arXiv subjects

Jian Xie

Publications and source records attributed to Jian Xie.

At least 19 recordsLinked to original sources

arXiv2026

GRAIN: Bridging Name and Narrative Shifts in Real-World Graph Reasoning through Invariance-Rewarded Agentic RL

Despite their potential in standardized graph tasks, Large Language Models (LLMs) remain brittle to real-world shifts in node identifiers and task formulation. While deterministic graph tools are invariant to such shifts, extracting topological structures from noisy text is highly fragile for LLMs, which often overfit to surface patterns. Moreover, mitigating these parsing failures via multi-agent systems incurs prohibitive latency. To address this, we propose GRAIN, a single-agent framework optimized via reinforcement learning. GRAIN models reasoning as a semantic parsing and tool-execution pipeline, guided by a Structure Invariance Reward. By validating extracted intermediate graphs against ground-truth topologies, this reward forces the LLM to learn robust text-to-structure mappings rather than memorizing linguistic artifacts. We also introduce GRIT, a benchmark evaluating sensitivity to such linguistic shifts. GRAIN outperforms multi-agent baselines by 16.45\% in accuracy with approximately 24\% lower latency. Furthermore, it demonstrates superior structural generalization, halving the out-of-distribution (OOD) gap of SFT models (from 15.77\% to 7.80\%) and maintaining robustness on large-scale graphs beyond the training distribution.

cs.AI↗

arXiv2026

Emergent trans-moir\'e orbitals and topology in rhombohedral graphene

The fractional quantum anomalous Hall effect (FQAHE) exhibited in fractional Chern insulators has recently been demonstrated in twisted MoTe2 and rhombohedral graphene/hBN moir\'e superlattices, promising new routes toward topological quantum computation. Central to realizing this promise is the understanding of the underlying microscopic mechanism. This, however, remains elusive in the case of rhombohedral graphene, with the crux being its two seemingly paradoxical conditions: a pronounced small-twist-angle ({\theta}) moir\'e interface, yet only when electrons are kept distant from it. Here, by scanning tunnelling microscopic imaging with both conditions fulfilled, we capture dramatic electronic structure reshaping in rhombohedral hexalayer graphene by unforeseen 'trans-moir\'e orbitals', which emerge on the other, distant side of the moir\'e interface but nevertheless enforce the moir\'e periodicity at all measured fillings. We visualize a hierarchy of spatially and energetically distinct trans-moir\'e orbitals which doped electrons must sequentially occupy--the lowest-energy orbital, expectedly responsible for the FQAHE at small fillings, carries a hollow-cage-like shape. Remarkably, these trans-moir\'e orbitals vanish at {\theta} {\gtrsim} 1{\deg}, and so do QAHE plateaus in similar devices. Simulations reveal an interaction-driven charge-redistribution mechanism which shapes the trans-moir\'e orbitals and corresponding Chern minibands. With our findings providing the missing microscopic link, the paradoxical conditions find a natural explanation: electrons are not simply kept distant from a small-{\theta} moir\'e interface; they are forced into topological trans-moir\'e orbitals, forged precisely under such conditions. Our microscopic diagnostics unlocks a wide range of possible 'synthetic' FQAHE platforms.

cond-mat.mes-hall↗

arXiv2026

Agent Learning via Early Experience

A long-term goal of language agents is to learn and improve through their own experience, ultimately outperforming humans in complex, real-world tasks. However, training agents from experience data with reinforcement learning remains difficult in many environments, which either lack verifiable rewards (e.g., websites) or require inefficient long-horizon rollouts (e.g., multi-turn tool use). As a result, most current agents rely on supervised fine-tuning on expert data, which is challenging to scale and generalizes poorly. This limitation stems from the nature of expert demonstrations: they capture only a narrow range of scenarios, and expose the agent to limited environment diversity. We address this limitation with a middle-ground paradigm we call early experience: interaction data generated by the agent's own actions, where the resulting future states serve as supervision without reward signals. Within this paradigm, we study two strategies of using such data: (1) implicit world modeling, which uses collected states to ground the policy in environment dynamics; and (2) self-reflection, where the agent learns from its suboptimal actions to improve reasoning and decision-making. Evaluation across eight diverse environments and multiple model families shows that our approaches consistently improve effectiveness and out-of-domain generalization, highlighting the value of early experience. Moreover, in environments with verifiable rewards, our results provide promising signals that early experience offers a strong foundation for subsequent reinforcement learning, making it a practical bridge between imitation learning and fully experience-driven agents.

cs.AI↗

arXiv2026

QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks

Deep research agents extend the role of search engines from retrieving keyword-matched pages to synthesizing knowledge, fundamentally changing how humans interact with information. However, frontier systems remain proprietary, while existing open agents often generalize poorly across different task types, leaving unclear how to train a broadly capable deep research agent. We release QUEST, a family of open models (ranging from 2B to 35B) that serve as general-purpose deep research agents designed to handle a wide range of long-horizon search tasks, with strong capabilities in fact seeking, citation grounding, and report synthesis. To build QUEST, we propose an effective training recipe combining mid-training, supervised fine-tuning, and reinforcement learning. Central to this recipe is a curated data synthesis pipeline based on unified rubric trees, which applies to different task types and enables synthesizing training data with verifiable rewards without human annotation. In addition, QUEST incorporates a built-in context management mechanism that enables effective long-horizon reasoning and knowledge synthesis. Using only 8K synthesized tasks, QUEST approaches or even surpasses frontier closed-source agents across eight deep research benchmarks spanning diverse task types, and achieves the best overall performance among recent open-weight agents. We released everything: models, data, and training scripts.

cs.CL↗

arXiv2026

MCGA: A Multi-task Classical Chinese Literary Genre Audio Corpus

With the rapid advancement of Multimodal Large Language Models (MLLMs), their potential has gained significant attention in Chinese Classical Studies (CCS). While existing research primarily focuses on text and visual modalities, the audio corpus within this domain remains largely underexplored. To bridge this gap, we introduce the Multi-task Classical Chinese Literary Genre Audio Corpus (MCGA), a 119-hour corpus comprising 22,000 audio samples. It encompasses a diverse range of literary genres across six tasks: Automatic Speech Recognition (ASR), Speech-to-Text Translation (S2TT), Speech Emotion Captioning (SEC), Spoken Question Answering (SQA), Speech Understanding (SU), and Speech Reasoning (SR). Through the evaluation of ten MLLMs, our experimental results demonstrate that current MLLMs still face substantial challenges on the MCGA test set. Furthermore, we introduce a domain-specific metric for SEC and a metric to measure the consistency between speech and text capabilities. We release MCGA to the public to facilitate the development of more robust MLLMs. MCGA Corpus: https://github.com/yxduir/MCGA

cs.CL↗

arXiv2026

The Model Agreed, But Didn't Learn: Diagnosing Surface Compliance in Large Language Models

Large Language Models (LLMs) internalize vast world knowledge as parametric memory, yet inevitably inherit the staleness and errors of their source corpora. Consequently, ensuring the reliability and malleability of these internal representations is imperative for trustworthy real-world deployment. Knowledge editing offers a pivotal paradigm for surgically modifying memory without retraining. However, while recent editors demonstrate high success rates on standard benchmarks, it remains questionable whether current evaluation frameworks that rely on assessing output under specific prompting conditions can reliably authenticate genuine memory modification. In this work, we introduce a simple diagnostic framework that subjects models to discriminative self-assessment under in-context learning (ICL) settings that better reflect real-world application environments, specifically designed to scrutinize the subtle behavioral nuances induced by memory modifications. This probing reveals a pervasive phenomenon of Surface Compliance, where editors achieve high benchmark scores by merely mimicking target outputs without structurally overwriting internal beliefs. Moreover, we find that recursive modifications accumulate representational residues, triggering cognitive instability and permanently diminishing the reversibility of the model's memory state. These insights underscore the risks of current editing paradigms and highlight the pivotal role of robust memory modification in building trustworthy, long-term sustainable LLM systems. Code is available at https://github.com/XiaojieGu/SA-MCQ.

cs.CL↗

arXiv2026

CODA: Difficulty-Aware Compute Allocation for Adaptive Reasoning

The emergence of large reasoning models demonstrates that scaling inference-time compute significantly enhances performance on complex tasks. However, it often falls into another trap: overthinking simple problems, where repetitive rationales yield minimal accuracy gains at a disproportionately high cost. This motivates adaptive reasoning: dynamically aligning reasoning depth with instance difficulty. In this paper, we study adaptive reasoning from an optimality perspective, formalizing it as a utility maximization problem where tokens are allocated until the marginal accuracy gain falls below the incremental cost. Based on this, we propose CODA (Compute Allocation by Difficulty Awareness), a method that operationalizes this principle by allocating tokens via a policy-internal difficulty signal. Specifically, CODA estimates difficulty via group-based rollouts and maps it to two non-negative gates that modulate a length-dependent shaping term on top of the binary base reward. The easy-side gate penalizes verbosity on simple instances, whereas the hard-side gate encourages more deliberative rollouts on challenging ones. Across model scales and benchmarks, CODA achieves adaptive reasoning without external annotations or user-provided budgets: on easy tasks, CODA reduces token costs by over 60% while maintaining strong accuracy, whereas on hard tasks it incentivizes more deliberative rollouts to maximize performance.

cs.CL↗

arXiv2026

FIRE: A Comprehensive Benchmark for Financial Intelligence and Reasoning Evaluation

We introduce FIRE, a comprehensive benchmark designed to evaluate both the theoretical financial knowledge of LLMs and their ability to handle practical business scenarios. For theoretical assessment, we curate a diverse set of examination questions drawn from widely recognized financial qualification exams, enabling evaluation of LLMs deep understanding and application of financial knowledge. In addition, to assess the practical value of LLMs in real-world financial tasks, we propose a systematic evaluation matrix that categorizes complex financial domains and ensures coverage of essential subdomains and business activities. Based on this evaluation matrix, we collect 3,000 financial scenario questions, consisting of closed-form decision questions with reference answers and open-ended questions evaluated by predefined rubrics. We conduct comprehensive evaluations of state-of-the-art LLMs on the FIRE benchmark, including XuanYuan 4.0, our latest financial-domain model, as a strong in-domain baseline. These results enable a systematic analysis of the capability boundaries of current LLMs in financial applications. We publicly release the benchmark questions and evaluation code to facilitate future research.

cs.AI↗

arXiv2026

LLM Agents for Education: Advances and Applications

Large Language Model (LLM) agents are transforming education by automating complex pedagogical tasks and enhancing both teaching and learning processes. In this survey, we present a systematic review of recent advances in applying LLM agents to address key challenges in educational settings, such as feedback comment generation, curriculum design, etc. We analyze the technologies enabling these agents, including representative datasets, benchmarks, and algorithmic frameworks. Additionally, we highlight key challenges in deploying LLM agents in educational settings, including ethical issues, hallucination and overreliance, and integration with existing educational ecosystems. Beyond the core technical focus, we include in Appendix A a comprehensive overview of domain-specific educational agents, covering areas such as science learning, language learning, and professional development.

cs.CY↗

arXiv2026

MMFCTUB: Multi-Modal Financial Credit Table Understanding Benchmark

The advent of multi-modal language models (MLLMs) has spurred research into their application across various table understanding tasks. However, their performance in credit table understanding (CTU) for financial credit review remains largely unexplored due to the following barriers: low data consistency, high annotation costs stemming from domain-specific knowledge and complex calculations, and evaluation paradigm gaps between benchmark and real-world scenarios. To address these challenges, we introduce MMFCTUB (Multi-Modal Financial Credit Table Understanding Benchmark), a practical benchmark, encompassing more than 7,600 high quality CTU samples across 5 table types. MMFCTUB employ a minimally supervised pipeline that adheres to inter-table constraints and maintains data distributions consistency. The benchmark leverages capacity-driven questions and mask-and-recovery strategy to evaluate models' cross-table structure perception, domain knowledge utilization, and numerical calculation capabilities. Utilizing MMFCTUB, we conduct comprehensive evaluations of both proprietary and open-source MLLMs, revealing their strengths and limitations in CTU tasks. MMFCTUB serves as a valuable resource for the research community, facilitating rigorous evaluation of MLLMs in the domain of CTU.

cs.CE↗

arXiv2025

Magnetic-Field-Driven Insulator-Superconductor Transition in Rhombohedral Graphene

Recent studies of rhombohedral multilayer graphene (RMG) have revealed a variety of superconducting states that can be induced or enhanced by magnetic fields, reinforcing RMG as a powerful platform for investigating novel superconductivity. Here we report an insulator-superconductor transition driven by in-plane magnetic fields B|| in rhombohedral hexalayer graphene. The upper critical in-plane field of 2T violates the Pauli limit, and an analysis based on isospin symmetry breaking supports a spin-polarized superconductor. At in-plane B = 0, such spin-polarized superconductor transitions into an insulator, exhibiting a thermally activated gap of 0.1 meV. In addition, we observe four superconducting states in the hole-doped regime, as well as phases with orbital multiferroicity near charge neutrality point. These findings substantially enrich the phase diagram of rhombohedral graphene and provide new insight into the microscopic mechanisms of superconductivity

cond-mat.supr-con↗

arXiv2025

Beware of Reasoning Overconfidence: Pitfalls in the Reasoning Process for Multi-solution Tasks

Large Language Models (LLMs) excel in reasoning tasks requiring a single correct answer, but they perform poorly in multi-solution tasks that require generating comprehensive and diverse answers. We attribute this limitation to \textbf{reasoning overconfidence}: a tendency to express undue certainty in an incomplete solution set. To examine the effect, we introduce \textit{MuSoBench}, a benchmark of multi-solution problems. Experiments show that the conventional short chain-of-thought (Short-CoT) prompting paradigm exhibits pronounced overconfidence, whereas the emerging long chain-of-thought (Long-CoT) approach mitigates it through iterative exploration and self-reflection. We further characterise observable behaviours and influential factors. To probe the underlying cause, we propose the \textbf{cognitive-rigidity hypothesis}, which posits that overconfidence arises when the reasoning process prematurely converges on a narrow set of thought paths. An attention-entropy analysis offers preliminary support for this view. These findings provide tools for assessing the completeness of LLM reasoning and highlight the need to move evaluation beyond single-answer accuracy toward comprehensive exploration.

cs.CL↗

arXiv2025

Enhancing Language Agent Strategic Reasoning through Self-Play in Adversarial Games

Existing language agents often encounter difficulties in dynamic adversarial games due to poor strategic reasoning. To mitigate this limitation, a promising approach is to allow agents to learn from game interactions automatically, without relying on costly expert-labeled data. Unlike static environments where agents receive fixed feedback or rewards, selecting appropriate opponents in dynamic adversarial games can significantly impact learning performance. However, the discussion of opponents in adversarial environments remains an area under exploration. In this paper, we propose a Step-level poliCy Optimization method through Play-And-Learn, SCO-PAL. Leveraging SCO-PAL, we conduct a detailed analysis of opponent selection by setting opponents at different levels and find that self-play is the most effective way to improve strategic reasoning in such adversarial environments. Utilizing SCO-PAL with self-play, we increase the average win rate against four opponents by approximately 30% compared to baselines and achieve a 54.76% win rate against GPT-4 in six adversarial games.

cs.CL↗

arXiv2025

Does Moire Matter? Critical Moire Dependence with Quantum Fluctuations in Graphene Based Integer and Fractional Chern Insulators

Rhombohedral multilayer graphene has emerged as a powerful platform for investigating flat-band-driven correlated phenomena, yet most aspects remain not understood. In this work, we systematically study the moire-dependent band topology in rhombohedral hexalayer graphene. For the first time we demonstrate that the moire twist angle plays a crucial role in the formation of the moire Chern insulators in rhombohedral hexalayer graphene/hexagonal boron nitride (RHG/hBN) moire superlattices. In the moire-distant regime at filling factor v = 1, only systems with a twist angle θ < 1.1° exhibit an integer moire Chern insulator, while the fractional Chern insulator at v = 2/3 requires smaller twist angle to be stabilized. Our theoretical modelling, which includes quantum fluctuations and exact diagonalization results, suggests that mean-field theory, which has been widely adopted, does not explain the twist-angle dependence of the v = 1 phase diagram, and that correlation effects are crucial. Moreover, we realize two distinct stacking configurations ( /Xi=0 and /Xi=1) between graphene and hBN, and find that both cases can yield a Chern insulator at v = 1. Our experimental work upends the current mean-field paradigm, illuminates how quantum fluctuations and moiré effects shape the RHG/hBN phase diagram, and paves the way for future understanding and engineering of topological correlated states in rhombohedral graphene moire systems.

cond-mat.mes-hall↗

arXiv2025

ARM2: Adaptive Reasoning Model with Vision Understanding and Executable Code

Large Reasoning Models (LRMs) often suffer from the ``over-thinking'' problem, generating unnecessarily long reasoning on simple tasks. Some strategies have been proposed to mitigate this issue, such as length penalties or routing mechanisms, but they are typically heuristic and task-specific, lacking a general framework for adaptive reasoning. In this paper, we present ARM2, a unified model that adaptively balances reasoning performance and efficiency across multiple formats through a reinforcement learning framework augmented with length-aware optimization. Beyond conventional natural language inference, ARM2 integrates vision understanding, extending its applicability to multimodal. Moreover, ARM2 integrates executable code into reasoning, enabling substantial reductions in token cost while preserving task performance compared to long CoT. Experiments demonstrate that ARM2 achieves performance on par with traditional reasoning models trained with GRPO, while reducing token usage by over 70% on average. We further conduct extensive analyses to validate the effectiveness of ARM2 and the soundness of its design.

cs.CL↗

arXiv2025

ARM: Adaptive Reasoning Model

While large reasoning models demonstrate strong performance on complex tasks, they lack the ability to adjust reasoning token usage based on task difficulty. This often leads to the "overthinking" problem -- excessive and unnecessary reasoning -- which, although potentially mitigated by human intervention to control the token budget, still fundamentally contradicts the goal of achieving fully autonomous AI. In this work, we propose Adaptive Reasoning Model (ARM), a reasoning model capable of adaptively selecting appropriate reasoning formats based on the task at hand. These formats include three efficient ones -- Direct Answer, Short CoT, and Code -- as well as a more elaborate format, Long CoT. To train ARM, we introduce Ada-GRPO, an adaptation of Group Relative Policy Optimization (GRPO), which addresses the format collapse issue in traditional GRPO. Ada-GRPO enables ARM to achieve high token efficiency, reducing tokens by an average of 30%, and up to 70%, while maintaining performance comparable to the model that relies solely on Long CoT. Furthermore, not only does it improve inference efficiency through reduced token generation, but it also brings a 2x speedup in training. In addition to the default Adaptive Mode, ARM supports two additional reasoning modes: 1) Instruction-Guided Mode, which allows users to explicitly specify the reasoning format via special tokens -- ideal when the appropriate format is known for a batch of tasks. 2) Consensus-Guided Mode, which aggregates the outputs of the three efficient formats and resorts to Long CoT in case of disagreement, prioritizing performance with higher token usage.

cs.CL↗

arXiv2025

UniEDU: A Unified Language and Vision Assistant for Education Applications

Education materials for K-12 students often consist of multiple modalities, such as text and images, posing challenges for models to fully understand nuanced information in these materials. In this paper, we propose a unified language and vision assistant UniEDU designed for various educational applications, including knowledge recommendation, knowledge tracing, time cost prediction, and user answer prediction, all within a single model. Unlike conventional task-specific models, UniEDU offers a unified solution that excels across multiple educational tasks while maintaining strong generalization capabilities. Its adaptability makes it well-suited for real-world deployment in diverse learning environments. Furthermore, UniEDU is optimized for industry-scale deployment by significantly reducing computational overhead-achieving approximately a 300\% increase in efficiency-while maintaining competitive performance with minimal degradation compared to fully fine-tuned models. This work represents a significant step toward creating versatile AI systems tailored to the evolving demands of education.

cs.CL↗

arXiv2025

HierSearch: A Hierarchical Enterprise Deep Search Framework Integrating Local and Web Searches

Recently, large reasoning models have demonstrated strong mathematical and coding abilities, and deep search leverages their reasoning capabilities in challenging information retrieval tasks. Existing deep search works are generally limited to a single knowledge source, either local or the Web. However, enterprises often require private deep search systems that can leverage search tools over both local and the Web corpus. Simply training an agent equipped with multiple search tools using flat reinforcement learning (RL) is a straightforward idea, but it has problems such as low training data efficiency and poor mastery of complex tools. To address the above issue, we propose a hierarchical agentic deep search framework, HierSearch, trained with hierarchical RL. At the low level, a local deep search agent and a Web deep search agent are trained to retrieve evidence from their corresponding domains. At the high level, a planner agent coordinates low-level agents and provides the final answer. Moreover, to prevent direct answer copying and error propagation, we design a knowledge refiner that filters out hallucinations and irrelevant evidence returned by low-level agents. Experiments show that HierSearch achieves better performance compared to flat RL, and outperforms various deep search and multi-source retrieval-augmented generation baselines in six benchmarks across general, finance, and medical domains.

cs.IR↗