Searcharxiv⌕ Search

arXiv subjects

Yiming Gao

Publications and source records attributed to Yiming Gao.

At least 19 recordsLinked to original sources
arXiv2026

PertReason: A Knowledge-Grounded Benchmark and Framework for Cell-State-Conditioned Mechanistic Reasoning of Perturbation Effects

Evaluating machine learning in scientific domains requires separating correct predictions from correct reasons under realistic distribution shifts. We introduce PertReason, a knowledge-grounded benchmark and framework suite for cell-state--conditioned reasoning about perturbation effects. At its core, PertReasonQA is a benchmark that tests whether models can generate mechanistically faithful explanations while remaining robust to complex shifts, such as new cells and unseen perturbations. PertReasonQA combines single-cell genetic and chemical perturbation data across multiple cellular contexts with knowledge graphs, and dynamically conditions pathways on cell-specific basal states to avoid generic memorization. Evaluations on state-of-the-art models reveal systematic gaps between predictive accuracy and mechanistic reasoning. Specifically, these models exhibit failure modes largely invisible to standard benchmarks, such as deriving correct answers through flawed logic, ignoring cellular context, and generating directionally inconsistent mechanisms. As a reference probe of the benchmark, we present PertReasonLM, a large language model trained to align outcome predictions with context-specific mechanistic reasoning. Our model targets the identified failure modes by grounding rationales in context-specific pathways and tightening agreement between outcomes and mechanisms. Together, we provide a diagnostic framework for exposing and mitigating failures in faithful reasoning in data-rich scientific systems.

cs.LG↗
arXiv2026

DiStash: A Disaggregated Multi-Stash Transactional Key-Value Store

A stash is a storage medium such as Dynamic Random Access Memory (DRAM), Solid State Disk (SSD), Hard Disk Drive (HDD), or Non-Volatile Memory (NVM). This paper presents a disaggregated transactional key-value (KV) store, DiStash, that governs KVs cross pools of stash types. It enables an application to use a single transaction to read and write different copies of one or more key-value pair across the different pools of stashes. It simplifies the application logic by (a) preventing undesirable race conditions that may cause copies of data across different stash pools to reflect different values and/or (b) failures that may result in loss of key-value pairs. A configuration of DiStash may use a pool of stashes as either ephemeral or durable storage. The application dictates whether the content of its participating stashes are inclusive (replicated) or exclusive (tiered). We implement a DiStash by extending FoundationDB. We quantify the tradeoffs with its design decisions using microbenchmarks and eBay's production workload. We open source our implementation at https://github.com/ebay-USC/DiStash.

cs.DB↗
arXiv2026

Senses Wide Shut: A Representation-Action Gap in Omnimodal LLMs

When an omnimodal large language model accepts a question whose textual premise contradicts what it actually sees or hears, does the failure lie in perception or in action? Recent omnimodal models are positioned as perception-grounded agents that jointly process video, audio, and text, yet a basic form of grounding remains untested: catching a textual claim that conflicts with the model's own sensory input. We introduce IMAVB, a curated 500-clip benchmark of long-form movies with a 2x2 design crossing target modality (vision, audio) and premise condition (standard, misleading), which lets us measure conflict detection separately from ordinary multimodal comprehension. Across eight open-source omnimodal LLMs and Gemini 3.1 Pro, we document a Representation-Action Gap: hidden states reliably encode premise-perception mismatches even when the same models almost never reject the false claim in their outputs. Behaviorally, models fall into two failure modes: under-rejection, in which they answer misleading questions as if the false premise were true; and over-rejection, in which they reject more often but also reject standard questions, sacrificing ordinary comprehension accuracy. The gap is modality-asymmetric (audio grounding underperforms vision) and prompt-resistant across seven variants. As an initial diagnostic intervention, a probe-guided logit adjustment (PGLA) re-injects the encoded mismatch signal into decoding and consistently improves rejection behavior. Together, these results suggest the bottleneck for omnimodal grounding lies in translation, not perception.

cs.AI↗
arXiv2026

Subset Balancing and Generalized Subset Sum via Lattices

We study the Subset Balancing problem: given $x \in \mathbb{Z}^n$ and a coefficient set $C \subseteq \mathbb{Z}$, find a nonzero vector $c \in C^n$ such that $c\cdot x = 0$. The standard meet-in-the-middle algorithm runs in time $\tilde{O}(|C|^{n/2})$, and recent improvements (SODA 2022, Chen, Jin, Randolph, and Servedio; STOC 2026, Randolph and Węgrzycki) beyond this barrier apply mainly when $d$ is constant. We give a reduction from Subset Balancing with $C = \{-d, \dots, d\}$ to a single instance of SVP$_{\infty}$ in dimension $n+1$. Instantiating this reduction with the best known $\ell_\infty$-SVP algorithms yields a deterministic $\tilde{O}((6\sqrt{2πe})^n)$-time algorithm and a randomized $\tilde{O}(2^{2.443n})$-time algorithm. The exponent depends only on $n$, improving on meet-in-the-middle for all $d\ge 15$. For sufficiently large $d$ we also obtain a polynomial-time algorithm. The reduction extends from the box constraint $[-d,d]^n$ to any centrally symmetric convex body $K\subseteq\mathbb{R}^n$, giving deterministic time $\tilde{O}(2^{c_K n})$ for a constant $c_K$ depending only on the shape of $K$. We further study the Generalized Subset Sum problem of finding $c \in C^n$ such that $c \cdot x = τ$. For $C = \{-d, \dots, d\}$ or $C = \{-d,\dots,d\}\setminus\{0\}$, we reduce the worst-case problem to CVP$_{\infty}$ in dimension $n+1$. We observe that distances in our lattice take only integer values, so an approximate CVP$_{\infty}$ oracle still suffices, yielding a deterministic worst-case algorithm running in time $2^{O(n\log\log d)}$. In the average-case setting, we demonstrate that for both coefficient sets the embedded CVP$_{\infty}$ instance satisfies a bounded-distance promise with high probability, removing the $\log\log d$ factor altogether and obtaining a deterministic algorithm running in time $\tilde{O}((18\sqrt{2πe})^n)$.

cs.DS↗
arXiv2026

L-PCN: A Point Cloud Accelerator Exploiting Spatial Locality through Octree-based Islandization

Existing Point Cloud Networks (PCNs) have proven to achieve great success in many point cloud tasks such as object part segmentation, shape classification, and so on. The most popular point-based PCNs are usually composed of two sequential steps: Data Structuring (DS) and Feature Computation (FC). In this paper, we first describe an important characteristic of the PCN-specific DS step that has not been addressed in existing PCN accelerators: the spatial locality resulting from overlapping points of the gathered point subsets. Using algorithm-hardware co-design, L-PCN (Locality-aware PCN) proposes two novel techniques to exploit this characteristic to reduce the large amount of repetitive operations in the overall PCN. The first of which is a point cloud partitioning technique, Octree-based Islandization. Using Octree-based adjacency gathering, a point cloud is partitioned into islands in L-PCN, where the point subsets inside the same island exhibit a strong spatial correlation. After partitioning, L-PCN performs the rest of PCN steps at the granularity of islands. The second method of L-PCN is scheduling the intra-island computation with a Hub-based Scheduling to exploit the intra-island data reuse by dynamically caching, updating, and reusing the repeated data. The two methods are implemented in an Islandization Unit, which can be seamlessly integrated into standard PCN workflow. Our evaluation shows that based on our methods for exploiting spatial locality, L-PCN achieves a theoretical reduction in feature fetching ranging from 55.2% to 93.8% and in feature computation ranging from 45.4% to 80.6% during the PCN process. For experimentation, prototype L-PCN accelerators are implemented on the Intel Arria 10 GX FPGA. Experimental results prove that with the Islandization Unit as a plug-in, state-of-the-art PCN accelerators can achieve an additional speedup ranging from 1.2x to 3.2x.

cs.AR↗
arXiv2026

UI-Voyager: A Self-Evolving GUI Agent Learning via Failed Experience

Autonomous mobile GUI agents have attracted increasing attention along with the advancement of Multimodal Large Language Models (MLLMs). However, existing methods still suffer from inefficient learning from failed trajectories and ambiguous credit assignment under sparse rewards for long-horizon GUI tasks. To that end, we propose UI-Voyager, a novel two-stage self-evolving mobile GUI agent. In the first stage, we employ Rejection Fine-Tuning (RFT), which enables the continuous co-evolution of data and models in a fully autonomous loop. The second stage introduces Group Relative Self-Distillation (GRSD), which identifies critical fork points in group rollouts and constructs dense step-level supervision from successful trajectories to correct failed ones. Extensive experiments on AndroidWorld show that our 4B model achieves an 81.0% Pass@1 success rate, outperforming numerous recent baselines and exceeding human-level performance. Ablation and case studies further verify the effectiveness of GRSD. Our method represents a significant leap toward efficient, self-evolving, and high-performance mobile GUI automation without expensive manual data annotation.

cs.LG↗
arXiv2026

Multi-Task Genetic Algorithm with Multi-Granularity Encoding for Protein-Nucleotide Binding Site Prediction

Accurate identification of protein-nucleotide binding sites is fundamental to deciphering molecular mechanisms and accelerating drug discovery. However, current computational methods often struggle with suboptimal performance due to inadequate feature representation and rigid fusion mechanisms, which hinder the effective exploitation of cross-task information synergy. To bridge this gap, we propose MTGA-MGE, a framework that integrates a Multi-Task Genetic Algorithm with Multi-Granularity Encoding to enhance binding site prediction. Specifically, we develop a Multi-Granularity Encoding (MGE) network that synergizes multi-scale convolutions and self-attention mechanisms to distill discriminative signals from high-dimensional, redundant biological data. To overcome the constraints of static fusion, a genetic algorithm is employed to adaptively evolve task-specific fusion strategies, thereby effectively improving model generalization. Furthermore, to catalyze collaborative learning, we introduce an External-Neighborhood Mechanism (ENM) that leverages biological similarities to facilitate targeted information exchange across tasks. Extensive evaluations on fifteen nucleotide datasets demonstrate that MTGA-MGE not only establishes a new state-of-the-art in data-abundant, high-resource scenarios but also maintains a robust competitive edge in rare, low-resource regimes, presenting a highly adaptive scheme for decoding complex protein-ligand interactions in the post-genomic era.

cs.LG↗
arXiv2026

scPilot: Large Language Model Reasoning Toward Automated Single-Cell Analysis and Discovery

We present scPilot, the first systematic framework to practice omics-native reasoning: a large language model (LLM) converses in natural language while directly inspecting single-cell RNA-seq data and on-demand bioinformatics tools. scPilot converts core single-cell analyses, i.e., cell-type annotation, developmental-trajectory reconstruction, and transcription-factor targeting, into step-by-step reasoning problems that the model must solve, justify, and, when needed, revise with new evidence. To measure progress, we release scBench, a suite of 9 expertly curated datasets and graders that faithfully evaluate the omics-native reasoning capability of scPilot w.r.t various LLMs. Experiments with o1 show that iterative omics-native reasoning lifts average accuracy by 11% for cell-type annotation and Gemini-2.5-Pro cuts trajectory graph-edit distance by 30% versus one-shot prompting, while generating transparent reasoning traces explain marker gene ambiguity and regulatory logic. By grounding LLMs in raw omics data, scPilot enables auditable, interpretable, and diagnostically informative single-cell analyses. Code, data, and package are available at https://github.com/maitrix-org/scPilot

cs.AI↗
arXiv2026

CellMaster: Collaborative Cell Type Annotation in Single-Cell Analysis

Single-cell RNA-seq (scRNA-seq) enables atlas-scale profiling of complex tissues, revealing rare lineages and transient states. Yet, assigning biologically valid cell identities remains a bottleneck because markers are tissue- and state-dependent, and novel states lack references. We present CellMaster, an AI agent that mimics expert practice for zero-shot cell-type annotation. Unlike existing automated tools, CellMaster leverages LLM-encoded knowledge (e.g., GPT-4o) to perform on-the-fly annotation with interpretable rationales, without pre-training or fixed marker databases. Across 9 datasets spanning 8 tissues, CellMaster improved accuracy by 7.1% over best-performing baselines (including CellTypist and scTab) in automatic mode. With human-in-the-loop refinement, this advantage increased to 18.6%, with a 22.1% gain on subtype populations. The system demonstrates particular strength in rare and novel cell states where baselines often fail. Source code and the web application are available at \href{https://github.com/AnonymousGym/CellMaster}{https://github.com/AnonymousGym/CellMaster}.

q-bio.GN↗
arXiv2026

Compendia: Automated Visual Storytelling Generation from Online Article Collection

In the digital age, readers value quantitative journalism that is clear, concise, analytical, and human-centred. To understand complex topics, they often piece together scattered facts from multiple articles. Visual storytelling can transform fragmented information into clear, engaging narratives, yet its use with unstructured online articles remains largely unexplored. To fill this gap, we present Compendia, an automated system that analyzes online articles in response to a user's query and generates a coherent data story tailored to the user's informational needs. Compendia addresses key challenges of storytelling from unstructured text through two modules covering: Online Article Retrieval, which gathers relevant articles; Data Fact Extraction, which identifies, validates, and refines quantitative facts; Fact Organization, which clusters and merges related facts into coherent thematic groups; and Visual Storytelling, which transforms the organized facts into narratives with visualizations in an interactive scrollytelling interface. We evaluated Compendia through a quantitative analysis, confirming the accuracy in fact extraction and organization, and through two user studies with 16 participants, demonstrating its usability, effectiveness, and ability to produce engaging visual stories for open-ended queries.

cs.HC↗
arXiv2026

ProAct: Agentic Lookahead in Interactive Environments

Existing Large Language Model (LLM) agents struggle in interactive environments requiring long-horizon planning, primarily due to compounding errors when simulating future states. To address this, we propose ProAct, a framework that enables agents to internalize accurate lookahead reasoning through a two-stage training paradigm. First, we introduce Grounded LookAhead Distillation (GLAD), where the agent undergoes supervised fine-tuning on trajectories derived from environment-based search. By compressing complex search trees into concise, causal reasoning chains, the agent learns the logic of foresight without the computational overhead of inference-time search. Second, to further refine decision accuracy, we propose the Monte-Carlo Critic (MC-Critic), a plug-and-play auxiliary value estimator designed to enhance policy-gradient algorithms like PPO and GRPO. By leveraging lightweight environment rollouts to calibrate value estimates, MC-Critic provides a low-variance signal that facilitates stable policy optimization without relying on expensive model-based value approximation. Experiments on both stochastic (e.g., 2048) and deterministic (e.g., Sokoban) environments demonstrate that ProAct significantly improves planning accuracy. Notably, a 4B parameter model trained with ProAct outperforms all open-source baselines and rivals state-of-the-art closed-source models, while demonstrating robust generalization to unseen environments. The codes and models are available at https://github.com/GreatX3/ProAct

cs.AI↗
arXiv2026

Artificial intelligence pioneers the double-strangeness factory

Artificial intelligence (AI) is transforming not only our daily experiences but also the technological development landscape and scientific research. In this study, we pioneered the application of AI in double-strangeness hypernuclear studies. These studies which investigate quantum systems with strangeness via hyperon interactions provide insights into fundamental baryon-baryon interactions and contribute to our understanding of the nuclear force and composition of neutron star cores. Specifically, we report the observation of a double hypernucleus in nuclear emulsion achieved via innovative integration of machine learning techniques. The proposed methodology leverages generative AI and Monte Carlo simulations to produce training datasets combined with object detection AI for effective event identification. Based on the kinematic analysis and charge identification, the observed event was uniquely identified as the production and decay of resulting from Ξ- capture by 14N in the nuclear emulsion. Assuming capture in the atomic 3D state, the binding energy of the two Λ hyperons in 13BΛΛ, BΛΛ, was determined as 25.57 +- 1.18(stat.) +- 0.07(syst.) MeV. The ΛΛ interaction energy obtained was 2.83 +- 1.18(stat.) +- 0.14(syst.) MeV. This study marks a new era in double-strangeness research.

nucl-ex↗
arXiv2026

NextFlow: Unified Sequential Modeling Activates Multimodal Understanding and Generation

We present NextFlow, a unified decoder-only autoregressive transformer trained on 6 trillion interleaved text-image discrete tokens. By leveraging a unified vision representation within a unified autoregressive architecture, NextFlow natively activates multimodal understanding and generation capabilities, unlocking abilities of image editing, interleaved content and video generation. Motivated by the distinct nature of modalities - where text is strictly sequential and images are inherently hierarchical - we retain next-token prediction for text but adopt next-scale prediction for visual generation. This departs from traditional raster-scan methods, enabling the generation of 1024x1024 images in just 5 seconds - orders of magnitude faster than comparable AR models. We address the instabilities of multi-scale generation through a robust training recipe. Furthermore, we introduce a prefix-tuning strategy for reinforcement learning. Experiments demonstrate that NextFlow achieves state-of-the-art performance among unified models and rivals specialized diffusion baselines in visual quality.

cs.CV↗
arXiv2025

On Factoring and Power Divisor Problems via Rank-3 Lattices and the Second Vector

We propose a deterministic algorithm based on Coppersmith's method that employs a rank-3 lattice to address factoring-related problems. An interesting aspect of our approach is that we utilize the second vector in the LLL-reduced basis to avoid trivial collisions in the Baby-step Giant-step method, rather than the shortest vector as is commonly used in the literature. Our results are as follows: 1. Compared to the result by Harvey and Hittmeir (Math. Comp. 91 (2022), 1367 - 1379), who achieved a complexity of O( N^(1/5) log^(16/5) N / (log log N)^(3/5)) for factoring a semiprime N = pq, we demonstrate that in the balanced p and q case, the complexity can be improved to O( N^(1/5) log^(13/5) N / (log log N)^(3/5) ). 2. For factoring sums and differences of powers, that is, numbers of the form N = a^n plus or minus b^n, we improve Hittmeir's result (Math. Comp. 86 (2017), 2947 - 2954) from O( N^(1/4) log^(3/2) N ) to O( N^(1/5) log^(13/5) N ). 3. For the problem of finding r-power divisors, that is, finding all integers p such that p^r divides N, Harvey and Hittmeir (Proceedings of ANTS XV, Research in Number Theory 8 (2022), no. 4, Paper No. 94) recently directly applied Coppersmith's method and achieved a complexity of O( N^(1/(4r)) log^(10+epsilon) N / r^3 ). By using faster LLL-type algorithms and sieving on small primes, we improve their result to O( N^(1/(4r)) log^(7+3 epsilon) N / ((log log N minus log(4r)) r^(2+epsilon)) ). The worst-case running time for their algorithm occurs when N = p^r q with q on the order of N^(1/2). By focusing on this case and employing our rank-3 lattice approach, we achieve a complexity of O( r^(1/4) N^(1/(4r)) log^(5/2) N ). In conclusion, we offer a new perspective on these problems, which we hope will provide further insights.

math.NT↗
arXiv2025

RELIC-GNN: Efficient State Registers Identification with Graph Neural Network for Reverse Engineering

Reverse engineering of gate-level netlist is critical for Hardware Trojans detection and Design Piracy counteracting. The primary task of gate-level reverse engineering is to separate the control and data signals from the netlist, which is mainly realized by identifying state registers with topological comparison.However, these methods become inefficient for large scale netlist. In this work, we propose RELIC-GNN, a graph neural network based state registers identification method, to address these issues. RELIC-GNN models the path structure of register as a graph and generates corresponding representation by considering node attributes and graph structure during training. The trained GNN model could be adopted to find the registers type very efficiently. Experimental results show that RELIC-GNN could achieve 100% in recall, 30.49% in precision and 88.37% in accuracy on average across different designs, which obtains significant improvements than previous approaches.

cs.CR↗
arXiv2025

IFEval-Audio: Benchmarking Instruction-Following Capability in Audio-based Large Language Models

Large language models (LLMs) have demonstrated strong instruction-following capabilities in text-based tasks. However, this ability often deteriorates in multimodal models after alignment with non-text modalities such as images or audio. While several recent efforts have investigated instruction-following performance in text and vision-language models, instruction-following in audio-based large language models remains largely unexplored. To bridge this gap, we introduce IFEval-Audio, a novel evaluation dataset designed to assess the ability to follow instructions in an audio LLM. IFEval-Audio contains 280 audio-instruction-answer triples across six diverse dimensions: Content, Capitalization, Symbol, List Structure, Length, and Format. Each example pairs an audio input with a text instruction, requiring the model to generate an output that follows a specified structure. We benchmark state-of-the-art audio LLMs on their ability to follow audio-involved instructions. The dataset is released publicly to support future research in this emerging area.

cs.CL↗
arXiv2025

DetailFlow: 1D Coarse-to-Fine Autoregressive Image Generation via Next-Detail Prediction

This paper presents DetailFlow, a coarse-to-fine 1D autoregressive (AR) image generation method that models images through a novel next-detail prediction strategy. By learning a resolution-aware token sequence supervised with progressively degraded images, DetailFlow enables the generation process to start from the global structure and incrementally refine details. This coarse-to-fine 1D token sequence aligns well with the autoregressive inference mechanism, providing a more natural and efficient way for the AR model to generate complex visual content. Our compact 1D AR model achieves high-quality image synthesis with significantly fewer tokens than previous approaches, i.e. VAR/VQGAN. We further propose a parallel inference mechanism with self-correction that accelerates generation speed by approximately 8x while reducing accumulation sampling error inherent in teacher-forcing supervision. On the ImageNet 256x256 benchmark, our method achieves 2.96 gFID with 128 tokens, outperforming VAR (3.3 FID) and FlexVAR (3.05 FID), which both require 680 tokens in their AR models. Moreover, due to the significantly reduced token count and parallel inference mechanism, our method runs nearly 2x faster inference speed compared to VAR and FlexVAR. Extensive experimental results demonstrate DetailFlow's superior generation quality and efficiency compared to existing state-of-the-art methods.

cs.CV↗
arXiv2025

TokenFlow: Unified Image Tokenizer for Multimodal Understanding and Generation

We present TokenFlow, a novel unified image tokenizer that bridges the long-standing gap between multimodal understanding and generation. Prior research attempt to employ a single reconstruction-targeted Vector Quantization (VQ) encoder for unifying these two tasks. We observe that understanding and generation require fundamentally different granularities of visual information. This leads to a critical trade-off, particularly compromising performance in multimodal understanding tasks. TokenFlow addresses this challenge through an innovative dual-codebook architecture that decouples semantic and pixel-level feature learning while maintaining their alignment via a shared mapping mechanism. This design enables direct access to both high-level semantic representations crucial for understanding tasks and fine-grained visual features essential for generation through shared indices. Our extensive experiments demonstrate TokenFlow's superiority across multiple dimensions. Leveraging TokenFlow, we demonstrate for the first time that discrete visual input can surpass LLaVA-1.5 13B in understanding performance, achieving a 7.2\% average improvement. For image reconstruction, we achieve a strong FID score of 0.63 at 384*384 resolution. Moreover, TokenFlow establishes state-of-the-art performance in autoregressive image generation with a GenEval score of 0.55 at 256*256 resolution, achieving comparable results to SDXL.

cs.CV↗