arXiv subjects

Yu Gu

Publications and source records attributed to Yu Gu.

At least 73 records · Page 4Linked to original sources

arXiv2026

Anticipatory Planning for Multimodal AI Agents

Recent advances in multimodal agents have improved computer-use interaction and tool-usage, yet most existing systems remain reactive, optimizing actions in isolation without reasoning about future states or long-term goals. This limits planning coherence and prevents agents from reliably solving high-level, multi-step tasks. We introduce TraceR1, a two-stage reinforcement learning framework that explicitly trains anticipatory reasoning by forecasting short-horizon trajectories before execution. The first stage performs trajectory-level reinforcement learning with rewards that enforce global consistency across predicted action sequences. The second stage applies grounded reinforcement fine-tuning, using execution feedback from frozen tool agents to refine step-level accuracy and executability. TraceR1 is evaluated across seven benchmarks, covering online computer-use, offline computer-use benchmarks, and multimodal tool-use reasoning tasks, where it achieves substantial improvements in planning stability, execution robustness, and generalization over reactive and single-stage baselines. These results show that anticipatory trajectory reasoning is a key principle for building multimodal agents that can reason, plan, and act effectively in complex real-world environments.

cs.AI↗

arXiv2026

Log Log Fluctuations of the Stochastic Heat Flow

We study the stochastic heat flow with constant initial data and analyze its spatial average on the scale of $\varepsilon\ll1$. We prove that the logarithm of the averaged process satisfies a pointwise central limit theorem: After being centered by $-\tfrac{1}{2}\log\log \varepsilon^{-1}$ and scaled down by $\sqrt{\log\log \varepsilon^{-1}}$, it converges in distribution to a standard Gaussian.

math.PR↗

arXiv2026

HIPPO: Enhancing the Table Understanding Capability of LLMs through Hybrid-Modal Preference Optimization

Tabular data contains rich structural semantics and plays a crucial role in organizing and manipulating information. Recent methods employ Multi-modal Large Language Models (MLLMs) to address table-related tasks across various modalities of table representations. However, existing studies mainly focus on exploring the table understanding ability of MLLMs using unimodal representations, which limits further exploration of multi-modal representations to enable more effective table reasoning. To better capture structural semantics from the tabular data, this paper introduces the HybrId-modal Preference oPtimizatiOn (HIPPO) model, which represents tables using both text and image, optimizing MLLMs by learning more comprehensive table information from these multiple modalities. Specifically, HIPPO samples MLLM responses from hybrid-modal table representations and designs a modality-consistent sampling strategy to enhance response diversity and mitigate modality bias during Direct Preference Optimization (DPO) training. Experiments on table question answering and table fact verification tasks demonstrate the effectiveness of HIPPO, achieving a 4% improvement over various table reasoning models. Further analysis reveals that HIPPO not only enhances the table reasoning capability based on unimodal representations but also facilitates the extraction of complementary semantics across modalities. The code is available at https://github.com/NEUIR/HIPPO.

cs.CL↗

arXiv2026

OJBKQ: Objective-Joint Babai-Klein Quantization

Post-training quantization (PTQ) is widely used to compress large language models without retraining. However, many existing weight-only methods rely on heuristic objectives and greedy rounding, thus leading to noticeable degradation under low-bit quantization. In this work, we introduce OJBKQ (Objective-Joint Babai-Klein Quantization with K-Best Sampling), a layer-wise PTQ method that formulates weight quantization as a joint optimization problem over activations and weights. This formulation results in a multiple-right-hand-side box-constrained integer least squares (BILS) problem in each layer, which is NP-hard. For each column of the weight matrix, we apply an extended Babai nearest-plane algorithm and an extended version of Klein's randomized Babai algorithm to find the minimum-residual Babai-Klein point, a sub-optimal solution to the BILS problem. Experimental results on large language models show that OJBKQ achieves lower perplexity at 3-4 bits compared to existing PTQ approaches, while maintaining comparable computational cost.

cs.LG↗

arXiv2026

SofT-GRPO: Surpassing Discrete-Token LLM Reinforcement Learning via Gumbel-Reparameterized Soft-Thinking Policy Optimization

The soft-thinking paradigm for Large Language Model (LLM) reasoning can outperform the conventional discrete-token Chain-of-Thought (CoT) reasoning in some scenarios, underscoring its research and application value. However, while the discrete-token CoT reasoning pattern can be reinforced through policy optimization algorithms such as group relative policy optimization (GRPO), extending the soft-thinking pattern with Reinforcement Learning (RL) remains challenging. This difficulty stems from the complexities of injecting stochasticity into soft-thinking tokens and updating soft-thinking policies accordingly. As a result, previous attempts to combine soft-thinking with GRPO typically underperform their discrete-token GRPO counterparts. To fully unlock the potential of soft-thinking, this paper presents a novel policy optimization algorithm, SofT-GRPO, to reinforce LLMs under the soft-thinking reasoning pattern. SofT-GRPO injects the Gumbel noise into logits, employs the Gumbel-Softmax technique to avoid soft-thinking tokens outside the pre-trained embedding space, and leverages the reparameterization trick in policy gradient. We conduct experiments across base LLMs ranging from 1.5B to 7B parameters, and results demonstrate that SofT-GRPO enables soft-thinking LLMs to slightly outperform discrete-token GRPO on Pass@1 (+0.13% on average accuracy), while exhibiting a substantial uplift on Pass@32 (+2.19% on average accuracy). Codes and weights are available on https://github.com/zz1358m/SofT-GRPO-master

cs.AI↗

arXiv2026

Varying coefficient model for longitudinal data with informative observation times

Varying coefficient models are widely used to characterize dynamic associations between longitudinal outcomes and covariates. Existing work on varying coefficient models, however, all assumes that observation times are independent of the longitudinal outcomes, which is often violated in real-world studies with outcome-driven or otherwise informative visit schedules. Such informative observation times can lead to biased estimation and invalid inference using existing methods. In this article, we develop estimation and inference procedures for varying coefficient models that account for informative observation times. We model the observation time process as a general counting process under a proportional intensity model, with time-varying covariates summarizing the observed history. To address potential bias, we incorporate inverse intensity weighting into a sieve estimation framework, yielding closed-form coefficient function estimators via weighted least squares. We establish consistency, convergence rates, and asymptotic normality of the proposed estimators, and construct pointwise confidence intervals for the coefficient functions. Extensive simulation studies demonstrate that the proposed weighted method substantially outperforms the conventional unweighted method when observation times are informative. Finally, we provide an application of our method to the Alzheimer's Disease Neuroimaging Initiative study.

stat.ME↗

arXiv2026

DAIEN-TTS: Disentangled Audio Infilling for Environment-Aware Text-to-Speech Synthesis

This paper presents DAIEN-TTS, a zero-shot text-to-speech (TTS) framework that enables ENvironment-aware synthesis through Disentangled Audio Infilling. By leveraging separate speaker and environment prompts, DAIEN-TTS allows independent control over the timbre and the background environment of the synthesized speech. Built upon F5-TTS, the proposed DAIEN-TTS first incorporates a pretrained speech-environment separation (SES) module to disentangle the environmental speech into mel-spectrograms of clean speech and environment audio. Two random span masks of varying lengths are then applied to both mel-spectrograms, which, together with the text embedding, serve as conditions for infilling the masked environmental mel-spectrogram, enabling the simultaneous continuation of personalized speech and time-varying environmental audio. To further enhance controllability during inference, we adopt dual classifier-free guidance (DCFG) for the speech and environment components and introduce a signal-to-noise ratio (SNR) adaptation strategy to align the synthesized speech with the environment prompt. Experimental results demonstrate that DAIEN-TTS generates environmental personalized speech with high naturalness, strong speaker similarity, and high environmental fidelity.

eess.AS↗

arXiv2026

Teaching LLMs to Learn Tool Trialing and Execution through Environment Interaction

Equipping Large Language Models (LLMs) with external tools enables them to solve complex real-world problems. However, the robustness of existing methods remains a critical challenge when confronting novel or evolving tools. Existing trajectory-centric paradigms primarily rely on memorizing static solution paths during training, which limits the ability of LLMs to generalize tool usage to newly introduced or previously unseen tools. In this paper, we propose ToolMaster, a framework that shifts tool use from imitating golden tool-calling trajectories to actively learning tool usage through interaction with the environment. To optimize LLMs for tool planning and invocation, ToolMaster adopts a trial-and-execution paradigm, which trains LLMs to first imitate teacher-generated trajectories containing explicit tool trials and self-correction, followed by reinforcement learning to coordinate the trial and execution phases jointly. This process enables agents to autonomously explore correct tool usage by actively interacting with environments and forming experiential knowledge that benefits tool execution. Experimental results demonstrate that ToolMaster significantly outperforms existing baselines in terms of generalization and robustness across unseen or unfamiliar tools. All code and data are available at https://github.com/NEUIR/ToolMaster.

cs.SE↗

arXiv2026

Long-Chain Reasoning Distillation via Adaptive Prefix Alignment

Large Language Models (LLMs) have demonstrated remarkable reasoning capabilities, particularly in solving complex mathematical problems. Recent studies show that distilling long reasoning trajectories can effectively enhance the reasoning performance of small-scale student models. However, teacher-generated reasoning trajectories are often excessively long and structurally complex, making them difficult for student models to learn. This mismatch leads to a gap between the provided supervision signal and the learning capacity of the student model. To address this challenge, we propose Prefix-ALIGNment distillation (P-ALIGN), a framework that fully exploits teacher CoTs for distillation through adaptive prefix alignment. Specifically, P-ALIGN adaptively truncates teacher-generated reasoning trajectories by determining whether the remaining suffix is concise and sufficient to guide the student model. Then, P-ALIGN leverages the teacher-generated prefix to supervise the student model, encouraging effective prefix alignment. Experiments on multiple mathematical reasoning benchmarks demonstrate that P-ALIGN outperforms all baselines by over 3%. Further analysis indicates that the prefixes constructed by P-ALIGN provide more effective supervision signals, while avoiding the negative impact of redundant and uncertain reasoning components. All code is available at https://github.com/NEUIR/P-ALIGN.

cs.CL↗

arXiv2026

Revealing the Attention Floating Mechanism in Masked Diffusion Models

Masked diffusion models (MDMs), which leverage bidirectional attention and a denoising process, are narrowing the performance gap with autoregressive models (ARMs). However, their internal attention mechanisms remain under-explored. This paper investigates the attention behaviors in MDMs, revealing the phenomenon of Attention Floating. Unlike ARMs, where attention converges to a fixed sink, MDMs exhibit dynamic, dispersed attention anchors that shift across denoising steps and layers. Further analysis reveals its Shallow Structure-Aware, Deep Content-Focused attention mechanism: shallow layers utilize floating tokens to build a global structural framework, while deeper layers allocate more capability toward capturing semantic content. Empirically, this distinctive attention pattern provides a mechanistic explanation for the strong in-context learning capabilities of MDMs, allowing them to double the performance compared to ARMs in knowledge-intensive tasks. All codes and datasets are available at https://github.com/NEUIR/Attention-Floating.

cs.LG↗

arXiv2026

What-If Analysis of Large Language Models: Explore the Game World Using Proactive Thinking

LLMs struggle with decision-making in high-stakes environments like MOBA games, primarily due to a lack of proactive reasoning and limited understanding of complex game dynamics. To address this, we propose What-if Analysis LLM (WiA-LLM), a framework that trains an LLM as an explicit, language-based world model. Instead of representing the environment in latent vectors, WiA-LLM uses natural language to simulate how the game state evolves over time in response to candidate actions, and provides textual justifications for these predicted outcomes. WiA-LLM is trained in two stages: supervised fine-tuning on human-like reasoning traces, followed by reinforcement learning with outcome-based rewards based on the alignment between predicted and actual future states. In the Honor of Kings (HoK) environment, WiA-LLM attains 74.2\% accuracy (27\%$\uparrow$ vs. base model) in forecasting game-state changes. In addition, WiA-LLM demonstrate strategic behavior more closely aligned with expert players than purely reactive LLMs, indicating enhanced foresight and expert-like decision-making.

cs.AI↗

arXiv2025

Lattice QCD calculation of charmed baryon decay constants at continuum limit and physical mass

We present the first principle calculation of charmed baryon decay constants employing 2+1 flavor gauge ensembles with lattice spacings ranging from 0.05 to 0.1 fm and pion masses between 136 and 310 MeV. Under $SU(3)$ flavor symmetry, we construct the charmed baryon interpolating operators and compute the corresponding hadronic matrix elements to extract the bare decay constants for each ensemble. The non-perturbative renormalization is performed via the symmetric momentum-subtraction scheme. After performing systematic chiral and continuum extrapolations, we obtain the decay constants with a precision of $8\sim 16\%$ from first principles.

hep-lat↗

arXiv2025

SimSort: A Data-Driven Framework for Spike Sorting by Large-Scale Electrophysiology Simulation

Spike sorting is an essential process in neural recording, which identifies and separates electrical signals from individual neurons recorded by electrodes in the brain, enabling researchers to study how specific neurons communicate and process information. Although there exist a number of spike sorting methods which have contributed to significant neuroscientific breakthroughs, many are heuristically designed, making it challenging to verify their correctness due to the difficulty of obtaining ground truth labels from real-world neural recordings. In this work, we explore a data-driven, deep learning-based approach. We begin by creating a large-scale dataset through electrophysiology simulations using biologically realistic computational models. We then present SimSort, a pretraining framework for spike sorting. Trained solely on simulated data, SimSort demonstrates zero-shot generalizability to real-world spike sorting tasks, yielding consistent improvements over existing methods across multiple benchmarks. These results highlight the potential of simulation-driven pretraining to enhance the robustness and scalability of spike sorting in experimental neuroscience.

q-bio.NC↗

arXiv2025

JoyVoice: Long-Context Conditioning for Anthropomorphic Multi-Speaker Conversational Synthesis

Large speech generation models are evolving from single-speaker, short sentence synthesis to multi-speaker, long conversation geneartion. Current long-form speech generation models are predominately constrained to dyadic, turn-based interactions. To address this, we introduce JoyVoice, a novel anthropomorphic foundation model designed for flexible, boundary-free synthesis of up to eight speakers. Unlike conventional cascaded systems, JoyVoice employs a unified E2E-Transformer-DiT architecture that utilizes autoregressive hidden representations directly for diffusion inputs, enabling holistic end-to-end optimization. We further propose a MM-Tokenizer operating at a low bitrate of 12.5 Hz, which integrates multitask semantic and MMSE losses to effectively model both semantic and acoustic information. Additionally, the model incorporates robust text front-end processing via large-scale data perturbation. Experiments show that JoyVoice achieves state-of-the-art results in multilingual generation (Chinese, English, Japanese, Korean) and zero-shot voice cloning. JoyVoice achieves top-tier results on both the Seed-TTS-Eval Benchmark and multi-speaker long-form conversational voice cloning tasks, demonstrating superior audio quality and generalization. It achieves significant improvements in prosodic continuity for long-form speech, rhythm richness in multi-speaker conversations, paralinguistic naturalness, besides superior intelligibility. We encourage readers to listen to the demo at https://jea-speech.github.io/JoyVoice

cs.SD↗

arXiv2025

Fluctuation exponents of the half-space KPZ at stationarity

We study the half-space KPZ equation with a Neumann boundary condition, starting from stationary Brownian initial data. We derive a variance identity that links the fluctuations of the height function to the transversal fluctuations of a half-space polymer model. Utilizing this identity, we obtain estimates for the polymer endpoints, leading to optimal fluctuation exponents for the height function in both the subcritical and critical regimes, as well as an optimal upper bound for the fluctuation exponents in the extended critical regime. We also compute the average growth rate as a function of the boundary parameter.

math.PR↗

arXiv2025

Moiré-modulated $Γ$ valley in twisted bilayer and twisted double-bilayer MoTe$_2$

Twisted MoTe$_2$ hosts intriguing correlated quantum phenomena including the fractional quantum anomalous Hall effect in twisted bilayer (t-BL) MoTe$_2$ near 3.7$^\circ$, which is sensitive to the twist angle and moiré superlattices. Here, we directly visualize the twist-angle-modulated electronic structure of t-BL and twisted double-bilayer (t-DBL) near this critical angle. We find that the moiré superlattice not only modifies the relative energy between $Γ$ and K valleys in t-BL MoTe$_2$, but also strongly reconstructs the $Γ$ valley for both t-BL and t-DBL. Specifically, the deep $p_z$-derived band at $Γ$ exhibits a distinct splitting that systematically varies with increasing twist angle. Theoretical analysis suggests that this modulation arises from the twist-angle-dependent lattice relaxation, especially interfacial corrugations. Our work directly visualizes the moiré-modulated electronic structure and provides key spectroscopic information of lattice relaxation and interlayer interactions underlying the physics of twisted MoTe$_2$.

cond-mat.str-el↗

arXiv2025

Periodic Pitman transforms and jointly invariant measures

We construct explicit jointly invariant measures for the periodic KPZ equation (and therefore also the stochastic Burgers' and stochastic heat equations) for general slope parameters and prove their uniqueness via a one force--one solution principle. The measures are given by polymer-like transforms of independent Brownian bridges. We describe several properties and limits of these measures, including an extension to a continuous process in the slope parameter that we term the periodic KPZ horizon. As an application of our construction, we prove a Gaussian process limit theorem with an explicit covariance function for the long-time height function fluctuations of the periodic KPZ equation when started from varying slopes. In connection with this, we conjecture a formula for the fluctuations of cumulants of the endpoint distribution for the periodic continuum directed random polymer. To prove joint invariance, we address the analogous problem for a semi-discrete system of SDEs related to the periodic O'Connell-Yor polymer model and then perform a scaling limit of the model and jointly invariant measures. For the semi-discrete system, we demonstrate a bijection that maps our systems of SDEs to another system with product invariant measure. Inverting the map on this product measure yields our invariant measures. This map relates to a periodic version of the discrete geometric Pitman transform that we introduce and probe. As a by-product of this, we show that the jointly invariant measures for a periodic version of the inverse-gamma polymer are the same as those for the O'Connell-Yor polymer.

math.PR↗

arXiv2025

The Illusion of Readiness in Health AI

Large language models have demonstrated remarkable performance in a wide range of medical benchmarks. Yet underneath the seemingly promising results lie salient growth areas, especially in cutting-edge frontiers such as multimodal reasoning. In this paper, we introduce a series of adversarial stress tests to systematically assess the robustness of flagship models and medical benchmarks. Our study reveals prevalent brittleness in the presence of simple adversarial transformations: leading systems can guess the right answer even with key inputs removed, yet may get confused by the slightest prompt alterations, while fabricating convincing yet flawed reasoning traces. Using clinician-guided rubrics, we demonstrate that popular medical benchmarks vary widely in what they truly measure. Our study reveals significant competency gaps of frontier AI in attaining real-world readiness for health applications. If we want AI to earn trust in healthcare, we must demand more than leaderboard wins and must hold AI systems accountable to ensure robustness, sound reasoning, and alignment with real medical demands.

cs.AI↗