arXiv subjects

Aijun Yang

Publications and source records attributed to Aijun Yang.

10 recordsLinked to original sources

arXiv2026

Self-Correcting Long-Horizon Search Agents via Tree-Structured Memory

Large language model (LLM)-based search agents answer questions through multi-step interactions with external environments. However, providing complete execution trajectories to the LLM causes unbounded context growth and introduces noise. Existing compression methods reduce context at the cost of important details and often replace erroneous facts without repairing downstream reasoning derived from them. To address this problem, we propose ReTree, a self-correcting tree-structured memory mechanism for search agents. ReTree constructs a bounded per-step reasoning context while preserving source-linked evidence. It models search as an evidence tree whose nodes store bounded summaries, evidence, and revision histories. When newly retrieved evidence contradicts an earlier claim, ReTree traces back to the node where the claim was introduced, replaces outdated evidence, regenerates summaries, prunes affected branches, and resumes search. Source-grounded evidence provenance supports reliable conflict localization and keeps final claims traceable to retrieved passages. Experiments on four public question-answering and search benchmarks show that ReTree consistently outperforms Full-Trajectory ReAct, improving answer accuracy by up to 25.6 percentage points (pp); the average maximum per-step reasoning context of Full-Trajectory ReAct is $1.27$--$1.51\times$ that of ReTree. These results establish ReTree as an effective self-correcting memory abstraction for long-horizon search.

cs.AI↗

arXiv2026

Kernel-Smith: A Unified Recipe for Evolutionary Kernel Optimization

We present Kernel-Smith, a framework for high-performance GPU kernel and operator generation that combines a stable evaluation-driven evolutionary agent with an evolution-oriented post-training recipe. On the agent side, Kernel-Smith maintains a population of executable candidates and iteratively improves them using an archive of top-performing and diverse programs together with structured execution feedback on compilation, correctness, and speedup. To make this search reliable, we build backend-specific evaluation services for Triton on NVIDIA GPUs and Maca on MetaX GPUs. On the training side, we convert long-horizon evolution trajectories into step-centric supervision and reinforcement learning signals by retaining correctness-preserving, high-gain revisions, so that the model is optimized as a strong local improver inside the evolutionary loop rather than as a one-shot generator. Under a unified evolutionary protocol, Kernel-Smith-235B-RL achieves state-of-the-art overall performance on KernelBench with Nvidia Triton backend, attaining the best average speedup ratio and outperforming frontier proprietary models including Gemini-3.0-pro and Claude-4.6-opus. We further validate the framework on the MetaX MACA backend, where our Kernel-Smith-MACA-30B surpasses large-scale counterparts such as DeepSeek-V3.2-think and Qwen3-235B-2507-think, highlighting potential for seamless adaptation across heterogeneous platforms. Beyond benchmark results, the same workflow produces upstream contributions to production systems including SGLang and LMDeploy, demonstrating that LLM-driven kernel optimization can transfer from controlled evaluation to practical deployment.

cs.CL↗

arXiv2025

Enhancing the Outcome Reward-based RL Training of MLLMs with Self-Consistency Sampling

Outcome-reward reinforcement learning (RL) is a common and increasingly significant way to refine the step-by-step reasoning of multimodal large language models (MLLMs). In the multiple-choice setting - a dominant format for multimodal reasoning benchmarks - the paradigm faces a significant yet often overlooked obstacle: unfaithful trajectories that guess the correct option after a faulty chain of thought receive the same reward as genuine reasoning, which is a flaw that cannot be ignored. We propose Self-Consistency Sampling (SCS) to correct this issue. For each question, SCS (i) introduces small visual perturbations and (ii) performs repeated truncation and resampling of an initial trajectory; agreement among the resulting trajectories yields a differentiable consistency score that down-weights unreliable traces during policy updates. Based on Qwen2.5-VL-7B-Instruct, plugging SCS into RLOO, GRPO, and REINFORCE++ series improves accuracy by up to 7.7 percentage points on six multimodal benchmarks with negligible extra computation. SCS also yields notable gains on both Qwen2.5-VL-3B-Instruct and InternVL3-8B, offering a simple, general remedy for outcome-reward RL in MLLMs.

cs.CV↗

arXiv2025

EvoSyn: Generalizable Evolutionary Data Synthesis for Verifiable Learning

Reliable verifiable data has become a key driver of capability gains in modern language models, enabling stable reinforcement learning with verifiable rewards and effective distillation that transfers competence across math, coding, and agentic tasks. Yet constructing generalizable synthetic verifiable data remains difficult due to hallucination-prone generation, and weak or trivial verification artifacts that fail to separate strong from weak solutions. Existing approaches often rely on task-specific heuristics or post-hoc filters that do not transfer across domains and lack a principled, universal evaluator of verifiability. In this work, we introduce an evolutionary, task-agnostic, strategy-guided, executably-checkable data synthesis framework that, from minimal seed supervision, jointly synthesizes problems, diverse candidate solutions, and verification artifacts, and iteratively discovers strategies via a consistency-based evaluator that enforces agreement between human-annotated and strategy-induced checks. This pipeline upgrades filtering into principled synthesis: it reliably assembles coherent, verifiable training instances and generalizes without domain-specific rules. Our experiments demonstrate the effectiveness of the proposed approach under both RLVR and model distillation training paradigms. The results show that training with our synthesized data yields significant improvements on both the LiveCodeBench and AgentBench-OS tasks, highlighting the robust generalization of our framework.

cs.LG↗

arXiv2025

Survival Analysis with Discrete Biomarkers Under a Semiparametric Bayesian Conditional Poisson Model

Discrete biomarkers derived as cell densities or counts from tissue microarrays and immunostaining are widely used to study immune signatures in relation to survival outcomes in cancer. Although routinely collected, these signatures are not measured with exact precision because the sampling mechanism involves examination of small tissue cores from a larger section of interest. We model these error-prone biomarkers as Poisson processes with latent rates, inducing heteroscedasticity in their conditional variance. While critical for tumor histology, such measurement error frameworks remain understudied for conditionally Poisson-distributed covariates. To address this, we propose a Bayesian joint model that incorporates a Dirichlet process (DP) mixture to flexibly characterize the latent covariate distribution. The proposed approach is evaluated using simulation studies which demonstrate a superior bias reduction and robustness to the underlying model in realistic settings when compared to existing methods. We further incorporate Bayes factors for hypothesis testing in the Bayesian semiparametric joint model. The methodology is applied to a survival study of high-grade serous carcinoma where comparisons are made between the proposed and existing approaches. Accompanying R software is available at the GitHub repository listed in the Web Appendices.

stat.ME↗

arXiv2025

Strong Consistency of the SIMEX Estimator in Linear Regression with a Conditionally Poisson Covariate

This paper considers estimation for linear regression analysis with covariate measurement error arising from Poisson surrogates. We consider cases where covariates follow a conditional Poisson distribution, capturing non-Gaussian and heteroscedastic error structures. To address this, we extend the simulation extrapolation (SIMEX) algorithm to the conditional Poisson setting (POI-SIMEX), enabling robust adjustment in the absence of internal validation data. Theoretical analysis establishes strong consistency of the POI-SIMEX estimator under a linear regression framework.

math.ST↗

arXiv2025

VisuLogic: A Benchmark for Evaluating Visual Reasoning in Multi-modal Large Language Models

Visual reasoning is a core component of human intelligence and a critical capability for advanced multimodal models. Yet current reasoning evaluations of multimodal large language models (MLLMs) often rely on text descriptions and allow language-based reasoning shortcuts, failing to measure genuine vision-centric reasoning. To address this, we introduce VisuLogic: a benchmark of 1,000 human-verified problems across six categories (e.g., quantitative shifts, spatial relations, attribute comparisons). These various types of questions can be evaluated to assess the visual reasoning capabilities of MLLMs from multiple perspectives. We evaluate leading MLLMs on this benchmark and analyze their results to identify common failure modes. Most models score below 30% accuracy-only slightly above the 25% random baseline and far below the 51.4% achieved by humans-revealing significant gaps in visual reasoning. Furthermore, we provide a supplementary training dataset and a reinforcement-learning baseline to support further progress.

cs.CV↗

arXiv2024

POI-SIMEX for Conditionally Poisson Distributed Biomarkers from Tissue Histology

Covariate measurement error in regression analysis is an important issue that has been studied extensively under the classical additive and the Berkson error models. Here, we consider cases where covariates are derived from tumor tissue histology, and in particular tissue microarrays. In such settings, biomarkers are evaluated from tissue cores that are subsampled from a larger tissue section so that these biomarkers are only estimates of the true cell densities. The resulting measurement error is non-negligible but is seldom accounted for in the analysis of cancer studies involving tissue microarrays. To adjust for this type of measurement error, we assume that these discrete-valued biomarkers are conditionally Poisson distributed, based on a Poisson process model governing the spatial locations of marker-positive cells. Existing methods for addressing conditional Poisson surrogates, particularly in the absence of internal validation data, are limited. We extend the simulation extrapolation (SIMEX) algorithm to accommodate the conditional Poisson case (POI-SIMEX), where measurement errors are non-Gaussian with heteroscedastic variance. The proposed estimator is shown to be strongly consistent in a linear regression model under the assumption of a conditional Poisson distribution for the observed biomarker. Simulation studies evaluate the performance of POI-SIMEX, comparing it with the naive method and an alternative corrected likelihood approach in linear regression and survival analysis contexts. POI-SIMEX is then applied to a study of high-grade serous cancer, examining the association between survival and the presence of triple-positive biomarker (CD3+CD8+FOXP3+ cells)

stat.ME↗

arXiv2024

Power-LLaVA: Large Language and Vision Assistant for Power Transmission Line Inspection

The inspection of power transmission line has achieved notable achievements in the past few years, primarily due to the integration of deep learning technology. However, current inspection approaches continue to encounter difficulties in generalization and intelligence, which restricts their further applicability. In this paper, we introduce Power-LLaVA, the first large language and vision assistant designed to offer professional and reliable inspection services for power transmission line by engaging in dialogues with humans. Moreover, we also construct a large-scale and high-quality dataset specialized for the inspection task. By employing a two-stage training strategy on the constructed dataset, Power-LLaVA demonstrates exceptional performance at a comparatively low training cost. Extensive experiments further prove the great capabilities of Power-LLaVA within the realm of power transmission line inspection. Code shall be released.

cs.CV↗

arXiv2021

Complementary Relation Contrastive Distillation

Knowledge distillation aims to transfer representation ability from a teacher model to a student model. Previous approaches focus on either individual representation distillation or inter-sample similarity preservation. While we argue that the inter-sample relation conveys abundant information and needs to be distilled in a more effective way. In this paper, we propose a novel knowledge distillation method, namely Complementary Relation Contrastive Distillation (CRCD), to transfer the structural knowledge from the teacher to the student. Specifically, we estimate the mutual relation in an anchor-based way and distill the anchor-student relation under the supervision of its corresponding anchor-teacher relation. To make it more robust, mutual relations are modeled by two complementary elements: the feature and its gradient. Furthermore, the low bound of mutual information between the anchor-teacher relation distribution and the anchor-student relation distribution is maximized via relation contrastive loss, which can distill both the sample representation and the inter-sample relations. Experiments on different benchmarks demonstrate the effectiveness of our proposed CRCD.

cs.CV↗