arXiv subjects

Shiyu Liang

Publications and source records attributed to Shiyu Liang.

At least 19 recordsLinked to original sources

arXiv2026

Enhancing Multimodal RLVR with Pairwise Token Contrast and Adaptive Trajectory Confidence

Reinforcement learning with verifiable rewards (RLVR) improves multimodal reasoning, but group-relative policy optimization broadcasts a single sequence-level advantage to every token of a response, although errors usually localize to a few decisive steps. Recent self-distilled variants refine credit with a teacher--student likelihood gap. However, this signal cannot tell whether it relies on the image or on a language shortcut, and is applied uniformly even when the teacher is uninformative. To address this, we propose Pairwise token contrast and Adaptive trajectory ConfidencE (PACE). Pairwise token contrast scores each response under four teacher views, crossing a correct or incorrect answer hint with the image present or removed, and compares the two hints for every token, which cancels shifts shared by both hints and reveals whether the token's support relies on the image. Adaptive trajectory confidence measures how well the teacher predicts each response given the correct hint and the image, normalizes this score within the batch, and uses it to set how far the token contrast redistributes credit within that response. PACE achieves the best average accuracy on three backbones and ranks first on all evaluated multimodal reasoning benchmarks with Qwen3-VL-8B-Instruct. Experiments show that both components matter: replacing pairwise token contrast with a single-view signal underperforms our method, while adaptive trajectory confidence adds further improvement.

cs.CL↗

arXiv2026

Beyond Action Entropy: Quotient-Space Exploration for Genome-Scale Metabolic Model Repair

Repairing scientific models from functional observations differs fundamentally from supervised prediction: feedback may certify a solution without revealing which structural correction is responsible. We study this setting for genome-scale metabolic model (GEM) repair, where multiple reaction edits can explain the same phenotypes and many apparently distinct edits correspond to the same biological mechanism. This many-to-one structure creates a hidden failure mode for conventional exploration: diversity in the output space need not translate into diversity of scientific hypotheses. We introduce QuotientPO, which collapses equivalent repairs into canonical mechanisms and optimizes exploration directly over the resulting quotient space. To make quotient exploration informative under finite rollouts, we derive a kernelized Rényi estimator that resolves graded crowding among distinct repair cores beyond coarse exact-match counts. On 2,212 held-out GEMs, QuotientPO improves Success@32 from 17.93% to 20.10% (+12.1% relative) while consistently increasing distinct successful-core discovery under the same sampling budget. These results establish quotient-space exploration as a principled approach to mechanism-level discovery under verifier-induced equivalence.

cs.LG↗

arXiv2026

REACT: Physically and Chemically Consistent Reconstruction of Marine Active Tracers

Reconstructing global sea surface pH from sparse observations is critical for monitoring ocean acidification and understanding marine carbon cycling. Traditional assimilation and inverse models are physically grounded but costly for large-scale reconstruction. Recent black-box and physics-guided AI models improve efficiency, but are mainly designed for passive tracers, where the reconstructed variable is also the transported inventory. In contrast, pH is an active carbonate tracer: it is the prediction target, while dissolved inorganic carbon (DIC) is the conserved carbon inventory. This mismatch can produce low pH error while violating carbonate closure and source-free carbon conservation. To address this, we introduce \textbf{REACT}, a carbon-first reconstruction framework that decouples transport, active correction, and chemical decoding. REACT transports a latent carbonate state with a conservative advection--diffusion solver, captures non-conservative carbon-cycle variations with a source module, decodes the corrected state into pH, and constrains the output through carbonate equilibrium. This design keeps pH as the target while enforcing consistency on the underlying carbon state. On simulation data, REACT reduces pH NRMSE by (14.7%) and chemical consistency error by (24.0%) over the best baseline. Cross-temporal-scale evaluations show robustness against error accumulation from coarse to fine temporal scales, and ablation studies validate the effectiveness of each component.

cs.AI↗

arXiv2026

RLCracker: Evaluating the Worst-Case Vulnerability of LLM Watermarks with Adaptive RL Attacks

Large language model (LLM) watermarking has shown promise in detecting AI-generated content and mitigating misuse, with prior work claiming robustness against paraphrasing and text editing. In this paper, we argue that existing evaluations are not sufficiently adversarial, obscuring critical vulnerabilities and overstating the security. To address this, we introduce the adaptive robustness radius, a formal metric that quantifies the worst-case resilience of watermarks against adaptive adversaries. By lifting the paraphrase space into a KL-divergence ball, we approximate this radius and theoretically demonstrate that optimizing the attack context and model parameters can significantly reduce the approximate radius, making watermarks highly vulnerable to paraphrase attacks. Leveraging this insight, we propose RLCracker, a reinforcement learning (RL)-based adaptive attack that erases watermark signals with limited watermarked examples and limited access to the detector. Despite weak supervision, it empowers a 3B model to achieve 98.5% removal success with minimal semantic shift on 1,500-token Unigram-marked texts after training on only 100 short samples. This performance dramatically exceeds 6.75% by GPT-4o and generalizes across five model sizes over ten watermarking schemes. Our code is available at https://github.com/OTT0-OTO/RLCracker.

cs.CR↗

arXiv2026

GGBound: A Genome-Grounded Agent for Microbial Life-Boundary Prediction

Characterizing the physiological life boundaries of microbial strains, including viable temperature, pH, salinity, substrate utilization, and morphology, is central to biotechnology and ecology, yet traditionally requires exhaustive in vitro screening. Existing computational approaches either treat physiological traits as isolated supervised targets or repurpose biological foundation models as static encoders, leaving the genotype-to-physiology gap largely unbridged. We formulate microbial life-boundary prediction as a unified genome-to-physiology task and address it with a genome-conditioned, tool-augmented LLM agent. To support this task, we curate a strain-centric benchmark from IJSEM, NCBI, and BacDive covering 1,525 strains and 6,448 instances across viability intervals, environmental optima, substrate utilization, categorical traits, and morphology. Architecturally, the agent injects frozen LucaOne genome embeddings into a Qwen backbone via lightweight token fusion, and reasons over a similarity-based RAG module and a Genome-scale Metabolic Model (GEM) perturbation tool. We optimize the agent through a three-stage pipeline of gene-text alignment, agentic SFT on distilled trajectories, and GRPO with a novel counterfactual gene-grounding reward that reinforces the policy only when the authentic genome embedding causally improves correct-token generation relative to a zero-gene ablation. The resulting 4B-parameter agent matches or surpasses substantially larger frontier LLMs, with ablations confirming that genome-token fusion, dynamic tool use, and the counterfactual reward each yield distinct, significant gains.

cs.CY↗

arXiv2026

Reflection Anchors for Propagation-Aware Visual Retention in Long-Chain Multimodal Reasoning

Long chain-of-thought (CoT) reasoning improves large vision--language models, but visual information often fades during generation, limiting long-horizon multimodal reasoning. Existing methods either re-inject vision at inference or train policies for stronger grounding, but where to intervene relies on perception heuristics rather than principled gain analysis, and how local visual influence propagates remains implicit. We study this problem from an information-theoretic standpoint and derive a lower bound on the downstream visual gain of a one-step intervention, which suggests two factors: local branching room (token entropy) and downstream visual propagation potential (suffix divergence from a vision-marginalized reference). Guided by this analysis, we propose reflection-anchor policy optimization (RAPO), a GRPO-based policy optimization method that selects high-entropy reflection anchors and optimizes a chain-masked finite-window KL surrogate for downstream visual dependence. Experiments on reasoning-intensive and general-domain benchmarks show that RAPO delivers substantial gains over strong baselines across multiple LVLM backbones. Mechanism analyses further indicate that reflection anchors are enriched for visually sensitive decision points and that RAPO increases contrastive visual-dependence signals along generated trajectories.

cs.CV↗

arXiv2026

VCORE: Variance-Controlled Optimization-based Reweighting for Chain-of-Thought Supervision

Supervised fine-tuning (SFT) on long chain-of-thought (CoT) trajectories has emerged as a crucial technique for enhancing the reasoning abilities of large language models (LLMs). However, the standard cross-entropy loss treats all tokens equally, ignoring their heterogeneous contributions across a reasoning trajectory. This uniform treatment leads to misallocated supervision and weak generalization, especially in complex, long-form reasoning tasks. To address this, we introduce \textbf{V}ariance-\textbf{C}ontrolled \textbf{O}ptimization-based \textbf{RE}weighting (VCORE), a principled framework that reformulates CoT supervision as a constrained optimization problem. By adopting an optimization-theoretic perspective, VCORE enables a principled and adaptive allocation of supervision across tokens, thereby aligning the training objective more closely with the goal of robust reasoning generalization. Empirical evaluations demonstrate that VCORE achieves the strongest overall average performance, with especially clear gains on lower-capacity models. Across both in-domain and out-of-domain settings, VCORE achieves substantial performance gains on mathematical and coding benchmarks, using models from the Qwen3 series (4B, 8B, 32B) and LLaMA-3.1-8B-Instruct. Moreover, we show that VCORE serves as a more effective initialization for subsequent reinforcement learning, establishing a stronger foundation for advancing the reasoning capabilities of LLMs. The Code will be released at https://github.com/coder-gx/VCORE.

cs.CL↗

arXiv2026

RLSpoofer: A Lightweight Evaluator for LLM Watermark Spoofing Resilience

Large language model (LLM) watermarking has emerged as a promising approach for detecting and attributing AI-generated text, yet its robustness to black-box spoofing remains insufficiently evaluated. Existing evaluation methods often demand extensive datasets and white-box access to algorithmic internals, limiting their practical applicability. In this paper, we study watermark resilience against spoofing fundamentally from a distributional perspective. We first establish a \textit{local capacity bottleneck}, which theoretically characterizes the probability mass that can be reallocated under KL-bounded local updates while preserving semantic fidelity. Building on this, we propose RLSpoofer, a reinforcement learning-based black-box spoofing attack that requires only 100 human-watermarked paraphrase training pairs and zero access to the watermarking internals or detectors. Despite weak supervision, it empowers a 4B model to achieve a 62.0\% spoof success rate with minimal semantic shift on PF-marked texts, dwarfing the 6\% of baseline models trained on up to 10,000 samples. Our findings expose the fragile spoofing resistance of current LLM watermarking paradigms, providing a lightweight evaluation framework and stressing the urgent need for more robust schemes.

cs.CR↗

arXiv2026

RL-100: Performant Robotic Manipulation with Real-World Reinforcement Learning

Real-world robotic manipulation in homes and factories demands reliability, efficiency, and robustness that approach or surpass those of skilled human operators. We present RL-100, a real-world reinforcement learning framework built on diffusion visuomotor policies. RL-100 unifies imitation and reinforcement learning under a single clipped PPO surrogate objective applied within the denoising process, yielding conservative and stable improvements across offline and online stages. To meet deployment latency requirements, a lightweight consistency distillation method compresses multi-step diffusion into a one-step controller for high-frequency control. The framework is task-, embodiment-, and representation-agnostic, and supports both single-action and action-chunking control. We evaluate RL-100 on eight diverse real-robot tasks, from dynamic pushing and agile bowling to pouring, cloth folding, unscrewing, multi-stage juicing, and long-horizon box folding. RL-100 attains 100 percent success across evaluated trials, for a total of 1000 out of 1000 episodes, including up to 250 out of 250 consecutive trials on one task. It matches or surpasses expert teleoperators in time to completion. Without retraining, a single policy attains approximately 90 percent zero-shot success under environmental and dynamics shifts, adapts in a few-shot regime to significant task variations (86.7 percent), and remains robust to aggressive human perturbations (about 96 percent). Notably, our juicing robot served random customers continuously for about seven hours without failure when deployed zero-shot in a shopping mall. These results suggest a practical path to deployment-ready robot learning by starting from human priors, aligning training objectives with human-grounded metrics, and reliably extending performance beyond human demonstrations.

cs.RO↗

arXiv2025

A Middle Path for On-Premises LLM Deployment: Preserving Privacy Without Sacrificing Model Confidentiality

Privacy-sensitive users require deploying large language models (LLMs) within their own infrastructure (on-premises) to safeguard private data and enable customization. However, vulnerabilities in local environments can lead to unauthorized access and potential model theft. To address this, prior research on small models has explored securing only the output layer within hardware-secured devices to balance model confidentiality and customization. Yet this approach fails to protect LLMs effectively. In this paper, we discover that (1) query-based distillation attacks targeting the secured top layer can produce a functionally equivalent replica of the victim model; (2) securing the same number of layers, bottom layers before a transition layer provide stronger protection against distillation attacks than top layers, with comparable effects on customization performance; and (3) the number of secured layers creates a trade-off between protection and customization flexibility. Based on these insights, we propose SOLID, a novel deployment framework that secures a few bottom layers in a secure environment and introduces an efficient metric to optimize the trade-off by determining the ideal number of hidden layers. Extensive experiments on five models (1.3B to 70B parameters) demonstrate that SOLID outperforms baselines, achieving a better balance between protection and downstream customization.

cs.LG↗

arXiv2025

From Parameters to Prompts: Understanding and Mitigating the Factuality Gap between Fine-Tuned LLMs

Factual knowledge extraction aims to explicitly extract knowledge parameterized in pre-trained language models for application in downstream tasks. While prior work has been investigating the impact of supervised fine-tuning data on the factuality of large language models (LLMs), its mechanism remains poorly understood. We revisit this impact through systematic experiments, with a particular focus on the factuality gap that arises when fine-tuning on known versus unknown knowledge. Our findings show that this gap can be mitigated at the inference stage, either under out-of-distribution (OOD) settings or by using appropriate in-context learning (ICL) prompts (i.e., few-shot learning and Chain of Thought (CoT)). We prove this phenomenon theoretically from the perspective of knowledge graphs, showing that the test-time prompt may diminish or even overshadow the impact of fine-tuning data and play a dominant role in knowledge extraction. Ultimately, our results shed light on the interaction between finetuning data and test-time prompt, demonstrating that ICL can effectively compensate for shortcomings in fine-tuning data, and highlighting the need to reconsider the use of ICL prompting as a means to evaluate the effectiveness of fine-tuning data selection methods.

cs.CL↗

arXiv2025

Origin Tracer: A Method for Detecting LoRA Fine-Tuning Origins in LLMs

As large language models (LLMs) continue to advance, their deployment often involves fine-tuning to enhance performance on specific downstream tasks. However, this customization is sometimes accompanied by misleading claims about the origins, raising significant concerns about transparency and trust within the open-source community. Existing model verification techniques typically assess functional, representational, and weight similarities. However, these approaches often struggle against obfuscation techniques, such as permutations and scaling transformations. To address this limitation, we propose a novel detection method Origin-Tracer that rigorously determines whether a model has been fine-tuned from a specified base model. This method includes the ability to extract the LoRA rank utilized during the fine-tuning process, providing a more robust verification framework. This framework is the first to provide a formalized approach specifically aimed at pinpointing the sources of model fine-tuning. We empirically validated our method on thirty-one diverse open-source models under conditions that simulate real-world obfuscation scenarios. We empirically analyze the effectiveness of our framework and finally, discuss its limitations. The results demonstrate the effectiveness of our approach and indicate its potential to establish new benchmarks for model verification.

cs.AI↗

arXiv2024

Scientific and technological knowledge grows linearly over time

The past few centuries have witnessed a dramatic growth in scientific and technological knowledge. However, the nature of that growth - whether exponential or otherwise - remains controversial, perhaps partly due to the lack of quantitative characterizations. We evaluated knowledge as a collective thinking structure, using citation networks as a representation, by examining extensive datasets that include 213 million publications (1800-2020) and 7.6 million patents (1976-2020). We found that knowledge - which we conceptualize as the reduction of uncertainty in a knowledge network - grew linearly over time in naturally formed citation networks that themselves expanded exponentially. Moreover, our results revealed inflection points in the growth of knowledge that often corresponded to important developments within fields, such as major breakthroughs, new paradigms, or the emergence of entirely new areas of study. Around these inflection points, knowledge may grow rapidly or exponentially on a local scale, although the overall growth rate remains linear when viewed globally. Previous studies concluding an exponential growth of knowledge may have focused primarily on these local bursts of rapid growth around key developments, leading to the misconception of a global exponential trend. Our findings help to reconcile the discrepancy between the perceived exponential growth and the actual linear growth of knowledge by highlighting the distinction between local and global growth patterns. Overall, our findings reveal major science development trends for policymaking, showing that producing knowledge is far more challenging than producing papers.

physics.soc-ph↗

arXiv2024

Electron densities of Transition Region Loops derived from IRIS O IV spectral data

Loops are fundamental structures in the magnetized atmosphere of the sun. Their physical properties are crucial for understanding the nature of the solar atmosphere. Transition region loops are relatively dynamic and their physical properties have not yet been fully understood. With spectral data of the line pair of O IV 1399.8 Å& 1401.2 Å($T_{max}=1.4\times10^5$ K) of 23 transition region loops obtained by IRIS, we carry out the first systematic analyses to their loop lengths ($L$), electron densities ($n_e$) and effective temperatures. We found electron densities, loop lengths and effective temperatures of these loops are in the ranges of $8.9\times10^{9}$-$3.5\times10^{11}$ cm$^{-3}$, 8-30 Mm and $1.9\times10^5$-$1.3\times10^6$ K, respectively. At a significant level of 90\%, regression analyses show that the relationship between electron densities and loop lengths is $n_e[cm^{-3}]\varpropto (L[Mm])^{-0.78\pm0.42}$, while the dependences of electron densities on effective temperatures and that on the line intensities are not obvious. These observations demonstrate that transition region loops are significantly different than their coronal counterparts. Further studies on the theoretical aspect based on the physical parameters obtained here are of significance for understanding the nature of transition region loops.

astro-ph.SR↗

arXiv2024

AceMap: Knowledge Discovery through Academic Graph

The exponential growth of scientific literature requires effective management and extraction of valuable insights. While existing scientific search engines excel at delivering search results based on relational databases, they often neglect the analysis of collaborations between scientific entities and the evolution of ideas, as well as the in-depth analysis of content within scientific publications. The representation of heterogeneous graphs and the effective measurement, analysis, and mining of such graphs pose significant challenges. To address these challenges, we present AceMap, an academic system designed for knowledge discovery through academic graph. We present advanced database construction techniques to build the comprehensive AceMap database with large-scale academic entities that contain rich visual, textual, and numerical information. AceMap also employs innovative visualization, quantification, and analysis methods to explore associations and logical relationships among academic entities. AceMap introduces large-scale academic network visualization techniques centered on nebular graphs, providing a comprehensive view of academic networks from multiple perspectives. In addition, AceMap proposes a unified metric based on structural entropy to quantitatively measure the knowledge content of different academic entities. Moreover, AceMap provides advanced analysis capabilities, including tracing the evolution of academic ideas through citation relationships and concept co-occurrence, and generating concise summaries informed by this evolutionary process. In addition, AceMap uses machine reading methods to generate potential new ideas at the intersection of different fields. Exploring the integration of large language models and knowledge graphs is a promising direction for future research in idea evolution. Please visit \url{https://www.acemap.info} for further exploration.

cs.DL↗

arXiv2024

Temporal Generalization Estimation in Evolving Graphs

Graph Neural Networks (GNNs) are widely deployed in vast fields, but they often struggle to maintain accurate representations as graphs evolve. We theoretically establish a lower bound, proving that under mild conditions, representation distortion inevitably occurs over time. To estimate the temporal distortion without human annotation after deployment, one naive approach is to pre-train a recurrent model (e.g., RNN) before deployment and use this model afterwards, but the estimation is far from satisfactory. In this paper, we analyze the representation distortion from an information theory perspective, and attribute it primarily to inaccurate feature extraction during evolution. Consequently, we introduce Smart, a straightforward and effective baseline enhanced by an adaptive feature extractor through self-supervised graph reconstruction. In synthetic random graphs, we further refine the former lower bound to show the inevitable distortion over time and empirically observe that Smart achieves good estimation performance. Moreover, we observe that Smart consistently shows outstanding generalization estimation on four real-world evolving graphs. The ablation studies underscore the necessity of graph reconstruction. For example, on OGB-arXiv dataset, the estimation metric MAPE deteriorates from 2.19% to 8.00% without reconstruction.

cs.LG↗

arXiv2023

Graph Out-of-Distribution Generalization with Controllable Data Augmentation

Graph Neural Network (GNN) has demonstrated extraordinary performance in classifying graph properties. However, due to the selection bias of training and testing data (e.g., training on small graphs and testing on large graphs, or training on dense graphs and testing on sparse graphs), distribution deviation is widespread. More importantly, we often observe \emph{hybrid structure distribution shift} of both scale and density, despite of one-sided biased data partition. The spurious correlations over hybrid distribution deviation degrade the performance of previous GNN methods and show large instability among different datasets. To alleviate this problem, we propose \texttt{OOD-GMixup} to jointly manipulate the training distribution with \emph{controllable data augmentation} in metric space. Specifically, we first extract the graph rationales to eliminate the spurious correlations due to irrelevant information. Secondly, we generate virtual samples with perturbation on graph rationale representation domain to obtain potential OOD training samples. Finally, we propose OOD calibration to measure the distribution deviation of virtual samples by leveraging Extreme Value Theory, and further actively control the training distribution by emphasizing the impact of virtual OOD samples. Extensive studies on several real-world datasets on graph classification demonstrate the superiority of our proposed method over state-of-the-art baselines.

cs.LG↗

arXiv2021

Achieving Small Test Error in Mildly Overparameterized Neural Networks

Recent theoretical works on over-parameterized neural nets have focused on two aspects: optimization and generalization. Many existing works that study optimization and generalization together are based on neural tangent kernel and require a very large width. In this work, we are interested in the following question: for a binary classification problem with two-layer mildly over-parameterized ReLU network, can we find a point with small test error in polynomial time? We first show that the landscape of loss functions with explicit regularization has the following property: all local minima and certain other points which are only stationary in certain directions achieve small test error. We then prove that for convolutional neural nets, there is an algorithm which finds one of these points in polynomial time (in the input dimension and the number of data points). In addition, we prove that for a fully connected neural net, with an additional assumption on the data distribution, there is a polynomial time algorithm.

cs.LG↗