Search arXiv⌕ Search

arXiv subjects

Yang Liu

Publications and source records attributed to Yang Liu.

At least 235 records · Page 13Linked to original sources
arXiv2026

ScalePRM: Training Process Reward Models by Scaling Verification Compute Without Ground Truth

Training process reward models (PRMs) requires step-level correctness labels, obtained either through expensive human annotation or by relying on ground-truth answers, limiting the ability to scale process-level supervision. We propose ScalePRM, which scales verification compute as an alternative: given a problem and a candidate solution, we generate multiple independent verifications of each reasoning step and aggregate their judgments to produce synthetic step-level labels without ground truth. We explore two representative inference-time scaling strategies, parallel scaling through self-consistency and sequential scaling through meta-critique, and train generative PRMs on the resulting synthetic data. On ProcessBench, a benchmark for identifying erroneous steps in mathematical reasoning, PRMs trained on step-level self-consistency data achieve 67.5 F1, surpassing reference-guided training with ground-truth access (66.4 F1) and GPT-4o as a critic (61.9 F1). When deployed as reward signals in RL training with Qwen2.5-Math-7B, our best PRM achieves 47.4% average accuracy across six mathematical reasoning benchmarks, outperforming ground-truth-based RLVR (43.9%). We also identify and address reward exploitation patterns unique to generative PRM-based RL. Our results demonstrate that scaling verification compute is a viable alternative to ground-truth supervision for training process reward models.

cs.LG↗
arXiv2026

LiST: Local-Simplex Test-Time LoRA Fusion

Task-specific LoRA adapters offer a modular way to specialize large language and vision-language models. However, existing adapter composition methods are mostly static and cannot adapt to individual test inputs. To address these issues, we propose \textbf{LiST}, a label-free test-time LoRA fusion framework that converts an existing LoRA bank into a target-conditioned local simplex and searches sample-specific fusion weights at inference time. LiST builds joint task representations from LoRA parameter anchors and prompt-level behavior vectors, retrieves neighboring adapters as a local search space, and performs branch-preserving fusion without updating the backbone or adapters. Candidate weights are selected by a prompt-level energy with prior, geometric, and stochastic-consistency constraints, and are deployed only when they pass a safe acceptance rule. Otherwise, LiST falls back to a target-conditioned prior. Experiments on multimodal and language benchmarks show that LiST outperforms static LoRA merging and conventional test-time adaptation baselines, while preserving task-specific adapter utility and improving robustness on unseen tasks.

cs.CV↗
arXiv2026

Hindsight Memory-PRM: Supervising Memory Management with Auditable Hindsight Credit

Memory operations of long-horizon LLM agents are hard to supervise: an operation's value is unobservable when it is taken. But they are special -- they leave machine-readable evidence in the trajectory: retrieval hits and answer-time citations. Hindsight Memory-PRM exploits this audit trail twice: offline to train an operation-conditioned memory-utility critic, and online, where retrievals, citations, and one controlled deletion-and-reanswer per probe settle an intervention-calibrated entry-level presence credit, propagated along version chains as an action-level proxy reward -- no per-operation human labels, no Monte-Carlo replay of continuations. On held-out LoCoMo a local 8B policy reaches 77.5% under a fixed shared reader, surpassing its API teacher (65.1%) and all reproduced external systems, at one eighth the context of Mem0's official operating point; on LongMemEval, 79.0%. Ablations attribute the gain to causal calibration rather than signal density, and the policy converges to a multi-version memory organization whose gains no tested open-loop baseline reproduces.

cs.CL↗
arXiv2026

Matrix-Game 3.5: Enhancing Real-Time Streaming Interactive World Models with Patch Memory

Interactive world models extend video generation from offline clip synthesis toward persistent simulation of interactive virtual worlds, enabling applications in games, robotics, embodied agents, and XR. Achieving stable long-horizon interactive generation, however, remains challenging, as the model must simultaneously preserve scene geometry, dynamic consistency, and camera control while supporting real-time autoregressive generation. Building upon Matrix-Game 3.0, we present Matrix-Game 3.5, as shown in Figure 1, which advances real-time interactive world generation toward geometry-aware and long-horizon consistent simulation through three key improvements. First, we propose a unified geometry-aware memory framework, whose patch-memory and tiled-PRoPE components introduce no additional learnable parameters, combining explicit 3D patch retrieval with projective camera conditioning to enable geometry-consistent camera control and faithful long-horizon scene recall. Second, we introduce a static-dynamic disentangled world representation that separately models static scene geometry and dynamic subjects, preserving both geometric consistency and subject identity throughout long-horizon generation. Third, we develop a two-stage progressive real-time distillation framework that converts a bidirectional diffusion model into a few-step causal generator through Perceptual Flow Matching and curriculum based Self-Rollout DMD, enabling minute-long real-time interactive generation. Extensive experiments demonstrate that, with a unified training corpus spanning Unreal simulation environments, open-world games, and internet videos, MatrixGame 3.5 achieves strong performance in long-horizon scene recall, precise camera control, subject consistency, prompt-driven world generation, and stable real-time open-world interaction.

cs.CV↗
arXiv2026

Auditing and Mitigating Privacy Leakage in Cloud-Edge Collaborative Decoding

Applications such as personalized assistance and proprietary document analysis require large language models (LLMs) to generate outputs from private data. Yet powerful LLMs typically cannot be deployed on the resource-constrained devices where private data resides, and uploading private data to cloud-hosted LLMs exposes sensitive information. Recent work addresses this tension with a cloud-edge collaborative decoding paradigm, where private data are kept on the edge with a small language model (SLM) producing next-token distributions, which are fused with predictions from a cloud LLM operating solely on public data. In this paper, we systematically analyze the privacy risks of such a paradigm with a novel evaluation framework using constructed QA datasets, which show that such collaboration can expose substantial private-context information. To address such privacy leakage, we propose CoVeil, a defense mechanism which dynamically optimizes transmitted signals to suppress leakage during decoding time while preserving the collaborative quality. Extensive evaluations demonstrate that CoVeil consistently improves the privacy-utility trade-off over existing baselines by reducing data leakage by up to 87.2%, with minimal accuracy loss.

cs.CR↗
arXiv2026

Learning Simple Test-Time Environments for LLM Web Agents

Large language model (LLM) agents have demonstrated remarkable proficiency in manually constructed environments, yet their performance frequently collapses when transitioned to complex real-world settings. Existing research largely attribute this degradation to the compositional generalization gaps in LLMs on combinations of multiple simple, well-structured environments. In this work, we propose that LLM web agents can learn simple environment observations at test time. Specifically, we introduce trial steps for agents to decompose a complex environment observation into sub-modules, and implement a label-free learning method, Test-Time Environment Decomposition (TTED), to adapt agent behaviors with experience during inference. Our empirical evaluations demonstrate the framework's efficacy across both synthetic and realistic benchmarks, showing (1) experience gains acquired within simpler sub-environments can be effectively composed to improve performance in the full one, and (2) test-time training on sub-environments can significantly enhance the compositional generalization of agents in real-world web automation tasks. We also provide key insights in the design of the label-free learning algorithm. As more complex environments are accessed by LLM agents, we believe learning environment decomposition skills at test time will be critical for robust real-world deployment.

cs.CL↗
arXiv2026

A Survey of LLM-based Automated Program Repair: Taxonomies, Design Paradigms, and Applications

Large language models (LLMs) are reshaping automated program repair. We present a reproducible hierarchical taxonomy that organizes 66 LLM-based repair systems according to where repair capability and control logic principally reside: task-adapted parameters, prompt and context design, designer-specified workflows, or LLM-directed runtime control. Adaptation, generation pattern, runtime control, and auxiliary evidence are preserved as separate coded dimensions. This representation exposes control distinctions hidden by utilization labels and supports cross-paradigm analysis of system design and evaluation evidence. To the best of our knowledge, it is the first publicly available LLM-based software repair survey to operationalize Fine-Tuning, Prompting, Procedural, and Agentic as one corpus-wide primary classification. Our hierarchy complements prior surveys through one corpus-wide primary decision rule, while the result-level protocol audit bounds which reported results are defensibly comparable. To analyze how repair systems use benchmarks, we record benchmark variants, metrics, base models, and fault-localization assumptions for each system's primary reported result, identifying protocol-aligned comparison windows and benchmark-family fragments. We clarify paradigm trade-offs in task alignment, deployment cost, controllability, and multi-hunk or cross-file repair. We outline open challenges and research directions. Our artifacts and scripted survey pipeline are publicly available at https://github.com/GLEAM-Lab/ProgramRepair.

cs.SE↗
arXiv2026

RegimeFormer: A Large Protein Model of Global Perturbation Regimes

Protein language models organize sequence and structure at scale, but a global representation of how proteins respond to mutation remains lacking. We present RegimeFormer, a large protein perturbation model coupled to RegimeAtlas, constructed by harmonizing and indexing 202,556,313 non-redundant protein sequences across the tree of life. A diversity-preserving one-million-protein subset provides the high-resolution training and inference layer, with 995,995 proteins yielding residue-level summaries across 407,048,356 residues and substitution-specific predictions available on demand. Across experimental deep mutational scanning, molecular benchmarks, structural confidence and evolutionary constraint, RegimeFormer identifies reproducible protein-level perturbation regimes that organize residue fragility, adaptability and predictive uncertainty. Regime conditioning improves substitution-specific prediction, with the largest relative gains under unseen-protein, unseen-family and low-homology evaluation. RegimeFormer-derived molecular priors further improve downstream transcriptomic and drug-response modelling. Together, RegimeFormer and RegimeAtlas provide a scalable framework for mapping, predicting and querying protein perturbation landscapes across global sequence space.

q-bio.QM↗
arXiv2026

Riemann-1.0: An Embodied World Action Model for Physical AI

We introduce Riemann-1.0, a fully causal autoregressive World Action Model for embodied intelligence. Riemann-1.0 jointly models multi-view visual observations, robot states, and embodiment-specific actions within a unified causal autoregressive sequence, representing robot actions and world evolution as causal state transitions. Unlike existing WAMs based on joint generation, video-first prediction, or decoupled modeling paradigms, Riemann-1.0 unifies online robot policy execution and action-conditioned world simulation within a single model, enabling it to function as both an executable robot policy and a multi-embodiment visual world simulator. To scale embodied experience across heterogeneous data sources, we further develop a progressive embodied pretraining framework that unifies learning from egocentric human videos, handheld-gripper demonstrations, and heterogeneous robot trajectories under a shared World Action Modeling objective. Built upon 200K+ hours of interaction data, Riemann-1.0 progressively transfers large-scale embodied experience into executable robot manipulation capabilities. Riemann-1.0 achieves state-of-the-art performance across both simulation benchmarks and real-world manipulation tasks. It achieves success rates of 94.3% on RoboTwin2.0, 99.0% on LIBERO, and 62.6% on the long-horizon compositional benchmark RoboCasa-365, outperforming the previous best method by 8.4% On long-horizon real-world manipulation tasks, Riemann-1.0 achieves a Success Rate (SR) of 85.0% and a Progress Success Rate (PSR) of 94.4%, exceeding the strongest open-source baseline by 15% in SR. These results demonstrate that unified World Action Modeling together with progressive embodied pretraining effectively transforms large-scale embodied experience into generalizable robot manipulation capabilities.

cs.RO↗
arXiv2026

LAC: Linear and Angular Compliance for Humanoid Whole-body Control

Real-world humanoid tasks involve physical interaction with objects and humans, yet current controllers either reject external forces as disturbances or restrict compliance to limited body links while ignoring angular effects. We present LAC, a general whole-body controller that simultaneously realizes commanded Linear and Angular Compliance for wrenches applied to the upper body. First, we synthesize whole-body compliant responses into a large-scale augmented dataset. Sampled force and couple events are imposed on contact frames extracted from human interaction data. At each contact link, the external force and a virtual torque from the passively yielding kinematic chain drive a virtual admittance under the commanded stiffness. Subsequently, teacher-student reinforcement learning trains a single policy to track the compliant motions under external wrenches. Finally, extensive simulation and real-world experiments demonstrate whole-body compliant responses to wrenches across the upper body, monotonic modulation over the full range of both stiffness commands, and applicability to teleoperated loco-manipulation tasks. Project website: https://lac-humanoid.github.io/

cs.RO↗
arXiv2026

SeVeR: Selective Visual Exposure and Retrieval for 3D Medical Image Question Answering

Volumetric medical VQA requires reasoning over long and redundant 3D visual token sequences, especially in multi-sequence MRI where complementary modalities provide diverse diagnostic cues but expose the decoder to many repeated anatomical regions. To investigate reasoning under multi-sequence visual redundancy, we first introduce BreMRIs-VQA, a clinically curated breast MRI benchmark with 1.19M QA pairs from 71.0K sequences and 12.9K patients, covering both free-text and multiple-choice questions. We further propose SeVeR, a selective visual exposure framework that compresses dense volumes into modality-wise prototypes and retrieves complementary multi-level evidence with change-aware gated attention during decoding, trained with a marginal-utility self-consistency objective that suppresses unhelpful retrieval. Experiments on BreMRIs-VQA and public benchmarks show that SeVeR improves both discriminative and generative performance while exposing substantially fewer visual tokens.

cs.CV↗
arXiv2026

Observation of Quantized Charge Accumulation in a Quantum Anomalous Hall System

The quantum anomalous Hall effect in magnetically doped topological insulators exhibits a quantized Hall conductance $σ_{xy} = e^2/h$ arising from the two-dimensional surface states. While conventional transport probes confirm this quantization, they remain insensitive to the field-induced surface charge accumulation as a direct manifestation of $σ_{xy}$. Here, we experimentally validate an out-of-plane capacitive method that directly detects this quantized charge accumulation in a quantum anomalous Hall system. Using Corbino and simple disk devices, we measure charge accumulation proportional to field variation $ΔB$, with dissipation characterized by longitudinal conductance $σ_{xx}$ and frequency $f$. A quantitative dissipation model extracts the intrinsic quantized charge density $η_0 = (e^2/h)ΔB$, which is confirmed through $f$- and $σ_{xx}$-dependent measurements. Under ultra-low dissipation ($σ_{xx} \approx 10^{-9}$ S), we directly resolve the fully quantized charge accumulation. This methodology establishes a direct charge-accumulation probe and provides a pathway toward detecting the topological magnetoelectric effect, a condensed matter manifestation of the four-dimensional quantum Hall effect.

cond-mat.mes-hall↗
arXiv2026

Participation, selection and indicative bidding in auctions with costly entry

We study auctions with costly entry in which bidders have incomplete information about their affiliated private valuations prior to entry. We examine whether indicative bidding--a mechanism requiring non-binding preliminary bids before entry--can stimulate participation and improve entrant selection, and we compare its performance with unrestricted and capped entry in a controlled laboratory experiment. When entry costs are high, indicative bidding generates significantly more revenue, primarily by increasing participation beyond theoretical predictions. When entry costs are low, its predicted revenue advantage is attenuated by higher-than-predicted selection inefficiency.

econ.GN↗
arXiv2026

Tunable Tool-Call Rates in LLM Agents via Representation Steering

Deciding whether to call a tool is a core competence of an LLM agent, and a costly one to get wrong: needless calls add latency, accrue cost, and may trigger irreversible side effects, while missing calls leave the model confidently wrong on questions it could only answer through tool-calls. Models manage this balance poorly, both over-using and under-using tools. Existing methods such as post-training and prompt engineering are expensive and difficult to modify at inference time. We show that whether an instruction-tuned model calls a tool can be controlled by a single linear direction in its residual stream, extracted without any training from the model's own tool-use preference signal and turned into an inference-time intervention with no prompt change. Adding the direction with strength $α$ moves the call rate monotonically from near $0\% $ to over $90\%$ while keeping calls well-formed. The steering works in both directions: dialing it down suppresses calls, and dialing it up induces new calls that land precisely on the questions the model cannot answer from its own knowledge. We also show that the direction generalizes to unseen tools with strength comparable to each tool's own direction and without favoring any specific tool choice. With live tool execution, a single sweep of the steering traces a cost/accuracy Pareto frontier and nearly doubles open-domain QA accuracy ($0.29 \! \rightarrow \! 0.56$); the same recipe transfers across a diverse range of models spanning dense, MoE, and multimodal architectures, without any training. Our code is publicly available at https://github.com/YuqiChen4188/Steering-Tool-Use-Propensity.

cs.AI↗
arXiv2026

Topological Surface Charge Detection via Active Capacitive Compensation: A Pathway to the 4D Quantum Hall Effect

The topological magnetoelectric effect (TME) in three-dimensional topological insulators (TIs), described by $ΔP = \frac{e^2}{2h} N_{\rm Ch}^{(2)} ΔB$, serves as a condensed-matter realization of the four-dimensional quantum Hall effect (4D QHE). In dual-gate axion-insulator devices, the TME-induced polarization yields a current $I_{\rm TME} \propto (C_{\rm total}/C_{\rm S})\,Q_{\rm 4D\mathrm{-}QHE}$, where the signal is suppressed by the capacitance ratio $C_{\rm total}/C_{\rm S}$. Here we propose an active compensation scheme that introduces a tunable negative capacitance $C_{\rm comp} \approx -C_{\rm gate}$ into the gate line, effectively canceling the gate dielectric capacitance and driving $C_{\rm total}/C_{\rm S} \to 1$. We validate the method using a quantum anomalous Hall (QAH) device, which shares the same surface-state physics as the axion insulator but permits direct charge measurement via a single gate, recovering over $95\%$ of the quantized charge signal from an initially half-attenuated state. This compensation method provides a robust means of resolving minute TME signals, offering a promising pathway toward direct measurements of the 4D QHE.

cond-mat.mes-hall↗
arXiv2026

Constraining the Baryon Fraction in Extragalactic Diffuse Ionized Gas with 124 Localized Fast Radio Bursts

Fast radio bursts (FRBs) are increasingly recognized as powerful cosmological tools for constraining the baryon fraction in extragalactic diffuse ionized gas, presenting a promising approach to address the missing baryon problem. In this paper, we constrain the baryon fraction in extragalactic diffuse ionized gas ($f_\mathrm{d}$) utilizing the latest sample of 124 localized FRBs across three different cosmological models. Our analysis models the probability distribution of the extragalactic diffuse ionized gas dispersion measure with a form that accurately reproduces mock observations. For a constant $f_\mathrm{d}$ model, we find that more than 90\% of baryons reside in the diffuse ionized gas phase. This result is robust against the choice of dark-energy parametrization under the current combination of datasets, although the fitted cosmological parameters shift accordingly. We also find that the inferred $f_\mathrm{d}$ is sensitive to the assumed dispersion measure distributions of both the Milky Way halo and the FRB host galaxies. Furthermore, the current data do not show statistically significant evidence for redshift evolution in $f_\mathrm{d}$, but the constraints are limited by the redshift distribution of the sample. Our conclusions are insensitive to the adopted baryonic feedback parameters and to the dispersion measure selection effect. These results provide strong evidence that the majority of the missing baryons reside in the diffuse ionized intergalactic medium.

astro-ph.CO↗
arXiv2026

Token-Level Credit Assignment Optimization for Generative Document Retrieval

Generative retrieval models perform document retrieval by autoregressively generating document identifiers (DocIDs). This process naturally forms a sequential decision problem, i.e., the model makes a sequence of token-level decisions, selecting a DocID token at each decoding step, with the resulting complete sequence identifying the retrieved document. However, relevance feedback is available only after the complete DocID has been generated and mapped to a document, resulting in a granularity mismatch between token-level generation decisions and document-level retrieval supervision. Consequently, existing reinforcement learning methods for generative retrieval rely on sequence-level rewards, assigning the same document-level relevance signal to every decoding step. Such uniform credit assignment obscures the contribution of individual token decisions, making it difficult to identify which decisions contribute to retrieval success or failure. In this paper, we propose Token-Level Credit Assignment for Generative Retrieval (TCA), a fine-grained reinforcement learning framework that aligns the granularity of credit assignment with that of autoregressive DocID generation. Unlike assigning a single reward to an entire generated DocID, TCA derives fine-grained rewards by comparing the hidden-state trajectory of each generated DocID with the gold DocID trajectory obtained from a frozen reference model. These trajectory-based rewards provide differentiated feedback across decoding steps, allowing the policy to reinforce generation paths that remain aligned with the target DocID. Moreover, TCA decouples token-level credit assignment from policy optimization and can be instantiated with both GRPO and PPO. Experiments on benchmarks show that our method consistently outperforms baselines, demonstrating the effectiveness of fine-grained supervision for aligning DocID generation.

cs.IR↗
arXiv2026

SkillNet: Create, Evaluate, and Connect AI Skills

Current AI agents can flexibly invoke tools and execute complex tasks, yet their long-term advancement is hindered by the lack of systematic accumulation and transfer of skills. Without a unified mechanism for skill consolidation, agents frequently ``reinvent the wheel'', rediscovering solutions in isolated contexts without leveraging prior strategies. To address this challenge, we introduce SkillNet, an open infrastructure for creating, evaluating, and organizing AI skills at scale. SkillNet structures skills within a unified ontology that supports creating skills from heterogeneous sources, establishing rich relational connections, and performing multi-dimensional evaluation across Safety, Completeness, Executability, Maintainability, and Cost-awareness. Our infrastructure integrates a repository of over 600,000 skills, an interactive platform, and a versatile Python toolkit. Experiments on ALFWorld, WebShop, and ScienceWorld show 40% higher average rewards and 30% fewer execution steps across multiple backbone models. Furthermore, SkillNet-Gym benchmarks skill retrieval, utilization, and composition, while SkillNet-Fabric enables task-specific skill routing through lightweight Wikis. By formalizing skills as evolving, composable assets, SkillNet provides a robust foundation for agents to move from transient experience to durable mastery.

cs.AI↗