arXiv subjects

Kaixuan Fan

Publications and source records attributed to Kaixuan Fan.

17 recordsLinked to original sources

arXiv2026

Where a New Concept Must Enter: Entry Point Gates Cross-Task Usability in Unified Multimodal Models

Unified multimodal models (UMMs) are motivated by the hope that understanding and generation reinforce each other but controlled ablations repeatedly find that adding a generation objective leaves understanding flat. Joint-training studies cannot settle the disagreement: with overlapping supervision, a gain cannot be attributed to the architecture rather than the data. To further investigate the relationship between the two directions in UMMs, we separate them by construction. A novel visual entity, a rendered 3D asset paired with a pseudo-word screened for absence from the frozen model's behavior, is bound through exactly one task direction, and the untrained direction is then measured. We find that the channel is real in both directions, but the directions differ in kind: generation training installs a name the model can only match among candidates; understanding training installs one it can also produce. What governs cross-task usability is where the binding enters the shared computation. An alignment probe predicts export across 36 configurations (Spearman $\rho = +0.68$). That objective's alignment term, maximized in closed form over activations with every weight frozen, makes a concept drawable when injected at layer 7 of 28 and is indistinguishable from the base model from layer 14 on, while the weight-based version of the same edit peaks at layers 10-14. In an observational series of four models, this window appears only where the understanding pathway is a semantic vision encoder, suggesting that unified weights are not enough: the two directions must share a semantic format at the entry point. Exploiting the rule, a mid-stack alignment objective acquires the concept for a $0.1\%$ relative loss of the model's general text-to-image ability, against $41\%$ for the standard generative route. Our code is at https://github.com/Zane-ZYQiu/entry-point-umm.

cs.CV↗

arXiv2026

Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA

Macaron-V1 is an open agent-model family for experiential intelligence: learning from experience in real environments and continuing to learn after deployment. It is organized around two system goals. Adaptation is pursued through recursive improvement of versioned model-harness pairs, where experience from one configuration is evaluated under an external contract and used to construct its successor. Collaboration is pursued via the Mixture-of-LoRA (MoL) architecture that freezes a base model, composes specialist LoRA adapters, and selects one LoRA per user turn. The flagship Macaron-V1-Venti (748B) combines a 744B GLM-5.2 base with four LoRAs for chat, agent, coding, and GenUI; the Qwen3.6-35B-based Macaron-V1-Tall (50B) uses the same design for local deployment. This report presents Macaron-V1 as a co-designed system spanning architecture, algorithms, and infrastructure. The MoL architecture supports continual learning through extensible LoRA specialists. The algorithm combines Model-Harness Co-design and recursive self-improvement loop, including the UI4A component-native GenUI harness, a stateful action substrate, versioned Harness Context Protocol contract, and the agentic RL framework MindForge. The supporting infrastructure includes the post-training platform MinT, the long-context RL method LongStraw, and stability techniques for sparse MoE and DSA base models. We evaluate Macaron-V1 on Personal Intelligence, GenUI, and general capability benchmarks against frontier baselines. Our results validate the current system, while compounding gains from continual learning and collective intelligence remain open questions.

cs.LG↗

arXiv2026

Magnetic-Free Quantum Interference and Universal Josephson Diode Effect Driven by a Supercurrent Gauge Field

The Josephson effect, a hallmark of superconducting phase coherence, drives modern quantum technologies. However, Josephson-based quantum interference has hitherto been tethered to magnetic fields, despite phase coherence being a quintessential, intrinsic trait of superconductivity. Moreover, the Josephson diode effect (JDE) is typically viewed as an anomalous phenomenon indicative of broken symmetries in exotic phases of matter. Here, in planar Josephson junctions made with $\mathrm{Bi}_2\mathrm{O}_2\mathrm{Se}$ and bilayer graphene, we demonstrate that the JDE is a missing universal property of the Josephson effect. Simultaneously, we present an all-electric technology that replaces magnetic flux for controlling and measuring supercurrent interference. Central to our approach is a supercurrent gauge field (SGF), generated and amplified through high-kinetic-inductance superconductors and novel device architectures. By establishing the physical equivalence between the SGF and a magnetic field, we eliminate the reliance on external fields in quantum interference and reveal a universal, field-free JDE mechanism with broad implications for detecting broken-symmetry states. Finally, we show that the SGF offers capabilities beyond those of a conventional magnetic field by experimentally demonstrating a magnetic-free, phase-sensitive technique to construct and characterize finite-momentum superconductivity, opening new frontiers for exploring novel phases of matter and superconducting quantum architectures.

cond-mat.supr-con↗

arXiv2026

On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters

Parameter-efficient fine-tuning (PEFT) is usually treated as a cheaper alternative to full fine-tuning. We study a broader role: small trainable adapters as persistent local state on top of strong shared foundation models. In this framing, the base model provides shared competence while adapters carry instance-specific behavior such as preferences, skills, tool habits, and memory-like updates. We organize the problem around three scaling axes: Scale Up, where stronger shared priors make small local updates more useful; Scale Down, where we study how small adapters can be while remaining reliable; and Scale Out, where many persistent adapted instances coexist. MinT provides one infrastructure example for managing adapter identity, revision, provenance, evaluation, and serving residency. Together, the results suggest that PEFT can be a compact substrate for persistent personal models rather than only a budget substitute for full fine-tuning.

cs.LG↗

arXiv2026

MinT: Managed Infrastructure for Training and Serving Millions of LLMs

We present MindLab Toolkit (MinT), a managed infrastructure system for Low-Rank Adaptation (LoRA) post-training and online serving. MinT targets a setting where many trained policies are produced over a small number of expensive base-model deployments. Instead of materializing each policy as a merged full checkpoint, MinT keeps the base model resident and moves exported LoRA adapter revisions through rollout, update, export, evaluation, serving, and rollback, hiding distributed training, serving, scheduling, and data movement behind a service interface. MinT scales this path along three axes. Scale Up extends LoRA RL to frontier-scale dense and MoE architectures, including MLA and DSA attention paths, with training and serving validated beyond 1T total parameters. Scale Down moves only the exported LoRA adapter, which can be under 1% of base-model size in rank-1 settings; adapter-only handoff reduces the measured step by 18.3x on a 4B dense model and 2.85x on a 30B MoE, while concurrent multi-policy GRPO shortens wall time by 1.77x and 1.45x without raising peak memory. Scale Out separates durable policy addressability from CPU/GPU working sets: a tensor-parallel deployment supports 10^6-scale addressable catalogs (measured single-engine sweeps through 100K) and thousand-adapter active waves at cluster scale, with cold loading treated as scheduled service work and packed MoE LoRA tensors improving live engine loading by 8.5-8.7x. MinT thus manages million-scale LoRA policy catalogs while training and serving selected adapter revisions over shared 1T-class base models.

cs.LG↗

arXiv2026

$\delta$-mem: Efficient Online Memory for Large Language Models

Large language models increasingly need to accumulate and reuse historical information in long-term assistants and agent systems. Simply expanding the context window is costly and often fails to ensure effective context utilization. We propose $\delta$-mem, a lightweight memory mechanism that augments a frozen full-attention backbone with a compact online state of associative memory. $\delta$-mem compresses past information into a fixed-size state matrix updated by delta-rule learning, and uses its readout to generate low-rank corrections to the backbone's attention computation during generation. With only an $8\times8$ online memory state, $\delta$-mem improves the average score to $1.10\times$ that of the frozen backbone and $1.15\times$ that of the strongest non-$\delta$-mem memory baseline. It achieves larger gains on memory-heavy benchmarks, reaching $1.31\times$ on MemoryAgentBench and $1.20\times$ on LoCoMo, while largely preserving general capabilities. These results show that effective memory can be realized through a compact online state directly coupled with attention computation, without full fine-tuning, backbone replacement, or explicit context extension.

cs.AI↗

arXiv2026

Observation of field-odd and field-free superconducting diode effects in $\mathrm{Mo}_2\mathrm{C}$ nanoflakes

The superconducting diode effect (SDE) enables nonreciprocal supercurrent flow, holding immense potential for ultra-low-power quantum electronics. Intrinsic SDE typically requires materials with inherent symmetry breakings. Here, we report the discovery of SDE in chemical vapor deposition-grown molybdenum carbide ($\mathrm{Mo}_2\mathrm{C}$) nanoflakes, a material traditionally considered centrosymmetric. Strikingly, this system uniquely hosts both field-odd and field-free SDEs. Transport measurements reveal a field-odd SDE with tunable efficiency exceeding 40% at 4 K under a perpendicular in-plane magnetic field. In a separate sample, a robust field-free SDE persists under zero-field and field-coolings. Out-of-plane field sweeps confirm the intrinsic nature of these phenomena. We propose that domain-boundary supercurrents or charge density wave-like orders drive this unexpected combination of symmetry breakings. Our findings establish air-stable $\mathrm{Mo}_2\mathrm{C}$ as an ideal platform for nonreciprocal superconducting electronics operating at liquid-helium temperatures, expanding the search for SDE into nominally centrosymmetric superconductors.

cond-mat.supr-con↗

arXiv2026

OpenGame: Open Agentic Coding for Games

Game development sits at the intersection of creative design and intricate software engineering, demanding the joint orchestration of game engines, real-time loops, and tightly coupled state across many files. While Large Language Models (LLMs) and code agents now solve isolated programming tasks with ease, they consistently stumble when asked to produce a fully playable game from a high-level design, collapsing under cross-file inconsistencies, broken scene wiring, and logical incoherence. We bridge this gap with OpenGame, the first open-source agentic framework explicitly designed for end-to-end web game creation. At its core lies Game Skill, a reusable, evolving capability composed of a Template Skill that grows a library of project skeletons from experience and a Debug Skill that maintains a living protocol of verified fixes - together enabling the agent to scaffold stable architectures and systematically repair integration errors rather than patch isolated syntax bugs. Powering this framework is GameCoder-27B, a code LLM specialized for game engine mastery through a three-stage pipeline of continual pre-training, supervised fine-tuning, and execution-grounded reinforcement learning. Since verifying interactive playability is fundamentally harder than checking static code, we further introduce OpenGame-Bench, an evaluation pipeline that scores agentic game generation along Build Health, Visual Usability, and Intent Alignment via headless browser execution and VLM judging. Across 150 diverse game prompts, OpenGame establishes a new state-of-the-art. We hope OpenGame pushes code agents beyond discrete software engineering problems and toward building complex, interactive real-world applications. Our framework will be fully open-sourced.

cs.SE↗

arXiv2026

Gen-Searcher: Reinforcing Agentic Search for Image Generation

Recent image generation models have shown strong capabilities in generating high-fidelity and photorealistic images. However, they are fundamentally constrained by frozen internal knowledge, thus often failing on real-world scenarios that are knowledge-intensive or require up-to-date information. In this paper, we present Gen-Searcher, as the first attempt to train a search-augmented image generation agent, which performs multi-hop reasoning and search to collect the textual knowledge and reference images needed for grounded generation. To achieve this, we construct a tailored data pipeline and curate two high-quality datasets, Gen-Searcher-SFT-10k and Gen-Searcher-RL-6k, containing diverse search-intensive prompts and corresponding ground-truth synthesis images. We further introduce KnowGen, a comprehensive benchmark that explicitly requires search-grounded external knowledge for image generation and evaluates models from multiple dimensions. Based on these resources, we train Gen-Searcher with SFT followed by agentic reinforcement learning with dual reward feedback, which combines text-based and image-based rewards to provide more stable and informative learning signals for GRPO training. Experiments show that Gen-Searcher brings substantial gains, improving Qwen-Image by around 16 points on KnowGen and 15 points on WISE. We hope this work can serve as an open foundation for search agents in image generation, and we fully open-source our data, models, and code.

cs.CV↗

arXiv2026

Available Energy and Ground States of Convective Hydrodynamic and Hydromagnetic Instabilities

We propose a method for predicting the nonlinear saturation level of convective instabilities in neutral and magnetized fluids. The method combines Gardner's restacking algorithm, which computes the available energy and ground states of collisionless plasmas in phase space, and Lagrangian relaxation, where fluid elements find lower-energy equilibria while preserving local invariants. For the incompressible Rayleigh-Taylor instability, the problem is formally equivalent to Gardner's and the restacking algorithm directly applies in configuration space. To treat compressibility, we follow restacking with Lagrangian relaxation to obtain the ground state, and the results show excellent agreement with direct numerical simulations. Successful extension to the $m=0$ interchange instability in a Z-pinch demonstrates the method's potential as a general framework for estimating the nonlinear extent of convective instabilities, which can facilitate the design and operation of fusion reactors.

physics.plasm-ph↗

arXiv2026

Exploring Reasoning Reward Model for Agents

Agentic Reinforcement Learning (Agentic RL) has achieved notable success in enabling agents to perform complex reasoning and tool use. However, most methods still relies on sparse outcome-based reward for training. Such feedback fails to differentiate intermediate reasoning quality, leading to suboptimal training results. In this paper, we introduce Agent Reasoning Reward Model (Agent-RRM), a multi-faceted reward model that produces structured feedback for agentic trajectories, including (1) an explicit reasoning trace , (2) a focused critique that provides refinement guidance by highlighting reasoning flaws, and (3) an overall score that evaluates process performance. Leveraging these signals, we systematically investigate three integration strategies: Reagent-C (text-augmented refinement), Reagent-R (reward-augmented guidance), and Reagent-U (unified feedback integration). Extensive evaluations across 12 diverse benchmarks demonstrate that Reagent-U yields substantial performance leaps, achieving 43.7% on GAIA and 46.2% on WebWalkerQA, validating the effectiveness of our reasoning reward model and training schemes. Code, models, and datasets are all released to facilitate future research.

cs.AI↗

arXiv2025

OneThinker: All-in-one Reasoning Model for Image and Video

Reinforcement learning (RL) has recently achieved remarkable success in eliciting visual reasoning within Multimodal Large Language Models (MLLMs). However, existing approaches typically train separate models for different tasks and treat image and video reasoning as disjoint domains. This results in limited scalability toward a multimodal reasoning generalist, which restricts practical versatility and hinders potential knowledge sharing across tasks and modalities. To this end, we propose OneThinker, an all-in-one reasoning model that unifies image and video understanding across diverse fundamental visual tasks, including question answering, captioning, spatial and temporal grounding, tracking, and segmentation. To achieve this, we construct the OneThinker-600k training corpus covering all these tasks and employ commercial models for CoT annotation, resulting in OneThinker-SFT-340k for SFT cold start. Furthermore, we propose EMA-GRPO to handle reward heterogeneity in multi-task RL by tracking task-wise moving averages of reward standard deviations for balanced optimization. Extensive experiments on diverse visual benchmarks show that OneThinker delivers strong performance on 31 benchmarks, across 10 fundamental visual understanding tasks. Moreover, it exhibits effective knowledge transfer between certain tasks and preliminary zero-shot generalization ability, marking a step toward a unified multimodal reasoning generalist. All code, model, and data are released.

cs.CV↗

arXiv2025

Asymmetric quantum Hall effect and diminished $\nu=0$ longitudinal resistance in graphene/InSe heterostructures

We investigate quantum transport in graphene/InSe heterostructures and find major asymmetries in the longitudinal resistance ($R_{xx}$) and vanishing $R_{xx}$ peaks at high magnetic fields, particularly at the charge-neutrality point. Our Landauer-Buttiker analysis and numerical simulations show that a monotonically varying density gradient combined with a full equilibration mechanism can explain these phenomena. Our results also suggest the presence of trivial long-range chiral edge current and offer a broadly applicable way to engineer transport properties in quantum Hall systems.

cond-mat.mes-hall↗

arXiv2025

SophiaVL-R1: Reinforcing MLLMs Reasoning with Thinking Reward

Recent advances have shown success in eliciting strong reasoning abilities in multimodal large language models (MLLMs) through rule-based reinforcement learning (RL) with outcome rewards. However, this paradigm typically lacks supervision over the thinking process leading to the final outcome. As a result, the model may learn sub-optimal reasoning strategies, which can hinder its generalization ability. In light of this, we propose SophiaVL-R1, as an attempt to add reward signals for the thinking process in this paradigm. To achieve this, we first train a thinking reward model that evaluates the quality of the entire thinking process. Given that the thinking reward may be unreliable for certain samples due to reward hacking, we propose the Trust-GRPO method, which assigns a trustworthiness weight to the thinking reward during training. This weight is computed based on the thinking reward comparison of responses leading to correct answers versus incorrect answers, helping to mitigate the impact of potentially unreliable thinking rewards. Moreover, we design an annealing training strategy that gradually reduces the thinking reward over time, allowing the model to rely more on the accurate rule-based outcome reward in later training stages. Experiments show that our SophiaVL-R1 surpasses a series of reasoning MLLMs on various benchmarks (e.g., MathVisita, MMMU), demonstrating strong reasoning and generalization capabilities. Notably, our SophiaVL-R1-7B even outperforms LLaVA-OneVision-72B on most benchmarks, despite the latter having 10 times more parameters. All code, models, and datasets are made publicly available at https://github.com/kxfan2002/SophiaVL-R1.

cs.CV↗

arXiv2024

Superconductivity at Pd/Bi$_2$Se$_3$ Interfaces Due to Self-Formed PdBiSe Interlayers

Understanding the physical and chemical processes at the interface of metals and topological insulators is crucial for developing the next generation of topological quantum devices. Here we report the discovery of robust superconductivity in Pd/Bi$_2$Se$_3$ bilayers fabricated by sputtering Pd on the surface of Bi$_2$Se$_3$. Through transmission electron microscopy measurements, we identify that the observed interfacial superconductivity originates from the diffusion of Pd into Bi$_2$Se$_3$. In the diffusion region, Pd chemically reacts with Bi$_2$Se$_3$ and forms a layer of PdBiSe, a known su-perconductor with a bulk transition temperature of 1.5 K. Our work provides a method for in-troducing superconductivity into Bi$_2$Se$_3$, laying the foundation for developing sophisticated Bi$_2$Se$_3$-based topological devices.

cond-mat.supr-con↗

arXiv2024

HypDAE: Hyperbolic Diffusion Autoencoders for Hierarchical Few-shot Image Generation

Few-shot image generation aims to generate diverse and high-quality images for an unseen class given only a few examples in that class. A key challenge in this task is balancing category consistency and image diversity, which often compete with each other. Moreover, existing methods offer limited control over the attributes of newly generated images. In this work, we propose Hyperbolic Diffusion Autoencoders (HypDAE), a novel approach that operates in hyperbolic space to capture hierarchical relationships among images from seen categories. By leveraging pre-trained foundation models, HypDAE generates diverse new images for unseen categories with exceptional quality by varying stochastic subcodes or semantic codes. Most importantly, the hyperbolic representation introduces an additional degree of control over semantic diversity through the adjustment of radii within the hyperbolic disk. Extensive experiments and visualizations demonstrate that HypDAE significantly outperforms prior methods by achieving a better balance between preserving category-relevant features and promoting image diversity with limited data. Furthermore, HypDAE offers a highly controllable and interpretable generation process.

cs.CV↗

arXiv2024

Theoretical and Global Simulation Analysis of Collisional Microtearing

This study delves into Microtearing Modes (MTMs) in tokamak plasmas, employing advanced simulations within the BOUT++ framework. The research, centering on collisional MTMs influenced by the time-dependent thermal force, enhances our understanding of plasma dynamics. It achieves this through the simplification and linearization of control equations in detailed linear simulations. The study meticulously evaluates various conductivity models, including those proposed by Larakers, Drake, and Hassam, under diverse plasma conditions and collision regimes. A notable achievement of this research is the derivation of a unified dispersion relation that encompasses both MTM and Drift-Alfven Wave (DAW) instabilities. It interestingly reveals that DAW and MTM exhibit instability at different proximities to the rational surface. Specifically, MTMs become unstable near the rational surface but stabilize farther away, whereas the drift-Alfven instability manifests away from the rational surface. Further, the study re-derives MTM dispersion relations based on Ohm's law and the vorticity equation, providing a thorough analysis of electromagnetic and electrostatic interactions in tokamaks. Global simulations demonstrate an inverse correlation between MTM growth rates and collisionality, and a direct correlation with temperature gradients. The nonalignment of the rational surface with the peak of electron local diamagnetic frequency stabilizes the MTMs. Nonlinear simulations highlight electron temperature relaxation as the primary saturation mechanism for MTMs, with magnetic flutter identified as the dominant mode of electron thermal transport.

physics.plasm-ph↗