arXiv subjects

Jie Yu

Publications and source records attributed to Jie Yu.

At least 37 records · Page 2Linked to original sources

arXiv2026

Pretraining a Foundation Model for Small-Molecule Natural Products

Natural products, as metabolites from microorganisms, animals, or plants, exhibit diverse biological activities, making them crucial for drug discovery. Nowadays, existing deep learning methods for natural products research primarily rely on supervised learning approaches designed for specific downstream tasks. However, such one-model-for-a-task paradigm often lacks generalizability and leaves significant room for performance improvement. Additionally, existing molecular characterization methods are not well-suited for the unique tasks associated with natural products. To address these limitations, we have pre-trained a foundation model for natural products based on their unique properties. Our approach employs a novel pretraining strategy that is especially tailored to natural products. By incorporating contrastive learning and masked graph learning objectives, we emphasize evolutional information from molecular scaffolds while capturing side-chain information. Our framework achieves state-of-the-art (SOTA) results in various downstream tasks related to natural product mining and drug discovery. We first compare taxonomy classification with synthesized molecule-focused baselines to demonstrate that current models are inadequate for understanding natural synthesis. Furthermore, by diving into a fine-grained analysis at both the gene and microbial levels, NaFM demonstrates the ability to capture evolutionary information. Eventually, our method is experimented with virtual screening, illustrating informative natural product representations that can lead to more effective identification of potential drug candidates.

q-bio.QM↗

arXiv2026

More Than Can Be Said: A Benchmark and Framework for Pre-Question Scientific Ideation

AI research agents have shown strong potential in automating literature search and manuscript refinement, yet most assume a clear and actionable initial input, operating only after a research question has been made explicit. In contrast, human research often begins with tacit friction, a sense of misalignment before a question can be formed. We introduce InciteResearch, a multi-agent framework designed to make a researcher's implicit understanding explicit, inspectable, and actionable. InciteResearch decomposes the logical chain of Socratic questioning and distributes it across the entire pipeline that: (1) Elicits a structured five-dimensional researcher profile state anchored by specific friction points from vague, even domain-unrelated inputs; (2) Violates hidden assumptions by maximizing the feasibility-novelty product with enforcing a 7-stage causal derivation trace; and (3) check whether the proposed method is a Necessary consequence of the reframed insight. We further introduce TF-Bench, the first benchmark for tacit-to-explicit research assistance that distinguishes domain-related from domain-unrelated inspirations across four scientific modes. On TF-Bench, InciteResearch achieves leapfrogging gains over a prompt-based baseline (novelty/impact from 3.671/3.806 to 4.250/4.397), shifting generated proposals from recombination to architectural insight. Our work demonstrates that AI can serve as an extension of thinking itself, rather than merely automating downstream execution.

cs.AI↗

arXiv2026

Nonvolatile single-ion memory with picosecond switching

The rapid development of artificial intelligence (AI), Internet of Things (IoT), and edge computing applications has posed severe challenges to conventional memory technologies in terms of density, speed, and energy consumption. Herein, a single-ion transport mechanism is proposed to achieve picosecond (ps) switching capability. For monolayer hexagonal boron nitride (h-BN) with single-atom vacancy defects, first-principles calculations reveal that single-ion penetration across the BN plane dominates the resistive switching. The trapping and release of a single ion correspond to different states of the memory device for one bit of information. Experimentally fabricated single-ion memory exhibits nonvolatile resistive switching with ultra-fast switching speed of 20 ps and ultra-low energy consumption of 310 aJ/bit. This high performance is attributed to the extremely short distance for the single ion to travel through. Such devices pave the way for the realization of high-performance nonvolatile memory with ultra-fast speed, ultra-low energy consumption, and high storage density, that is called the "Unified Memory" long desired by the whole industry.

physics.app-ph↗

arXiv2026

AkiraRust: Re-thinking LLM-aided Rust Repair Using a Feedback-guided Thinking Switch

Eliminating undefined behaviors (UBs) in Rust programs requires a deep semantic understanding to enable accurate and reliable repair. While existing studies have demonstrated the potential of LLMs to support Rust code analysis and repair, most frameworks remain constrained by inflexible templates or lack grounding in executable semantics, resulting in limited contextual awareness and semantic incorrectness. Here, we present AkiraRust, an LLM-driven repair and verification framework that incorporates a finite-state machine to dynamically adapt its detection and repair flow to runtime semantic conditions. AkiraRust introduces a dual-mode reasoning strategy that coordinates fast and slow thinking across multiple agents. Each agent is mapped to an FSM state, and a waveform-driven transition controller manages state switching, rollback decisions, and semantic check pointing, enabling context-aware and runtime-adaptive repair. Experimental results show that AkiraRust achieves about 92% semantic correctness and delivers a 2.2x average speedup compared to SOTA.

cs.SE↗

arXiv2026

Cognitive Chunking for Soft Prompts: Accelerating Compressor Learning via Block-wise Causal Masking

Providing extensive context via prompting is vital for leveraging the capabilities of Large Language Models (LLMs). However, lengthy contexts significantly increase inference latency, as the computational cost of self-attention grows quadratically with sequence length. To mitigate this issue, context compression-particularly soft prompt compressio-has emerged as a widely studied solution, which converts long contexts into shorter memory embeddings via a trained compressor. Existing methods typically compress the entire context indiscriminately into a set of memory tokens, requiring the compressor to capture global dependencies and necessitating extensive pre-training data to learn effective patterns. Inspired by the chunking mechanism in human working memory and empirical observations of the spatial specialization of memory embeddings relative to original tokens, we propose Parallelized Iterative Compression (PIC). By simply modifying the Transformer's attention mask, PIC explicitly restricts the receptive field of memory tokens to sequential local chunks, thereby lowering the difficulty of compressor training. Experiments across multiple downstream tasks demonstrate that PIC consistently outperforms competitive baselines, with superiority being particularly pronounced in high compression scenarios (e.g., achieving relative improvements of 29.8\% in F1 score and 40.7\% in EM score on QA tasks at the $64\times$ compression ratio). Furthermore, PIC significantly expedites the training process. Specifically, when training the 16$\times$ compressor, it surpasses the peak performance of the competitive baseline while effectively reducing the training time by approximately 40\%.

cs.AI↗

arXiv2026

EMSEdit: Efficient Multi-Step Meta-Learning-based Model Editing

Large Language Models (LLMs) power numerous AI applications, yet updating their knowledge remains costly. Model editing provides a lightweight alternative through targeted parameter modifications, with meta-learning-based model editing (MLME) demonstrating strong effectiveness and efficiency. However, we find that MLME struggles in low-data regimes and incurs high training costs due to the use of KL divergence. To address these issues, we propose $\textbf{E}$fficient $\textbf{M}$ulti-$\textbf{S}$tep $\textbf{Edit (EMSEdit)}$, which leverages multi-step backpropagation (MSBP) to effectively capture gradient-activation mapping patterns within editing samples, performs multi-step edits per sample to enhance editing performance under limited data, and introduces norm-based regularization to preserve unedited knowledge while improving training efficiency. Experiments on two datasets and three LLMs show that EMSEdit consistently outperforms state-of-the-art methods in both sequential and batch editing. Moreover, MSBP can be seamlessly integrated into existing approaches to yield additional performance gains. Further experiments on a multi-hop reasoning editing task demonstrate EMSEdit's robustness in handling complex edits, while ablation studies validate the contribution of each design component. Our code is available at https://github.com/xpq-tech/emsedit.

cs.CL↗

arXiv2026

RV$\times$TESS I: Modeling Asteroseismic Signals with Simultaneous Photometry and RVs

Detecting small planets via the radial velocity method remains challenged by signals induced by stellar variability, versus the effects of the planet(s). Here, we explore using Gaussian Process (GP) regression with Transiting Exoplanet Survey Satellite (TESS) photometry in modeling radial velocities (RVs) to help to mitigate stellar jitter from oscillations and granulation for exoplanet detection. We applied GP regression to simultaneous TESS photometric and RV data of HD 5562, a G-type subgiant ($M_\star=1.09M_{\odot}$, $R_\star=1.88R_{\odot}$) with a V magnitude of 7.17, using photometry to inform the priors for RV fitting. The RV data is obtained by the Magellan Planet Finder Spectrograph (PFS). The photometry-informed GP regression reduced the RV scatter of HD~5562 from 2.03 to 0.51 m/s. We performed injection and recovery tests to evaluate the potential of GPs for discovering small exoplanets around evolved stars, which demonstrate that the GP provides comparable noise reduction to the binning method. We also found that the necessity of photometric data depends on the quality of the RV dataset. For long baseline and high-cadence RV observations, GP regression can effectively mitigate stellar jitter without photometric data. However, for intermittent RV observations, incorporating photometric data improves GP fitting and enhances detection capabilities.

astro-ph.EP↗

arXiv2026

Rethinking Residual Distribution in Locate-then-Edit Model Editing

Model editing enables targeted updates to the knowledge of large language models (LLMs) with minimal retraining. Among existing approaches, locate-then-edit methods constitute a prominent paradigm: they first identify critical layers, then compute residuals at the final critical layer based on the target edit, and finally apply least-squares-based multi-layer updates via $\textbf{residual distribution}$. While empirically effective, we identify a counterintuitive failure mode: residual distribution, a core mechanism in these methods, introduces weight shift errors that undermine editing precision. Through theoretical and empirical analysis, we show that such errors increase with the distribution distance, batch size, and edit sequence length, ultimately leading to inaccurate or suboptimal edits. To address this, we propose the $\textbf{B}$oundary $\textbf{L}$ayer $\textbf{U}$pdat$\textbf{E (BLUE)}$ strategy to enhance locate-then-edit methods. Sequential batch editing experiments on three LLMs and two datasets demonstrate that BLUE not only delivers an average performance improvement of 35.59\%, significantly advancing the state of the art in model editing, but also enhances the preservation of LLMs' general capabilities. Our code is available at https://github.com/xpq-tech/BLUE.

cs.CL↗

arXiv2026

EMP: Enhance Memory in Data Pruning

Recently, large language and vision models have shown strong performance, but due to high pre-training and fine-tuning costs, research has shifted towards faster training via dataset pruning. Previous methods used sample loss as an evaluation criterion, aiming to select the most "difficult" samples for training. However, when the pruning rate increases, the number of times each sample is trained becomes more evenly distributed, which causes many critical or general samples to not be effectively fitted. We refer to this as Low-Frequency Learning (LFL). In other words, LFL prevents the model from remembering most samples. In our work, we decompose the scoring function of LFL, provide a theoretical explanation for the inefficiency of LFL, and propose adding a memory term to the scoring function to enhance the model's memory capability, along with an approximation of this memory term. Similarly, we explore memory in Self-Supervised Learning (SSL), marking the first discussion on SSL memory. Using contrastive learning, we derive the memory term both theoretically and experimentally. Finally, we propose Enhance Memory Pruning (EMP), which addresses the issue of insufficient memory under high pruning rates by enhancing the model's memory of data, thereby improving its performance. We evaluated the performance of EMP in tasks such as image classification, natural language understanding, and model pre-training. The results show that EMP can improve model performance under extreme pruning rates. For example, in the CIFAR100-ResNet50 pre-training task, with 70\% pruning, EMP outperforms current methods by 2.2\%.

cs.LG↗

arXiv2026

Seeing Right but Saying Wrong: Inter- and Intra-Layer Refinement in MLLMs without Training

Multimodal Large Language Models (MLLMs) have demonstrated strong capabilities across a variety of vision-language tasks. However, their internal reasoning often exhibits a critical inconsistency: although deeper layers may attend to the correct visual regions, final predictions are frequently misled by noisy attention from earlier layers. This results in a disconnect between what the model internally understands and what it ultimately expresses, a phenomenon we describe as seeing it right but saying it wrong. To address this issue, we propose DualPD, a dual-perspective decoding refinement strategy that enhances the visual understanding without any additional training. DualPD consists of two components. (1) The layer-wise attention-guided contrastive logits module captures how the belief in the correct answer evolves by comparing output logits between layers that exhibit the largest attention shift. (2) The head-wise information filtering module suppresses low-contribution attention heads that focus on irrelevant regions, thereby improving attention quality within each layer. Experiments conducted on both the LLaVA and Qwen-VL model families across multiple multimodal benchmarks demonstrate that DualPD consistently improves accuracy without training, confirming its effectiveness and generalizability. The code will be released upon publication.

cs.CV↗

arXiv2026

JPU: Bridging Jailbreak Defense and Unlearning via On-Policy Path Rectification

Despite extensive safety alignment, Large Language Models (LLMs) often fail against jailbreak attacks. While machine unlearning has emerged as a promising defense by erasing specific harmful parameters, current methods remain vulnerable to diverse jailbreaks. We first conduct an empirical study and discover that this failure mechanism is caused by jailbreaks primarily activating non-erased parameters in the intermediate layers. Further, by probing the underlying mechanism through which these circumvented parameters reassemble into the prohibited output, we verify the persistent existence of dynamic $\textbf{jailbreak paths}$ and show that the inability to rectify them constitutes the fundamental gap in existing unlearning defenses. To bridge this gap, we propose $\textbf{J}$ailbreak $\textbf{P}$ath $\textbf{U}$nlearning (JPU), which is the first to rectify dynamic jailbreak paths towards safety anchors by dynamically mining on-policy adversarial samples to expose vulnerabilities and identify jailbreak paths. Extensive experiments demonstrate that JPU significantly enhances jailbreak resistance against dynamic attacks while preserving the model's utility.

cs.CR↗

arXiv2025

LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Training

We present LLaVA-OneVision-1.5, a novel family of Large Multimodal Models (LMMs) that achieve state-of-the-art performance with significantly reduced computational and financial costs. Different from the existing works, LLaVA-OneVision-1.5 provides an open, efficient, and reproducible framework for building high-quality vision-language models entirely from scratch. The LLaVA-OneVision-1.5 release comprises three primary components: (1) Large-Scale Curated Datasets: We construct an 85M concept-balanced pretraining dataset LLaVA-OneVision-1.5-Mid-Traning and a meticulously curated 22M instruction dataset LLaVA-OneVision-1.5-Instruct. (2) Efficient Training Framework: We develop a complete end-to-end efficient training framework leveraging an offline parallel data packing strategy to facilitate the training of LLaVA-OneVision-1.5 within a $16,000 budget. (3) State-of-the-art Performance: Experimental results demonstrate that LLaVA-OneVision-1.5 yields exceptionally competitive performance across a broad range of downstream tasks. Specifically, LLaVA-OneVision-1.5-8B outperforms Qwen2.5-VL-7B on 18 of 27 benchmarks, and LLaVA-OneVision-1.5-4B surpasses Qwen2.5-VL-3B on all 27 benchmarks. (4) RL-based Post-training: We unlock the model's latent potential through a lightweight RL stage, effectively eliciting robust chain-of-thought reasoning to significantly boost performance on complex multimodal reasoning tasks.

cs.CV↗

arXiv2025

Automatic nodule identification and differentiation in ultrasound videos to facilitate per-nodule examination

Ultrasound is a vital diagnostic technique in health screening, with the advantages of non-invasive, cost-effective, and radiation free, and therefore is widely applied in the diagnosis of nodules. However, it relies heavily on the expertise and clinical experience of the sonographer. In ultrasound images, a single nodule might present heterogeneous appearances in different cross-sectional views which makes it hard to perform per-nodule examination. Sonographers usually discriminate different nodules by examining the nodule features and the surrounding structures like gland and duct, which is cumbersome and time-consuming. To address this problem, we collected hundreds of breast ultrasound videos and built a nodule reidentification system that consists of two parts: an extractor based on the deep learning model that can extract feature vectors from the input video clips and a real-time clustering algorithm that automatically groups feature vectors by nodules. The system obtains satisfactory results and exhibits the capability to differentiate ultrasound videos. As far as we know, it's the first attempt to apply re-identification technique in the ultrasonic field.

eess.IV↗

arXiv2025

GaiaUnlimited: The old stellar disc of the Milky Way as traced by the Red Clump

We present an exploration of the Milky Way's structural parameters using an all-sky sample of RC giants to map the stellar density from the Galactic disc beyond 3 kpc. These evolved giants are considered to be standard candles due to their low intrinsic variance in their absolute luminosities, allowing us to estimate their distances with reasonable confidence. We exploit all-sky photometry from the AllWISE mid-infrared survey and the Gaia survey, along with astrometry from Gaia Data Release 3 and recent 3D extinction maps, to develop a probabilistic scheme in order to select with high confidence RC-like stars. Our curated catalogue contains about 10 million sources, for which we estimate photometric distances based on the WISE W1 photometry. We then derive the selection function for our sample, which is the combined selection function of sources with both Gaia and AllWISE photometry. Using the distances and accounting for the full selection function of our observables, we are able to fit a two-disc, multi-parameter model to constrain the scale height (hz), scale-length (rd), flaring, and the relative mass ratios of the two disc components. We illustrate and verify our methodology using mock catalogues of RC stars. We find that the RC population is best described by a flared disc with scale length rd=$4.24\pm0.32$ kpc and scale height at the Sun of hz(at Sun)=$0.18\pm0.01$ kpc, and a shorter and thicker disc with rd=$2.66\pm0.11$ kpc, hz(at Sun)=$0.48\pm0.11$ kpc, with no flare. The thicker disc constitutes 66\% of the RC stellar mass beyond 3 kpc, while the flared disc shows evidence of being warped beyond 9 kpc from the Galactic center. The residuals between the predicted number density of RC stars from our axisymmetric model and the measured counts show possible evidence of a two-armed spiral perturbation in the disc of the Milky Way.

astro-ph.GA↗

arXiv2025

Stand on The Shoulders of Giants: Building JailExpert from Previous Attack Experience

Large language models (LLMs) generate human-aligned content under certain safety constraints. However, the current known technique ``jailbreak prompt'' can circumvent safety-aligned measures and induce LLMs to output malicious content. Research on Jailbreaking can help identify vulnerabilities in LLMs and guide the development of robust security frameworks. To circumvent the issue of attack templates becoming obsolete as models evolve, existing methods adopt iterative mutation and dynamic optimization to facilitate more automated jailbreak attacks. However, these methods face two challenges: inefficiency and repetitive optimization, as they overlook the value of past attack experiences. To better integrate past attack experiences to assist current jailbreak attempts, we propose the \textbf{JailExpert}, an automated jailbreak framework, which is the first to achieve a formal representation of experience structure, group experiences based on semantic drift, and support the dynamic updating of the experience pool. Extensive experiments demonstrate that JailExpert significantly improves both attack effectiveness and efficiency. Compared to the current state-of-the-art black-box jailbreak methods, JailExpert achieves an average increase of 17\% in attack success rate and 2.7 times improvement in attack efficiency. Our implementation is available at \href{https://github.com/xiZAIzai/JailExpert}{XiZaiZai/JailExpert}

cs.CR↗

arXiv2025

The highest mass Kepler red giants -- II. Spectroscopic parameters, the amplitude-activity relation, and unexpected halo orbits

The high-mass (M$>$2 \Msolar{}) Kepler red giant stars are less well-studied than their lower-mass counterparts. In the previous article, we presented a sample of 48 high-mass Kepler red giants and measured their asteroseismic parameters. This article presents spectroscopic measurements from the same sample, using high-resolution Keck/HIRES spectra to determine \Teff{}, [Fe/H], \logg{}, and $v \sin i$. We refined our previous estimates of the stellar masses and radii based on the new \Teff{}. We also examined spectral features that could indicate binary activity, such as the Li line and [C/N] ratios. We found no Li-rich stars or clear [C/N] anomalies, but we observed a correlation between [C/N] and [Fe/H]. We measured chromospheric activity using the $S$-index of the Ca II H \& K lines and found no correlation with internal magnetic fields. However, we confirmed an anti-correlation between surface chromospheric activity and radial mode oscillation amplitudes, which indicates that strong surface magnetic fields weaken stellar oscillations. Finally, we used the Gaia DR3 astrometric data to show that our sample of stars have orbits consistent with all three Galactic kinematic regions. Although these stars are quite young, their orbits carry them into the thick disk and even the halo, raising questions about the accuracy and viability of kinematics in unravelling Galactic history. In future work, we plan to use the spectroscopic parameters measured here to provide better constraints for boutique frequency modelling, which will allow us to test the asteroseismic scaling relations at the high-mass regime.

astro-ph.SR↗

arXiv2025

Camel: Energy-Aware LLM Inference on Resource-Constrained Devices

Most Large Language Models (LLMs) are currently deployed in the cloud, with users relying on internet connectivity for access. However, this paradigm faces challenges such as network latency, privacy concerns, and bandwidth limits. Thus, deploying LLMs on edge devices has become an important research focus. In edge inference, request latency is critical as high latency can impair real-time tasks. At the same time, edge devices usually have limited battery capacity, making energy consumption another major concern. Balancing energy consumption and inference latency is essential. To address this, we propose an LLM inference energy management framework that optimizes GPU frequency and batch size to balance latency and energy consumption. By effectively managing the exploration-exploitation dilemma in configuration search, the framework finds the optimal settings. The framework was implemented on the NVIDIA Jetson AGX Orin platform, and a series of experimental validations were conducted. Results demonstrate that, compared to the default configuration, our framework reduces energy delay product (EDP) by 12.4%-29.9%, achieving a better balance between energy consumption and latency.

cs.NI↗

arXiv2025

Precise Asteroseismic Ages for the Helmi Streams

The Helmi streams are remnants of a dwarf galaxy that was accreted by the Milky Way and whose stars now form a distinct kinematic and chemical substructure in the Galactic halo. Precisely age-dating these typically faint stars of extragalactic origin has been notoriously difficult due to the limitations of using only spectroscopic data, interferometry, or coarse asteroseismic measurements. Using observations from NASA's Transiting Exoplanet Survey Satellite, we report the detailed asteroseismic modeling of two of the brightest red giants within the Helmi streams, HD 175305 and HD 128279. By modeling the individual oscillation mode frequencies and the spectroscopic properties of both stars, we determine their fundamental properties including mass, radius, and age ($τ$). We report $τ= 11.16 \pm 0.91$ Gyr for HD 175305 and $τ= 12.52 \pm 1.05$ Gyr for HD 128279, consistent with previously inferred star-formation histories for the Helmi streams and the differential chemical abundances between the two stars. With precise ages for individual stream members, our results reinforce the hypothesis that the Helmi streams' progenitor must have existed at least 12 Gyr ago. Our results also highlight that the ages of metal-poor, $α$-enhanced red giants can be severely underestimated when inferred using global asteroseismic parameters instead of individual mode frequencies.

astro-ph.SR↗