Searcharxiv⌕ Search

arXiv subjects

Jiayu Liu

Publications and source records attributed to Jiayu Liu.

At least 55 records · Page 3Linked to original sources
arXiv2026

UniCog: Uncovering Cognitive Abilities of LLMs through Latent Mind Space Analysis

A growing body of research suggests that the cognitive processes of large language models (LLMs) differ fundamentally from those of humans. However, existing interpretability methods remain limited in explaining how cognitive abilities are engaged during LLM reasoning. In this paper, we propose UniCog, a unified framework that analyzes LLM cognition via a latent mind space. Formulated as a latent variable model, UniCog encodes diverse abilities from dense model activations into sparse, disentangled latent dimensions. Through extensive analysis on six advanced LLMs, including DeepSeek-V3.2 and GPT-4o, we reveal a Pareto principle of LLM cognition, where a shared reasoning core is complemented by ability-specific signatures. Furthermore, we discover that reasoning failures often manifest as anomalous intensity in latent activations. These findings opens a new paradigm in LLM analysis, providing a cognition grounded view of reasoning dynamics. Finally, leveraging these insights, we introduce a latent-informed candidate prioritization strategy, which improves reasoning performance by up to 7.5% across challenging benchmarks. Our code is available at https://github.com/milksalute/unicog.

cs.AI↗
arXiv2026

Absence of magnetic order in epitaxial RuO2 revealed by X-ray linear dichroism

Recently, the topic of altermagnetism has attracted tremendous attention and RuO2 have been demonstrated to be one of the most promising altermagnetic candidates. However, disputes still remain on the existence of magnetic order in RuO2. Here in this work, we employ X-ray linear dichroism (XLD), a widely utilized technique for characterizing antiferromagnets, in conjunction with photoemission electron microscopy and multiple scattering calculation to provide clear evidence of the absence of magnetic order in epitaxial RuO2 films. The observed XLD signal is nearly invariant with temperature and independent on cooling field direction, in stark contrast to the substantial magnetic order-related XLD signal predicted by multiple scattering calculation. This finding strongly suggests a nonmagnetic origin for RuO2. Furthermore, we observed significantly distinct XLD signals at the Ru M3 and O K edges in RuO2 films grown on TiO2 substrate with different surface orientations, which can be attributed to the low-symmetry crystal field. These results unequivocally demonstrate the absence of magnetic order in RuO2 and establishes XLD measurement as a robust technique for probing the low-symmetry magnetic materials.

cond-mat.mtrl-sci↗
arXiv2025

Step-DeepResearch Technical Report

As LLMs shift toward autonomous agents, Deep Research has emerged as a pivotal metric. However, existing academic benchmarks like BrowseComp often fail to meet real-world demands for open-ended research, which requires robust skills in intent recognition, long-horizon decision-making, and cross-source verification. To address this, we introduce Step-DeepResearch, a cost-effective, end-to-end agent. We propose a Data Synthesis Strategy Based on Atomic Capabilities to reinforce planning and report writing, combined with a progressive training path from agentic mid-training to SFT and RL. Enhanced by a Checklist-style Judger, this approach significantly improves robustness. Furthermore, to bridge the evaluation gap in the Chinese domain, we establish ADR-Bench for realistic deep research scenarios. Experimental results show that Step-DeepResearch (32B) scores 61.4% on Scale AI Research Rubrics. On ADR-Bench, it significantly outperforms comparable models and rivals SOTA closed-source models like OpenAI and Gemini DeepResearch. These findings prove that refined training enables medium-sized models to achieve expert-level capabilities at industry-leading cost-efficiency.

cs.CL↗
arXiv2025

Mathematical Proof as a Litmus Test: Revealing Failure Modes of Advanced Large Reasoning Models

Large reasoning models (e.g., R1, o3) have demonstrated remarkable mathematical problem-solving abilities. However, the high reported accuracy of these advanced models on popular datasets, reliance on purely numerical evaluation and potential benchmark leakage, often masks their true reasoning shortcomings. To address this, we propose leveraging the inherent rigor and methodological complexity of mathematical proofs as a diagnostic tool to expose these hidden failures. Specifically, we introduce the RFMDataset (Reveal Failure Modes), a collection of 200 diverse mathematical proof problems, and thoroughly evaluate advanced models' performance on it. Our in-depth analysis of their failures uncovers 10 fine-grained error types, which shows fundamental limitations in current large reasoning models: 1) large reasoning models grapple profoundly with mathematical proofs, with some generating entirely correct proofs for less than 20% of problems and failing even on basic ones; 2) models exhibit a diverse spectrum of reasoning failures, prominently demonstrating the lack of guarantees for the correctness and rigor of single-step reasoning; and 3) models show hallucination and incompleteness during the reasoning process. Our findings reveal that models' self-reflection is insufficient to resolve the current logical dilemmas, necessitating formalized and fine-grained logical training.

cs.AI↗
arXiv2025

Verifying Large Language Models' Reasoning Paths via Correlation Matrix Rank

Despite the strong reasoning ability of large language models~(LLMs), they are prone to errors and hallucinations. As a result, how to check their outputs effectively and efficiently has become a critical problem in their applications. Existing checking methods heavily rely on external resources, such as trained verifiers (e.g., process/outcome reward models) or elaborate prompts, which lead to high computational overhead and are only applicable to specific domains. In this paper, we investigate whether the internal behaviors of LLMs have already implied the credibility of their reasoning paths. Specifically, we find that the rank of the correlation matrix between the input problem and the output reasoning path is a robust indicator of reasoning correctness. Different from other correctness indicators for LLMs, the calculation of the correlation matrix only relies on the LLM itself, which avoids the hassle of training a separate model or designing complicated prompts. Based on it, we design a simple, plug-and-play Self-Indicator method to reweight candidate reasoning paths, which achieves significant performance improvements than other voting and verification methods with very few computational overhead. Our experiments across multiple LLMs of varying scales and model families have further shown the effectiveness of Self-Indicator. It achieves over 75% accuracy in distinguishing correct reasoning paths from incorrect ones, and, in turn, improves the accuracies on three reasoning benchmarks by more than 8%.

cs.AI↗
arXiv2025

CritiCal: Can Critique Help LLM Uncertainty or Confidence Calibration?

Accurate confidence calibration in Large Language Models (LLMs) is critical for safe use in high-stakes domains, where clear verbalized confidence enhances user trust. Traditional methods that mimic reference confidence expressions often fail to capture the reasoning needed for accurate confidence assessment. We propose natural language critiques as a solution, ideally suited for confidence calibration, as precise gold confidence labels are hard to obtain and often require multiple generations. This paper studies how natural language critiques can enhance verbalized confidence, addressing: (1) What to critique: uncertainty (question-focused) or confidence (answer-specific)? Analysis shows confidence suits multiple-choice tasks, while uncertainty excels in open-ended scenarios. (2) How to critique: self-critique or critique calibration training? We propose Self-Critique, enabling LLMs to critique and optimize their confidence beyond mere accuracy, and CritiCal, a novel Critique Calibration training method that leverages natural language critiques to improve confidence calibration, moving beyond direct numerical optimization. Experiments show that CritiCal significantly outperforms Self-Critique and other competitive baselines, even surpassing its teacher model, GPT-4o, in complex reasoning tasks. CritiCal also shows robust generalization in out-of-distribution settings, advancing LLM's reliability.

cs.CL↗
arXiv2025

Foundation of Intelligence: Review of Math Word Problems from Human Cognition Perspective

Math word problem (MWP) serves as a fundamental research topic in artificial intelligence (AI) dating back to 1960s. This research aims to advance the reasoning abilities of AI by mirroring the human-like cognitive intelligence. The mainstream technological paradigm has evolved from the early rule-based methods, to deep learning models, and is rapidly advancing towards large language models. However, the field still lacks a systematic taxonomy for the MWP survey along with a discussion of current development trends. Therefore, in this paper, we aim to comprehensively review related research in MWP solving through the lens of human cognition, to demonstrate how recent AI models are advancing in simulating human cognitive abilities. Specifically, we summarize 5 crucial cognitive abilities for MWP solving, including Problem Understanding, Logical Organization, Associative Memory, Critical Thinking, and Knowledge Learning. Focused on these abilities, we review two mainstream MWP models in recent 10 years: neural network solvers, and LLM based solvers, and discuss the core human-like abilities they demonstrated in their intricate problem-solving process. Moreover, we rerun all the representative MWP solvers and supplement their performance on 5 mainstream benchmarks for a unified comparison. To the best of our knowledge, this survey first comprehensively analyzes the influential MWP research of the past decade from the perspective of human reasoning cognition and provides an integrative overall comparison across existing approaches. We hope it can inspire further research in AI reasoning. Our repository is released on https://github.com/Ljyustc/FoI-MWP.

cs.AI↗
arXiv2025

DixitWorld: Evaluating Multimodal Abductive Reasoning in Vision-Language Models with Multi-Agent Dixit Gameplay

Multimodal abductive reasoning--the generation and selection of explanatory hypotheses from partial observations--is a cornerstone of intelligence. Current evaluations of this ability in vision-language models (VLMs) are largely confined to static, single-agent tasks. Inspired by Dixit, we introduce DixitWorld, a comprehensive evaluation suite designed to deconstruct this challenge. DIXITWORLD features two core components: DixitArena, a dynamic, multi-agent environment that evaluates both hypothesis generation (a "storyteller" crafting cryptic clues) and hypothesis selection ("listeners" choosing the target image from decoys) under imperfect information; and DixitBench, a static QA benchmark that isolates the listener's task for efficient, controlled evaluation. Results from DixitArena reveal distinct, role-dependent behaviors: smaller open-source models often excel as creative storytellers, producing imaginative yet less discriminative clues, whereas larger proprietary models demonstrate superior overall performance, particularly as listeners. Performance on DixitBench strongly correlates with listener results in DixitArena, validating it as a reliable proxy for hypothesis selection. Our findings reveal a key trade-off between generative creativity and discriminative understanding in multimodal abductive reasoning, a central challenge for developing more balanced and capable vision-language agents.

cs.AI↗
arXiv2025

Surface reconstruction and orthogonal decoupling in SrAl4 and EuAl4

Surface-induced symmetry breaking in quantum materials can stabilize exotic electronic phases distinct from those in the bulk, yet its momentum-space manifestations remain elusive due to domain-averaging effects. Here, using angle-resolved photoemission spectroscopy (ARPES) and scanning tunneling microscopy (STM), we present a microscopic investigation of the electronic structures of SrAl4 and EuAl4, layered tetragonal intermetallic compounds that exhibit well-characterized incommensurate charge-density-wave (CDW) transitions. Below the CDW transition temperatures, we uncover linearly dispersing electronic states and pronounced unidirectional replica bands orthogonal to the bulk CDW wave vector, evidencing the emergence of an in-plane C4 symmetry-breaking electronic order that is not dictated by the bulk incommensurate CDW. STM measurements further reveal a 1 times 2 surface reconstruction with quasi-one-dimensional modulations and half-unit-cell steps, traced to ordered 50 percent Sr/Eu vacancies, which vanish irreversibly upon thermal cycling, indicating decoupled surface and bulk orders. These findings establish SrAl4 and EuAl4 as model platforms for exploring surface-confined nematicity and emergent low-dimensional phases in quantum materials.

cond-mat.mtrl-sci↗
arXiv2025

Step-3 is Large yet Affordable: Model-system Co-design for Cost-effective Decoding

Large language models (LLMs) face low hardware efficiency during decoding, especially for long-context reasoning tasks. This paper introduces Step-3, a 321B-parameter VLM with hardware-aware model-system co-design optimized for minimizing decoding costs. Step-3 innovates in two key dimensions: (1) A novel Multi-Matrix Factorization Attention (MFA) mechanism that significantly reduces both KV cache size and computation while maintaining high attention expressiveness, and (2) Attention-FFN Disaggregation (AFD), a distributed inference system that decouples attention and Feed-Forward Network (FFN) layers into specialized subsystems. This co-design achieves unprecedented cost efficiency: Step-3 significantly reduces theoretical decoding costs compared with models like DeepSeek-V3 and Qwen3 MoE 235B, with the gains widening at longer context. Step-3 achieves low cost while activating 38B parameters per token (more than DeepSeek-V3 and Qwen3 MoE 235B), demonstrating that hardware-aligned attention arithmetic intensity, MoE sparsity, and AFD are critical to cost-effectiveness. We perform a head-to-head comparison with DeepSeek-V3 in its favorable scenarios. Our implementation on Hopper GPUs achieves a decoding throughput of up to 4,039 tokens per second per GPU under 50ms TPOT SLA (4K context, FP8, no MTP). It is higher than DeepSeek-V3's 2,324 in the same setup and sets a new Pareto frontier for LLM decoding.

cs.LG↗
arXiv2025

Blind and Topological Interference Managements for Bistatic Integrated Sensing and Communication

Integrated sensing and communication (ISAC) systems provide significant enhancements in performance and resource efficiency compared to individual sensing and communication systems, primarily attributed to the collaborative use of wireless resources, radio waveforms, and hardware platforms. This paper focuses on the bistatic ISAC systems with dispersed multi-receiver and one sensor. Compared to a monostatic ISAC system, the main challenge in the bistatic setting is that the information messages are unknown to the sensor and therefore they are seen as interference, while the channel between the transmitters (TX) and the sensor is unknown to the transmitters. In order to mitigate the interference at the sensor while maximizing the communication degree of freedom, we introduce two strategies, namely, blind interference alignment and topological interference management. Although well-known in the context of Gaussian interference channels, these strategies are novel in the context of bistatic ISAC. For the bistatic ISAC models with heterogeneous coherence times or with heterogeneous connectivity, the achieved ISAC tradeoff points in terms of communication and sensing degrees of freedom are characterized. In particular, we show that the new tradeoff outperforms the time-sharing between the sensing-only and the communication-only schemes. Simulation results demonstrate that the proposed schemes significantly improve the channel estimation error for the sensing task, compared to treating interference as noise at the sensor and successive interference cancellation.

cs.IT↗
arXiv2025

Advancing Tool-Augmented Large Language Models via Meta-Verification and Reflection Learning

Empowering large language models (LLMs) with effective tool utilization capabilities is crucial for enabling AI agents to solve complex problems. However, current models face two major limitations: (1) unreliable tool planning and invocation due to low-quality instruction datasets (e.g., widespread hallucinated API calls), and (2) weak tool reflection abilities (over 90% of errors cannot be corrected) resulting from static imitation learning. To address these critical limitations, we propose Tool-MVR, a novel Tool-Augmented LLM that achieves comprehensive System 2 reasoning through two key innovations. Specifically, we first introduce Multi-Agent Meta-Verification (MAMV), a systematic pipeline that rigorously validates APIs, queries, and reasoning trajectories to construct ToolBench-V, a new high-quality instruction dataset that addresses the limitation of unreliable tool planning and invocation. Second, we propose Exploration-based Reflection Learning (EXPLORE), which enhances tool reflection capabilities by leveraging tool feedback through a dynamic "Error -> Reflection -> Correction" learning paradigm, resulting in our reflection dataset ToolBench-R and addressing the critical weakness in tool reflection. Finally, we obtain Tool-MVR by finetuning open-source LLMs (e.g., Qwen-7B) on both ToolBench-V and ToolBench-R. Our experiments demonstrate that Tool-MVR achieves state-of-the-art performance on StableToolBench, surpassing both ToolLLM (by 23.9%) and GPT-4 (by 15.3%) while reducing API calls by 31.4%, with strong generalization capabilities across unseen tools and scenarios. Additionally, on our proposed RefineToolBench, the first benchmark specifically designed to evaluate tool reflection capabilities, Tool-MVR achieves a 58.9% error correction rate, significantly outperforming ToolLLM's 9.1%.

cs.CL↗
arXiv2025

CogMath: Assessing LLMs' Authentic Mathematical Ability from a Human Cognitive Perspective

Although large language models (LLMs) show promise in solving complex mathematical tasks, existing evaluation paradigms rely solely on a coarse measure of overall answer accuracy, which are insufficient for assessing their authentic capabilities. In this paper, we propose \textbf{CogMath}, which comprehensively assesses LLMs' mathematical abilities through the lens of human cognition. Specifically, inspired by psychological theories, CogMath formalizes human reasoning process into 3 stages: \emph{problem comprehension}, \emph{problem solving}, and \emph{solution summarization}. Within these stages, we investigate perspectives such as numerical calculation, knowledge, and counterfactuals, and design a total of 9 fine-grained evaluation dimensions. In each dimension, we develop an ``\emph{Inquiry}-\emph{Judge}-\emph{Reference}'' multi-agent system to generate inquiries that assess LLMs' mastery from this dimension. An LLM is considered to truly master a problem only when excelling in all inquiries from the 9 dimensions. By applying CogMath on three benchmarks, we reveal that the mathematical capabilities of 7 mainstream LLMs are overestimated by 30\%-40\%. Moreover, we locate their strengths and weaknesses across specific stages/dimensions, offering in-depth insights to further enhance their reasoning abilities.

cs.AI↗
arXiv2025

Observation of multiple surface states in naturally cleavable chiral crystal PdSbSe

Chiral multifold fermions in solids exhibit unique band structures and topological properties, making them ideal for exploring fundamental physical phenomena related to nontrivial topology, chirality, and symmetry breaking. However, the challenge of obtaining clean, flat surfaces through cleavage has hindered the investigation of their unique electronic states. In this study, we utilize high-resolution angle-resolved photoemission spectroscopy and density functional theory calculations to investigate the low-energy electronic structure of the cleavable single-crystal PdSbSe. Our combined experimental and theoretical analysis reveals the presence of multifold degenerate fermions within this chiral crystal. We also observe multiple chiral Fermi arc surface states and spin-splitting behavior in the associated bulk bands. These findings provide unique insights into chiral, multifold fermionic states in easily cleavable crystals and offer a robust platform for further research into their unique electronic properties and potential applications in novel electronic devices.

cond-mat.mtrl-sci↗
arXiv2025

Three-dimensional flat band in ultra-thin Kagome metal Mn3Sn film

Flat bands with small energy dispersion can give rise to strongly correlated electronic and topological phases, especially when located at the Fermi level. Whilst flat bands have been experimentally realized in two-dimensional (2D) twisted van der Waals heterostructures, they are highly sensitive to twist angle, necessitating complex fabrication techniques. Geometrically frustrated kagome lattices have emerged as an attractive platform as they natively host flat bands that have been observed experimentally in quasi-2D bulk-crystal kagome metals. An outstanding experimental question is whether flat bands can be realized in atomically thin metals, with opportunities for stronger electron-electron interactions through tuning of the surrounding dielectric environment. Here we use angle-resolved photoelectron spectroscopy, scanning tunnelling microscopy and band structure calculations to show that ultra-thin films of the kagome metal Mn3Sn host a robust dispersionless flat band with a bandwidth of 50 meV. Furthermore, we demonstrate chemical tuning of the flat band to near the Fermi level via manganese defect engineering. The realization of tunable kagome-derived flat bands in an ultra-thin kagome metal, represents a promising platform to study strongly correlated and topological phenomena, with applications in quantum computing, spintronics and low-energy electronics.

cond-mat.str-el↗
arXiv2025

Data-and-Semantic Dual-Driven Spectrum Map Construction for 6G Spectrum Management

Spectrum maps reflect the utilization and distribution of spectrum resources in the electromagnetic environment, serving as an effective approach to support spectrum management. However, the construction of spectrum maps in urban environments is challenging because of high-density connection and complex terrain. Moreover, the existing spectrum map construction methods are typically applied to a fixed frequency, which cannot cover the entire frequency band. To address the aforementioned challenges, a UNet-based data-and-semantic dual-driven method is proposed by introducing the semantic knowledge of binary city maps and binary sampling location maps to enhance the accuracy of spectrum map construction in complex urban environments with dense communications. Moreover, a joint frequency-space reasoning model is exploited to capture the correlation of spectrum data in terms of space and frequency, enabling the realization of complete spectrum map construction without sampling all frequencies of spectrum data. The simulation results demonstrate that the proposed method can infer the spectrum utilization status of missing frequencies and improve the completeness of the spectrum map construction. Furthermore, the accuracy of spectrum map construction achieved by the proposed data-and-semantic dual-driven method outperforms the benchmark schemes, especially in scenarios with low sampling density.

cs.LG↗
arXiv2024

SOR-like iteration and FPI are consistent when they are equipped with certain optimal iterative parameters

Two common methods for solving absolute value equations (AVE) are SOR-like iteration method and fixed point iteration (FPI) method. In this paper, novel convergence analysis, which result wider convergence range, of the SOR-like iteration and the FPI are given. Based on the new analysis, a new optimal iterative parameter with a analytical form is obtained for the SOR-like iteration. In addition, an optimal iterative parameter with a analytical form is also obtained for FPI. Surprisingly, the SOR-like iteration and the FPI are the same whenever they are equipped with our optimal iterative parameters. As a by product, we give two new constructive proof for a well known sufficient condition such that AVE has a unique solution for any right hand side. Numerical results demonstrate our claims.

math.NA↗