arXiv subjects

Xueqi Li

Publications and source records attributed to Xueqi Li.

At least 19 recordsLinked to original sources

arXiv2026

Same Agent, Different Answers: A Repeat-Aware Audit of Corpus-Induced Answer Churn in Retrieval-Augmented QA

A retrieval-augmented QA system can return different answers after an index expansion even when its requested model identifier, prompt, retrieval policy, evidence depth, rendering, and exposed generation controls are held fixed. Aggregate accuracy may hide these changes when gains and losses cancel, while ordinary generation variability makes one-shot comparisons overstate update effects. We call the hidden phenomenon accuracy-blind answer churn and introduce the \emph{Snapshot Compatibility Audit}, which estimates excess answer churn by subtracting same-snapshot repeat disagreement from cross-snapshot disagreement. We instantiate it by expanding one frozen FineWeb prefix from one to seven shards. In a preregistered 400-question Natural Questions study, normalized-exact and blinded-semantic excess churn are 6.44 and 10.25 percentage points while exact-match accuracy changes by only $-1.50$ points. A post-hoc analysis finds repeat-stable semantic flips on 40/400 questions. A separately preregistered 200-question TriviaQA study yields smaller, directionally consistent excess churn while exact-match accuracy moves in the opposite direction. An outcome-blind post-hoc 100-question subset replication with a second DeepSeek generator and serving configuration finds 8.75 pp of semantic excess churn even as exact match rises by 3.00 percentage points. Answer-level compatibility can therefore fail without a conspicuous or consistently directed utility shift. Retrieval-augmented releases should audit compatibility alongside utility.

cs.IR↗

arXiv2026

Never Too Late for Force: Accelerating VLA Post-Training with Reactive Force Injection

Pretrained vision-language-action (VLA) policies provide strong language-conditioned manipulation knowledge, but they remain largely vision-driven and can struggle once manipulation enters contact states where the scene is occluded, depth is ambiguous, or small force errors push execution off the offline demonstration distribution. We present LIFT (Late Reactive Injection of Force for VLA Post-Training), a force-aware post-training framework that adds contact reactivity to a pretrained VLA policy while preserving its general manipulation knowledge. LIFT grafts a reactive action expert beside the original action expert, initializes it from pretrained action weights, and injects recent 6D end-effector force through causal force memory and zero-initialized cross attention, enabling actions to be refreshed during execution. To address the policy-dependent distribution shift of contact feedback, LIFT further couples reactive force injection with an online DAgger loop that trains on a mixture of offline task-alignment data and human-corrected online rollouts. Across towel folding, book insertion, and Hanoi ring placement, LIFT learns faster and reaches higher performance than vision-only post-training, while ablations show that reactive force memory and online corrective data are both important for robust contact-rich manipulation. Our code and data will be publicly available.

cs.RO↗

arXiv2026

FlashAccel: Leveraging High-Bandwidth Flash (HBF) for High-Throughput LLM Inference

Large language model (LLM) inference is increasingly limited by the capacity of High-Bandwidth Memory (HBM) in GPUs, as model weights and KV cache grow rapidly. High-Bandwidth Flash (HBF) provides higher capacity than HBM while offering comparable bandwidth, making it a promising substrate for capacity-constrained LLM inference. However, its inherently high access latency, low bandwidth utilization, and lack of support for heterogeneous resource management make it difficult to integrate HBF into GPUs for LLM inference. We present FlashAccel, a co-designed system that enables efficient LLM inference using HBF. FlashAccel integrates HBF into HBM-based GPUs, providing architectural support to mitigate access latency. It improves bandwidth utilization through specialized data layouts for both model weights and KV cache, and introduces an HBF-aware storage management layer together with a programming model to organize persistent data in HBF and coordinate heterogeneous memory resources at the system level. Experimental results demonstrate that integrating six HBF stacks into the GPU enables FlashAccel to deliver an average improvement of 2.49$\times$ and 1.93$\times$ in throughput per GPU and energy efficiency over the HBM-only GPU under a 100ms latency constraint, respectively.

cs.AR↗

arXiv2026

Fermion mass relations in one-parameter modular models

Modular flavour symmetries provide a possible organizing principle for the Standard Model Yukawa sector, by replacing generic couplings with a potentially small number of modular forms controlled by a single complex modulus. We study the extreme limit of this idea: \acp{OPM}, in which each charged-fermion mass matrix is fixed by a single modular invariant contraction. We develop a systematic method to construct such models, showing that the \ac{OPM} requirement is already highly constraining at the level of possible fermion hierarchies. In a concrete realization, the charged-lepton and down-quark sectors are controlled by the common modulus, leading to exact mass relations at the flavour scale, \[ m_s^5 = 2\sqrt{2}\,m_d^3m_b^2, \qquad m_\mu^3 = \sqrt{2}\,m_e m_\tau^2, \qquad m_s^2m_\tau = \sqrt{2}\,m_e m_b^2. \] We show that, once renormalization-group evolution and selective supersymmetric threshold effects are included, these high-scale relations can be made compatible with low-energy charged-fermion data. Our results provide a working proof of principle for \acp{OPM} and point towards a possible route to the flavour puzzle through highly

hep-ph↗

arXiv2026

Scaling by Diversified Experience for Vision-Language-Action Models

Vision-Language-Action models face significant challenges in real-world deployment due to the entanglement of high-level reasoning with low-level control, and the instability of policy optimization. In this paper, we introduce SyVLA, a robust VLA model trained with diversified experiences. We propose an Intention Decoupling algorithm to isolate control-relevant features from reasoning contexts and a similar-sample guided RL pipeline to stabilize policy updates and mitigate distribution shift. Extensive experiments on real-world robotic tasks and multi-modal benchmarks demonstrate that SyVLA achieves superior task success rates and stronger out-of-distribution generalization compared to existing methods, while effectively preserving core vision-language capabilities. Codes and Datasets is released on \href{https://sy-vla.github.io/}{project page}.

cs.CV↗

arXiv2026

World-Language-Action Model for Unified World Modeling, Language Reasoning, and Action Synthesis

We propose world-language-action (WLA) models as a new class of embodied foundation models. WLA takes textual instructions, images, and robot states as inputs to jointly predict textual subtasks, subgoal images, and robot actions, conjoining the \emph{world modeling interface} to learn from extensive egocentric videos as in the world-action model (WAM) and the \emph{language reasoning} capacities to solve complex long-horizon tasks as in vision-language-action (VLA) models. At the core of WLA lies an \emph{autoregressive (AR)} Transformer backbone, instead of a bidirectional diffusion Transformer as in WAMs, to predict the \emph{next state}, comprising the \emph{semantic-level} textual intention and complementary \emph{fine-grained} physical dynamics. The physical dynamics are supervised by the world modeling objective based on a dedicated World Expert, and are leveraged to ease the characterization of the state-action correlation for the Action Expert. WLA leverages meta-queries to make the world prediction \emph{implicitly} impact the action generation so that the former can be disabled during inference. The world prediction can also be activated to enable test-time scaling for improved robot control. Our WLA-0 prototype, with 2B active parameters, achieves 40 ms per inference on an NVIDIA RTX 5090. Evaluations across simulated and real-world environments demonstrate that WLA-0 achieves state-of-the-art multi-task and long-horizon learning abilities, e.g., 92.94\% success rate on RoboTwin2.0 Clean and 56.5\% success rate on RMBench. WLA-0 also holds the promise to learn novel tasks directly from \emph{cross-embodiment robot videos} without action annotations.

cs.RO↗

arXiv2026

TokenStack: A Heterogeneous HBM-PIM Architecture and Runtime for Efficient LLM Inference

Large language model (LLM) serving is now limited by the key-value (KV) cache. During decode, each new token rereads prior KV state, so attention becomes a bandwidth- and capacity-heavy memory task. HBM-PIM helps by moving attention closer to memory, but current stack organizations still waste resources. In practice, only hot KV blocks benefit from near-memory compute. Weights, activations, and cold KV mainly need dense storage and GPU-visible bandwidth. A uniform HBM-PIM stack makes all layers pay for PIM logic, while a dedicated-PIM design such as AttAcc recovers capacity but shrinks the HBM bandwidth left for GPU-side work. We propose TokenStack, a vertically heterogeneous HBM-PIM architecture for KV-centric LLM serving that leverages HBM4's logic-die substrate. TokenStack separates each stack into dense capacity layers and PIM-enabled compute layers, then uses the logic base die as a stack-local control point that manages cross-layer movement without host-side overhead. The base-die controller handles cross-layer DMA, layered address translation, attention-side gather/broadcast coordination, and inline quantization during migration. On top of this hardware, TokenStack uses topology-aware KV placement, workload-aware eviction, and bounded replication to keep hot KV near PIM compute while moving colder state to dense layers. Using production-derived traces across four models, completed multi-QPS runs show that TokenStack increases geometric-mean token throughput by 1.62x and SLO-compliant serving capacity by 1.70x over AttAcc, and reduces per-token energy by 30-47%.

cs.AR↗

arXiv2026

Revision or Re-Solving? Decomposing Second-Pass Gains in Multi-LLM Pipelines

Multi-LLM revision pipelines, in which a second model reviews and improves a draft produced by a first, are widely assumed to derive their gains from genuine error correction. We question this assumption with a controlled decomposition experiment that uses four matched conditions to separate second-pass gains into three additive components: re-solving, scaffold, and content. We evaluate this design across two model pairs on three benchmarks spanning knowledge-intensive MCQ and competitive programming. Our results show that the gains of multi-LLM revision are not monolithic, but depend on task structure, draft quality, and the type of draft information. On MCQ tasks, where the answer space is constrained and drafts provide little structural guidance, most gains are consistent with stronger-model re-solving, and directly routing queries to the stronger model can be more effective than revising a weak draft. On code generation tasks, however, two-stage prompting remains useful because even semantically null drafts can provide substantial structural scaffolding, while weak draft content can be harmful. Finally, role-reversed experiments show that strong drafts clearly benefit weak reviewers. Ultimately, our findings demonstrate that the utility of multi-LLM revision is dynamically bottlenecked by task structure and draft quality, necessitating more targeted pipeline designs rather than blanket revision strategies.

cs.SE↗

arXiv2026

INFLAVON: CMB as cosmic tracer of Flavor physics

We unify one of the most widely studied frameworks to explain the hierarchical structure of the flavor sector in the Standard Model, the Froggatt-Nielsen mechanism, with cosmic inflation. We propose that the complex scalar field, the so-called flavon, which breaks the Froggatt-Nielsen $U(1)$ symmetry and generates the Yukawa couplings of the Standard Model, to also drive inflation, which we dub as Inflavon. After inflation ends, the decay of the inflavon reheats the Universe, establishing a novel link between early Universe cosmology and flavor physics. As concrete examples, we present realizations where the inflavon potential is described by an $\alpha$-attractor potential. We then compute the resulting CMB observables, specifically the spectral index ($n_s$), the tensor-to-scalar ratio ($r$), and the amplitude of scalar perturbations ($A_s$) as functions of the underlying Froggatt-Nielsen model parameters. We identify the parameter space in Froggatt-Nielsen models involving the scale of Flavor symmetry breaking $\Lambda_{\rm FN}$ and FN charges which is ruled out by Planck and ACT data, as well as the region that could be probed by next-generation CMB experiments like CMB-S4, SO and LiteBIRD. We also discuss inflavon as dark matter and its isocurvature constraints.

hep-ph↗

arXiv2025

Mantis: A Versatile Vision-Language-Action Model with Disentangled Visual Foresight

Recent advances in Vision-Language-Action (VLA) models demonstrate that visual signals can effectively complement sparse action supervisions. However, letting VLA directly predict high-dimensional visual states can distribute model capacity and incur prohibitive training cost, while compressing visual states into more compact supervisory signals inevitably incurs information bottlenecks. Moreover, existing methods often suffer from poor comprehension and reasoning capabilities due to the neglect of language supervision. This paper introduces Mantis, a novel framework featuring a Disentangled Visual Foresight (DVF) to tackle these issues. Specifically, Mantis decouples visual foresight prediction from the backbone with the combination of meta queries and a diffusion Transformer (DiT) head. With the current visual state provided to the DiT via a residual connection, a simple next-state prediction objective enables the meta queries to automatically capture the latent actions that delineate the visual trajectory, and hence boost the learning of explicit actions. The disentanglement reduces the burden of the VLA backbone, enabling it to maintain comprehension and reasoning capabilities through language supervision. Empirically, pretrained on human manipulation videos, robot demonstrations, and image-text pairs, Mantis achieves a 96.7% success rate on LIBERO benchmark after fine-tuning, surpassing powerful baselines while exhibiting high convergence speed. Real-world evaluations show that Mantis outperforms $\pi_{0.5}$, a leading open-source VLA model, particularly in instruction-following capability, generalization to unseen instructions, and reasoning ability. Code and weights are released to support the open-source community.

cs.CV↗

arXiv2025

Multivariate Diffusion Transformer with Decoupled Attention for High-Fidelity Mask-Text Collaborative Facial Generation

While significant progress has been achieved in multimodal facial generation using semantic masks and textual descriptions, conventional feature fusion approaches often fail to enable effective cross-modal interactions, thereby leading to suboptimal generation outcomes. To address this challenge, we introduce MDiTFace--a customized diffusion transformer framework that employs a unified tokenization strategy to process semantic mask and text inputs, eliminating discrepancies between heterogeneous modality representations. The framework facilitates comprehensive multimodal feature interaction through stacked, newly designed multivariate transformer blocks that process all conditions synchronously. Additionally, we design a novel decoupled attention mechanism by dissociating implicit dependencies between mask tokens and temporal embeddings. This mechanism segregates internal computations into dynamic and static pathways, enabling caching and reuse of features computed in static pathways after initial calculation, thereby reducing additional computational overhead introduced by mask condition by over 94% while maintaining performance. Extensive experiments demonstrate that MDiTFace significantly outperforms other competing methods in terms of both facial fidelity and conditional consistency.

cs.CV↗

arXiv2025

TurboFuzz: FPGA Accelerated Hardware Fuzzing for Processor Agile Verification

Verification is a critical process for ensuring the correctness of modern processors. The increasing complexity of processor designs and the emergence of new instruction set architectures (ISAs) like RISC-V have created demands for more agile and efficient verification methodologies, particularly regarding verification efficiency and faster coverage convergence. While simulation-based approaches now attempt to incorporate advanced software testing techniques such as fuzzing to improve coverage, they face significant limitations when applied to processor verification, notably poor performance and inadequate test case quality. Hardware-accelerated solutions using FPGA or ASIC platforms have tried to address these issues, yet they struggle with challenges including host-FPGA communication overhead, inefficient test pattern generation, and suboptimal implementation of the entire multi-step verification process. In this paper, we present TurboFuzz, an end-to-end hardware-accelerated verification framework that implements the entire Test Generation-Simulation-Coverage Feedback loop on a single FPGA for modern processor verification. TurboFuzz enhances test quality through optimized test case (seed) control flow, efficient inter-seed scheduling, and hybrid fuzzer integration, thereby improving coverage and execution efficiency. Additionally, it employs a feedback-driven generation mechanism to accelerate coverage convergence. Experimental results show that TurboFuzz achieves up to 2.23x more coverage collection than software-based fuzzers within the same time budget, and up to 571x performance speedup when detecting real-world issues, while maintaining full visibility and debugging capabilities with moderate area overhead.

cs.AR↗

arXiv2025

Step-Audio 2 Technical Report

This paper presents Step-Audio 2, an end-to-end multi-modal large language model designed for industry-strength audio understanding and speech conversation. By integrating a latent audio encoder and reasoning-centric reinforcement learning (RL), Step-Audio 2 achieves promising performance in automatic speech recognition (ASR) and audio understanding. To facilitate genuine end-to-end speech conversation, Step-Audio 2 incorporates the generation of discrete audio tokens into language modeling, significantly enhancing its responsiveness to paralinguistic information such as speaking styles and emotions. To effectively leverage the rich textual and acoustic knowledge in real-world data, Step-Audio 2 integrates retrieval-augmented generation (RAG) and is able to call external tools such as web search to mitigate hallucination and audio search to switch timbres. Trained on millions of hours of speech and audio data, Step-Audio 2 delivers intelligence and expressiveness across diverse conversational scenarios. Evaluation results demonstrate that Step-Audio 2 achieves state-of-the-art performance on various audio understanding and conversational benchmarks compared to other open-source and commercial solutions. Please visit https://github.com/stepfun-ai/Step-Audio2 for more information.

cs.CL↗

arXiv2025

Modular Flavor Symmetries and Fermion Mass Hierarchies

We investigate fermion mass hierarchies in models with modular flavor symmetries. Several key conclusions arise from the observation that the determinants of mass matrices transform as 1-dimensional vector-valued modular forms. We demonstrate that, under some fairly general assumptions, achieving hierarchical fermion masses requires the vacuum expectation value of the modulus $\tau$ to be located near one of the critical points, $i$, $i\infty$, or $\omega$. We also revisit the universal near-critical behavior around these points and classify the resulting mass hierarchies for the critical points $i$ and $\omega$. We compare the traditional Froggatt--Nielsen mechanism with its modular variant. The knowledge and boundedness of Fourier and Taylor coefficients are crucial to the predictive power of modular flavor symmetries.

hep-ph↗

arXiv2025

Flavor Symmetries and Winding Modes

Modular flavor symmetries have been proposed as a new way to address the flavor problem. It is known that they can emerge from string compactifications. We discuss this connection in detail, and show how the congruence subgroups of SL(2,Z), which underlie many modular flavor symmetries, emerge from stringy duality symmetries by orbifolding. This requires an analysis of massive states, which reveals a picture that is more intricate than the well-known situation on the torus. It involves towers of states of different quantum numbers, related by modular transformations. Members of different towers become massless at different points in moduli space. We also show that, at least in the Z_3 orbifold, the string selection rules can be understood as discrete remnants of continuous gauge symmetries. Non-Abelian discrete flavor symmetries arise as relics of various, relatively misaligned, continuous Abelian gauge symmetries. The generators of these U(1) symmetries give rise to CP-violating Clebsch-Gordan coefficients. If the modulus settles close to a critical point, the corresponding gauge bosons may be light enough to be searched for at future colliders.

hep-th↗

arXiv2025

Multiple realizations of modular flavor symmetries and their phenomenology

We point out that specifying the finite modular group does not uniquely fix a modular flavor symmetry. We illustrate this using the finite modular group $T'$. Otherwise equivalent models based on different $T'$ lead to modular forms with different properties and, hence, produce different phenomenological features. We exemplify this in various scenarios, and show that the ability of a given model to accommodate mass and other observed hierarchies depends sensitively on the way the $T'$ is implemented.

hep-ph↗

arXiv2022

AlphaSparse: Generating High Performance SpMV Codes Directly from Sparse Matrices

Sparse Matrix-Vector multiplication (SpMV) is an essential computational kernel in many application scenarios. Tens of sparse matrix formats and implementations have been proposed to compress the memory storage and speed up SpMV performance. We develop AlphaSparse, a superset of all existing works that goes beyond the scope of human-designed format(s) and implementation(s). AlphaSparse automatically \emph{creates novel machine-designed formats and SpMV kernel implementations} entirely from the knowledge of input sparsity patterns and hardware architectures. Based on our proposed Operator Graph that expresses the path of SpMV format and kernel design, AlphaSparse consists of three main components: Designer, Format \& Kernel Generator, and Search Engine. It takes an arbitrary sparse matrix as input while outputs the performant machine-designed format and SpMV implementation. By extensively evaluating 843 matrices from SuiteSparse Matrix Collection, AlphaSparse achieves significant performance improvement by 3.2$\times$ on average compared to five state-of-the-art artificial formats and 1.5$\times$ on average (up to 2.7$\times$) over the up-to-date implementation of traditional auto-tuning philosophy.

cs.DC↗

arXiv2020

Semi-Disentangled Representation Learning in Recommendation System

Disentangled representation has been widely explored in many fields due to its maximal compactness, interpretability and versatility. Recommendation system also needs disentanglement to make representation more explainable and general for downstream tasks. However, some challenges slow its broader application -- the lack of fine-grained labels and the complexity of user-item interactions. To alleviate these problems, we propose a Semi-Disentangled Representation Learning method (SDRL) based on autoencoders. SDRL divides each user/item embedding into two parts: the explainable and the unexplainable, so as to improve proper disentanglement while preserving complex information in representation. The explainable part consists of $internal\ block$ for individual-based features and $external\ block$ for interaction-based features. The unexplainable part is composed by $other\ block$ for other remaining information. Experimental results on three real-world datasets demonstrate that the proposed SDRL could not only effectively express user and item features but also improve the explainability and generality compared with existing representation methods.

cs.SI↗