arXiv subjects

Yixuan Wang

Publications and source records attributed to Yixuan Wang.

At least 91 records · Page 5Linked to original sources

arXiv2025

Knowing the Answer Isn't Enough: Fixing Reasoning Path Failures in LVLMs

We reveal a critical yet underexplored flaw in Large Vision-Language Models (LVLMs): even when these models know the correct answer, they frequently arrive there through incorrect reasoning paths. The core issue is not a lack of knowledge, but a path selection bias within the vast reasoning search space. Although LVLMs are often capable of sampling correct solution trajectories, they disproportionately favor unstable or logically inconsistent ones, leading to erratic and unreliable outcomes. The substantial disparity between Pass@K (with large K) and Pass@1 across numerous models provides compelling evidence that such failures primarily stem from misreasoning rather than ignorance. To systematically investigate and address this issue, we propose PSO (Path-Select Optimization), a two-stage post-training framework designed to enhance both the reasoning performance and stability of existing LVLMs. In the first stage, we employ Group Relative Policy Optimization (GRPO) with template and answer-based rewards to cultivate structured, step-by-step reasoning. In the second stage, we conduct online preference optimization, where the model samples reasoning paths from GRPO-generated data, self-evaluates them, and aligns itself toward the preferred trajectories. Incorrect or suboptimal paths are concurrently stored in a Negative Replay Memory (NRM) as hard negatives, which are periodically revisited to prevent the model from repeating prior mistakes and to facilitate continual reasoning refinement. Extensive experiments show that PSO effectively prunes invalid reasoning paths, substantially enhances reasoning accuracy (with 7.4% improvements on average), and yields more stable and consistent chains of thought. Our code will be available at https://github.com/aiming-lab/PSO.

cs.CV↗

arXiv2025

CAMERA: Multi-Matrix Joint Compression for MoE Models via Micro-Expert Redundancy Analysis

Large Language Models (LLMs) with Mixture-of-Experts (MoE) architectures are distinguished by their strong performance scaling with increasing parameters across a wide range of tasks, yet they also suffer from substantial computational and storage overheads. Notably, the performance gains of MoE models do not scale proportionally with the growth in expert parameters. While prior works attempt to reduce parameters via expert-level pruning, merging, or decomposition, they still suffer from challenges in both performance and computational efficiency. In this paper, we address these challenges by introducing micro-expert as a finer-grained compression unit that spans across matrices. We first establish a more fundamental perspective, viewing MoE layers as mixtures of micro-experts, and present CAMERA, a lightweight and training-free framework for identifying micro-expert redundancy. Our analysis uncovers significant variance in micro-expert contributions during decoding. Based on this insight, we further propose CAMERA-P, a structured micro-expert pruning framework, and CAMERA-Q, a mixed-precision quantization idea designed for micro-experts. Extensive experiments on nine downstream tasks show that CAMERA-P consistently outperforms strong baselines under pruning ratios ranging from 20% to 60%. Furthermore, CAMERA-Q achieves superior results under aggressive 2-bit quantization, surpassing existing matrix- and channel-level ideas. Notably, our method enables complete micro-expert analysis of Qwen2-57B-A14B in less than 5 minutes on a single NVIDIA A100-40GB GPU.

cs.CL↗

arXiv2025

Time-aware Motion Planning in Dynamic Environments with Conformal Prediction

Safe navigation in dynamic environments remains challenging due to uncertain obstacle behaviors and the lack of formal prediction guarantees. We propose two motion planning frameworks that leverage conformal prediction (CP): a global planner that integrates Safe Interval Path Planning (SIPP) for uncertainty-aware trajectory generation, and a local planner that performs online reactive planning. The global planner offers distribution-free safety guarantees for long-horizon navigation, while the local planner mitigates inaccuracies in obstacle trajectory predictions through adaptive CP, enabling robust and responsive motion in dynamic environments. To further enhance trajectory feasibility, we introduce an adaptive quantile mechanism in the CP-based uncertainty quantification. Instead of using a fixed confidence level, the quantile is automatically tuned to the optimal value that preserves trajectory feasibility, allowing the planner to adaptively tighten safety margins in regions with higher uncertainty. We validate the proposed framework through numerical experiments conducted in dynamic and cluttered environments. The project page is available at https://time-aware-planning.github.io

cs.RO↗

arXiv2025

Lookahead Q-Cache: Achieving More Consistent KV Cache Eviction via Pseudo Query

Large language models (LLMs) rely on key-value cache (KV cache) to accelerate decoding by reducing redundant computations. However, the KV cache memory usage grows substantially with longer text sequences, posing challenges for efficient deployment. Existing KV cache eviction methods prune tokens using prefilling-stage attention scores, causing inconsistency with actual inference queries, especially under tight memory budgets. In this paper, we propose Lookahead Q-Cache (LAQ), a novel eviction framework that generates low-cost pseudo lookahead queries to better approximate the true decoding-stage queries. By using these lookahead queries as the observation window for importance estimation, LAQ achieves more consistent and accurate KV cache eviction aligned with real inference scenarios. Experimental results on LongBench and Needle-in-a-Haystack benchmarks show that LAQ outperforms existing methods across various budget levels, achieving a 1 $\sim$ 4 point improvement on LongBench under limited cache budget. Moreover, LAQ is complementary to existing approaches and can be flexibly combined to yield further improvements.

cs.CL↗

arXiv2025

Robustifying Learning-Augmented Caching Efficiently without Compromising 1-Consistency

The online caching problem aims to minimize cache misses when serving a sequence of requests under a limited cache size. While naive learning-augmented caching algorithms achieve ideal $1$-consistency, they lack robustness guarantees. Existing robustification methods either sacrifice $1$-consistency or introduce excessive computational overhead. In this paper, we introduce Guard, a lightweight robustification framework that enhances the robustness of a broad class of learning-augmented caching algorithms to $2H_{k-1} + 2$, while preserving their $1$-consistency. Guard achieves the current best-known trade-off between consistency and robustness, with only O(1) additional per-request overhead, thereby maintaining the original time complexity of the base algorithm. Extensive experiments across multiple real-world datasets and prediction models validate the effectiveness of Guard in practice.

cs.DS↗

arXiv2025

Conformal Prediction for Signal Temporal Logic Inference

Signal Temporal Logic (STL) inference seeks to extract human-interpretable rules from time-series data, but existing methods lack formal confidence guarantees for the inferred rules. Conformal prediction (CP) is a technique that can provide statistical correctness guarantees, but is typically applied as a post-training wrapper without improving model learning. Instead, we introduce an end-to-end differentiable CP framework for STL inference that enhances both reliability and interpretability of the resulting formulas. We introduce a robustness-based nonconformity score, embed a smooth CP layer directly into training, and employ a new loss function that simultaneously optimizes inference accuracy and CP prediction sets with a single term. Following training, an exact CP procedure delivers statistical guarantees for the learned STL formulas. Experiments on benchmark time-series tasks show that our approach reduces uncertainty in predictions (i.e., it achieves high coverage while reducing prediction set size), and improves accuracy (i.e., the number of misclassifications when using a fixed threshold) over state-of-the-art baselines.

cs.LG↗

arXiv2025

Automated Snippet-Alignment Data Augmentation for Code Translation

Code translation aims to translate the code from its source language to the target language and is used in various software development scenarios. Recent developments in Large Language Models (LLMs) have showcased their capabilities in code translation, and parallel corpora play a crucial role in training models for code translation. Parallel corpora can be categorized into program-alignment (PA) and snippet-alignment (SA) data. Although PA data has complete context and is suitable for semantic alignment learning, it may not provide adequate fine-grained training signals due to its extended length, while the brevity of SA data enables more fine-grained alignment learning. Due to limited parallel corpora, researchers explore several augmentation methods for code translation. Previous studies mainly focus on augmenting PA data. In this paper, we propose a data augmentation method that leverages LLMs to generate SA data automatically. To fully leverage both PA data and SA data, we explore a simple yet effective two-stage training strategy, which consistently enhances model performance compared to fine-tuning solely on PA data. Experiments on TransCoder-test demonstrate that our augmented SA data combined with the two-stage training approach yields consistent improvements over the baseline, achieving a maximum gain of 3.78% on pass@k.

cs.SE↗

arXiv2025

RE$^2$: Improving Chinese Grammatical Error Correction via Retrieving Appropriate Examples with Explanation

The primary objective of Chinese grammatical error correction (CGEC) is to detect and correct errors in Chinese sentences. Recent research shows that large language models (LLMs) have been applied to CGEC with significant results. For LLMs, selecting appropriate reference examples can help improve their performance. However, existing methods predominantly rely on text similarity for example retrieval, a strategy that frequently mismatches actual error patterns and retrieves lexically similar yet grammatically irrelevant sentences. To address this problem, we propose a method named RE$^2$, which retrieves appropriate examples with explanations of grammatical errors. Instead of using text similarity of the input sentence, we use explanations of grammatical errors to select reference examples, which are used by LLMs to improve the performance of CGEC. We conduct experiments on two CGEC datasets and create a high-quality grammatical error explanation (GEE) dataset, which is not only used in our research but also serves as a valuable resource for future studies in both CGEC and GEE. The experimental results on the two datasets indicate that our proposed method effectively improves the performance of CGEC.

cs.CL↗

arXiv2025

CommonKV: Compressing KV Cache with Cross-layer Parameter Sharing

Large Language Models (LLMs) confront significant memory challenges due to the escalating KV cache with increasing sequence length. As a crucial technique, existing cross-layer KV cache sharing methods either necessitate modified model architectures with subsequent pre-training or incur significant performance degradation at high compression rates. To mitigate these challenges, we propose CommonKV, a training-free method for cross-layer KV cache compression through adjacent parameters sharing. Inspired by the high similarity observed in cross-layer hidden states, we utilize Singular Value Decomposition (SVD) to achieve weight sharing across adjacent parameters, resulting in a more easily mergeable latent KV cache. Furthermore, we also introduce an adaptive budget allocation strategy. It dynamically assigns compression budgets based on cosine similarity, ensuring that dissimilar caches are not over-compressed. Experiments across multiple backbone models and benchmarks including LongBench and Ruler demonstrate that the proposed method consistently outperforms existing low-rank and cross-layer approaches at various compression ratios. Moreover, we find that the benefits of CommonKV are orthogonal to other quantization and eviction methods. By integrating these approaches, we can ultimately achieve a 98\% compression ratio without significant performance loss.

cs.LG↗

arXiv2025

Finite time blowup for Keller-Segel equation with logistic damping in three dimensions

The Keller-Segel equation, a classical chemotaxis model, and many of its variants have been extensively studied for decades. In this work, we focus on 3D Keller-Segel equation with a quadratic logistic damping term $-μρ^2$ (modeling density-dependent mortality rate) and show the existence of finite-time blowup solutions with nonnegative density and finite mass for any $μ\in \big[0,\frac{1}{3}\big)$. This range of $μ$ is sharp; for $μ\ge \frac{1}{3}$, the logistic damping effect suppresses the blowup as shown in [Kang-Stevens, 2016] and [Tello-Winkler, 2007]. A key ingredient is to construct a self-similar blowup solution to a related aggregation equation as an approximate solution, with subcritical scaling relative to the original model. Based on this construction, we employ a robust weighted $L^2$ method to prove the stability of this approximate solution, where modulation ODEs are introduced to enforce local vanishing conditions for the perturbation lying in a singular-weighted $L^2$ space. As a byproduct, we exhibit a new family of type I blowup mechanisms for the classical 3D Keller-Segel equation.

math.AP↗

arXiv2025

Scale-Consistent Learning for Partial Differential Equations

Machine learning (ML) models have emerged as a promising approach for solving partial differential equations (PDEs) in science and engineering. Previous ML models typically cannot generalize outside the training data; for example, a trained ML model for the Navier-Stokes equations only works for a fixed Reynolds number ($Re$) on a pre-defined domain. To overcome these limitations, we propose a data augmentation scheme based on scale-consistency properties of PDEs and design a scale-informed neural operator that can model a wide range of scales. Our formulation leverages the facts: (i) PDEs can be rescaled, or more concretely, a given domain can be re-scaled to unit size, and the parameters and the boundary conditions of the PDE can be appropriately adjusted to represent the original solution, and (ii) the solution operators on a given domain are consistent on the sub-domains. We leverage these facts to create a scale-consistency loss that encourages matching the solutions evaluated on a given domain and the solution obtained on its sub-domain from the rescaled PDE. Since neural operators can fit to multiple scales and resolutions, they are the natural choice for incorporating scale-consistency loss during training of neural PDE solvers. We experiment with scale-consistency loss and the scale-informed neural operator model on the Burgers' equation, Darcy Flow, Helmholtz equation, and Navier-Stokes equations. With scale-consistency, the model trained on $Re$ of 1000 can generalize to $Re$ ranging from 250 to 10000, and reduces the error by 34% on average of all datasets compared to baselines.

cs.LG↗

arXiv2025

High precision PINNs in unbounded domains: application to singularity formulation in PDEs

We investigate the high-precision training of Physics-Informed Neural Networks (PINNs) in unbounded domains, with a special focus on applications to singularity formulation in PDEs. We propose a modularized approach and study the choices of neural network ansatz, sampling strategy, and optimization algorithm. When combined with rigorous computer-assisted proofs and PDE analysis, the numerical solutions identified by PINNs, provided they are of high precision, can serve as a powerful tool for studying singularities in PDEs. For 1D Burgers equation, our framework can lead to a solution with very high precision, and for the 2D Boussinesq equation, which is directly related to the singularity formulation in 3D Euler and Navier-Stokes equations, we obtain a solution whose loss is $4$ digits smaller than that obtained in \cite{wang2023asymptotic} with fewer training steps. We also discuss potential directions for pushing towards machine precision for higher-dimensional problems.

cs.LG↗

arXiv2025

CodeDiffuser: Attention-Enhanced Diffusion Policy via VLM-Generated Code for Instruction Ambiguity

Natural language instructions for robotic manipulation tasks often exhibit ambiguity and vagueness. For instance, the instruction "Hang a mug on the mug tree" may involve multiple valid actions if there are several mugs and branches to choose from. Existing language-conditioned policies typically rely on end-to-end models that jointly handle high-level semantic understanding and low-level action generation, which can result in suboptimal performance due to their lack of modularity and interpretability. To address these challenges, we introduce a novel robotic manipulation framework that can accomplish tasks specified by potentially ambiguous natural language. This framework employs a Vision-Language Model (VLM) to interpret abstract concepts in natural language instructions and generates task-specific code - an interpretable and executable intermediate representation. The generated code interfaces with the perception module to produce 3D attention maps that highlight task-relevant regions by integrating spatial and semantic information, effectively resolving ambiguities in instructions. Through extensive experiments, we identify key limitations of current imitation learning methods, such as poor adaptation to language and environmental variations. We show that our approach excels across challenging manipulation tasks involving language ambiguity, contact-rich manipulation, and multi-object interactions.

cs.RO↗

arXiv2025

Directly Forecasting Belief for Reinforcement Learning with Delays

Reinforcement learning (RL) with delays is challenging as sensory perceptions lag behind the actual events: the RL agent needs to estimate the real state of its environment based on past observations. State-of-the-art (SOTA) methods typically employ recursive, step-by-step forecasting of states. This can cause the accumulation of compounding errors. To tackle this problem, our novel belief estimation method, named Directly Forecasting Belief Transformer (DFBT), directly forecasts states from observations without incrementally estimating intermediate states step-by-step. We theoretically demonstrate that DFBT greatly reduces compounding errors of existing recursively forecasting methods, yielding stronger performance guarantees. In experiments with D4RL offline datasets, DFBT reduces compounding errors with remarkable prediction accuracy. DFBT's capability to forecast state sequences also facilitates multi-step bootstrapping, thus greatly improving learning efficiency. On the MuJoCo benchmark, our DFBT-based method substantially outperforms SOTA baselines. Code is available at https://github.com/QingyuanWuNothing/DFBT.

cs.LG↗

arXiv2025

Second Order Ensemble Langevin Method for Sampling and Inverse Problems

We propose a sampling method based on an ensemble approximation of second order Langevin dynamics. The log target density is appended with a quadratic term in an auxiliary momentum variable and damped-driven Hamiltonian dynamics introduced; the resulting stochastic differential equation is invariant to the Gibbs measure, with marginal on the position coordinates given by the target. A preconditioner based on covariance under the law of the dynamics does not change this invariance property, and is introduced to accelerate convergence to the Gibbs measure. The resulting mean-field dynamics may be approximated by an ensemble method; this results in a gradient-free and affine-invariant stochastic dynamical system. Numerical results demonstrate its potential as the basis for a numerical sampler in Bayesian inverse problems.

math.DS↗

arXiv2025

Tag-Evol: Achieving Efficient Instruction Evolving via Tag Injection

Evol-Instruct has made significant improvements as a data synthesis method in several areas. Existing methods typically rely on a fixed set of strategies to evolve, which require manual design and are monolithic in form. In addition, iterative evolution also makes the acquisition of hard samples expensive. In view of this, we propose the Tag-Evol framework, a more diverse and efficient instruction evolving method. Specifically, Tag-Evol uses diverse and specific knowledge tags as strategies to achieve controlled evolution by injecting different combinations of tags into the original instructions. Experiments with multiple backbones in diverse domain benchmarks show that the proposed method generates significantly better evolved data than other methods. Furthermore, we conduct a thorough analysis of the evolved data, demonstrating that Tag-Evol is not only efficient but also generates more diverse and challenging data.

cs.CL↗

arXiv2025

Turning Trash into Treasure: Accelerating Inference of Large Language Models with Token Recycling

Massive parameters of LLMs have made inference latency a fundamental bottleneck. Speculative decoding represents a lossless approach to accelerate inference through a guess-and-verify paradigm. Some methods rely on additional architectures to guess draft tokens, which need extra training before use. Alternatively, retrieval-based training-free techniques build libraries from pre-existing corpora or by n-gram generation. However, they face challenges like large storage requirements, time-consuming retrieval, and limited adaptability. Observing that candidate tokens generated during the decoding process are likely to reoccur in future sequences, we propose Token Recycling. It stores candidate tokens in an adjacency matrix and employs a breadth-first-search (BFS)-like algorithm to construct a draft tree, which is then validated through tree attention. New candidate tokens from the decoding process are then used to update the matrix. Token Recycling requires \textless2MB of additional storage and achieves approximately 2x speedup across all sizes of LLMs. It significantly outperforms existing train-free methods by 30\% and even a widely recognized training method by 25\%.

cs.CL↗

arXiv2025

Think Before You Accept: Semantic Reflective Verification for Faster Speculative Decoding

Large language models (LLMs) suffer from high inference latency due to the auto-regressive decoding process. Speculative decoding accelerates inference by generating multiple draft tokens using a lightweight model and verifying them in parallel. However, existing verification methods rely heavily on distributional consistency while overlooking semantic correctness, thereby limiting the potential speedup of speculative decoding. While some methods employ additional models for relaxed verification of draft tokens, they often fail to generalize effectively to more diverse or open-domain settings. In this work, we propose Reflective Verification, a training-free and semantics-aware approach that achieves a better trade-off between correctness and efficiency. Specifically, we leverage the inherent reflective capacity of LLMs to semantically assess the correctness of draft tokens in parallel during verification. Using prompt-based probing, we obtain both the original and reflective distributions of draft tokens in a single forward pass. The fusion of these distributions enables semantic-level verification of draft tokens that incorporates both consistency and correctness. Experiments across multiple domain benchmarks and model scales demonstrate that our method significantly increases the acceptance length of draft tokens without compromising model performance. Furthermore, we find that the proposed Reflective Verification is orthogonal to existing statistical verification methods, and their combination yields additional 5$\sim$15\% improvements in decoding speed.

cs.LG↗