In order to better learn these neutral expression-disentangled features (NDFs) and to alleviate the non-convex optimization problem, a self-paced learning (SPL) strategy based on NDFs is proposed in the training stage. SPL learns samples from easy to complex by increasing the number of samples selected into the training process, which enables to effectively suppress the negative impacts introduced by low-quality samples and inconsistently distributed NDFs. Experiments on three popular databases (i.e., CK+, Oulu-CASIA, and RAF-DB) show the effectiveness of our proposed method.

cs.CV↗

arXiv2022

Mechanism, measurement, and quantification of stress in decision process: a model based systematic-review protocol

Every human action begins with decision-making. Stress is a significant source of biases that can influence human decision-making. In order to understand the relationship between stress and decision-making, stress quantification is fundamental. Different methods of measuring and quantifying stress in decision-making have been described in the literature while an up-to-date systematic review of the existing methods is lacking. Moreover, mental stress, mental effort, cognitive workload, and workload are often used interchangeably but should be distinguished to enable in-depth investigations of decision-mechanisms. Our objectives are to clarify stress related concepts and review the measurement, quantification, and application of stress during decision making activities.

cs.HC↗

arXiv2018

Real-Time Computability of Real Numbers by Chemical Reaction Networks

We explore the class of real numbers that are computed in real time by deterministic chemical reaction networks that are integral in the sense that all their reaction rate constants are positive integers. We say that such a reaction network computes a real number $α$ in real time if it has a designated species $X$ such that, when all species concentrations are set to zero at time $t = 0$, the concentration $x(t)$ of $X$ is within $2^{-t}$ of $|α|$ at all times $t \ge 1$, and the concentrations of all other species are bounded. We show that every algebraic number and some transcendental numbers are real time computable by chemical reaction networks in this sense. We discuss possible implications of this for the 1965 Hartmanis-Stearns conjecture, which says that no irrational algebraic number is real time computable by a Turing machine.

cs.CC↗