arXiv subjects

Jian Li

Publications and source records attributed to Jian Li.

At least 91 records · Page 5Linked to original sources

arXiv2026

Harnessing Pre-Resolution Signals for Future Prediction Agents

Many high-stakes decisions depend on forecasts made before outcomes are known. In this future prediction setting, the central challenge is that public evidence evolves over time, while the main supervision signal arrives only after resolution: the realized outcome mainly assesses final correctness, offering only coarse guidance on what to track, what to verify, and which judgments to leave uncertain along the way. Our key observation is that revisiting the same unresolved question over time creates informative temporal contrasts across evolving evidence and repeated forecasts, exposing what earlier attempts missed before resolution and yielding a diagnostic signal we call the pre-resolution signal. We instantiate this idea in Milkyway, a future prediction agent with a persistent future prediction harness, an editable external state that stores reusable procedural guidance across revisits to the same unresolved question. As the same unresolved question is revisited, Milkyway extracts pre-resolution signals from evolving evidence and repeated forecasts, uses them to update the harness, and improves later forecasts on that question before resolution. After resolution, the realized outcome serves as a post-resolution check of provisional updates. On the FutureX and FutureWorld benchmarks, Milkyway achieves strong performance against competitive baselines, and a mechanism study suggests that the gains stem from harness evolution driven by pre-resolution signals rather than repeated prediction alone.

cs.AI↗

arXiv2026

Nuclear level densities in the relativistic Hartree-Bogoliubov plus combinatorial framework

A systematic study of nuclear level densities has been carried out within the relativistic Hartree-Bogoliubov plus combinatorial framework. Calculations were performed for even-even nuclei with available experimental data, based on the relativistic energy density functionals DD-ME2, DD-PC1, and PC-PK1. The overall performance of the model is assessed against experimental data. On this basis, the effects of different functionals, pairing correlations, deformation, and other relevant factors on nuclear level densities are examined. The results show that the present framework provides a good description of the experimental level density and reproduces the s-wave neutron resonance spacings with an accuracy comparable to that of the best existing global models. Furthermore, differences among the adopted relativistic density functionals in the nucleon effective mass at saturated nuclear matter are transmitted to the predicted level densities and constitute the main source of the differences among the results obtained with the three functionals.

nucl-th↗

arXiv2026

GUI Agents with Reinforcement Learning: Toward Digital Inhabitants

Graphical User Interface (GUI) agents have emerged as a promising paradigm for intelligent systems that perceive and interact with graphical interfaces visually. Yet supervised fine-tuning alone cannot handle long-horizon credit assignment, distribution shifts, and safe exploration in irreversible environments, making Reinforcement Learning (RL) a central methodology for advancing automation. In this work, we present the first comprehensive overview of the intersection between RL and GUI agents, and examine how this research direction may evolve toward digital inhabitants. We propose a principled taxonomy that organizes existing methods into Offline RL, Online RL, and Hybrid Strategies, and complement it with analyses of reward engineering, data efficiency, and key technical innovations. Our analysis reveals several emerging trends: the tension between reliability and scalability is motivating the adoption of composite, multi-tier reward architectures; GUI I/O latency bottlenecks are accelerating the shift toward world-model-based training, which can yield substantial performance gains; and the spontaneous emergence of System-2-style deliberation suggests that explicit reasoning supervision may not be necessary when sufficiently rich reward signals are available. We distill these findings into a roadmap covering process rewards, continual RL, cognitive architectures, and safe deployment, aiming to guide the next generation of robust GUI automation and its agent-native infrastructure.

cs.AI↗

arXiv2026

Context-Fidelity Boosting: Enhancing Faithful Generation through Watermark-Inspired Decoding

Large language models (LLMs) often produce content that contradicts or overlooks information provided in the input context, a phenomenon known as faithfulness hallucination. In this paper, we propose Context-Fidelity Boosting (CFB), a lightweight and general decoding-time framework that reduces such hallucinations by increasing the generation probability of source-supported tokens. Motivated by logit-shaping principles from watermarking techniques, CFB applies additive token-level logit adjustments based on a token's degree of support from the input context. Specifically, we develop three boosting strategies: static boosting, which applies a fixed bias to source-supported tokens; context-aware boosting, which scales this bias using the divergence between next-token distributions with and without context; and token-aware boosting, which further redistributes the adaptive bias according to local relevance estimated from source-position attention and source-scoped semantic similarity. CFB requires no retraining or architectural changes, making it compatible with a wide range of LLMs. Experiments on summarization and question answering tasks across multiple open-source LLMs show that CFB consistently improves faithfulness metrics with minimal generation overhead. Our implementation is fully open-sourced.

cs.CL↗

arXiv2026

AD-Copilot: A Vision-Language Assistant for Industrial Anomaly Detection via Visual In-context Comparison

Multimodal Large Language Models (MLLMs) have achieved impressive success in natural visual understanding, yet they consistently underperform in industrial anomaly detection (IAD). This is because MLLMs trained mostly on general web data differ significantly from industrial images. Moreover, they encode each image independently and can only compare images in the language space, making them insensitive to subtle visual differences that are key to IAD. To tackle these issues, we present AD-Copilot, an interactive MLLM specialized for IAD via visual in-context comparison. We first design a novel data curation pipeline to mine inspection knowledge from sparsely labeled industrial images and generate precise samples for captioning, VQA, and defect localization, yielding a large-scale multimodal dataset Chat-AD rich in semantic signals for IAD. On this foundation, AD-Copilot incorporates a novel Comparison Encoder that employs cross-attention between paired image features to enhance multi-image fine-grained perception, and is trained with a multi-stage strategy that incorporates domain knowledge and gradually enhances IAD skills. In addition, we introduce MMAD-BBox, an extended benchmark for anomaly localization with bounding-box-based evaluation. The experiments show that AD-Copilot achieves 82.3% accuracy on the MMAD benchmark, outperforming all other models without any data leakage. In the MMAD-BBox test, it achieves a maximum improvement of $3.35\times$ over the baseline. AD-Copilot also exhibits excellent generalization of its performance gains across other specialized and general-purpose benchmarks. Remarkably, AD-Copilot surpasses human expert-level performance on several IAD tasks, demonstrating its potential as a reliable assistant for real-world industrial inspection. All datasets and models will be released for the broader benefit of the community.

cs.CV↗

arXiv2026

FOREVER: Forgetting Curve-Inspired Memory Replay for Language Model Continual Learning

Continual learning (CL) for large language models (LLMs) aims to enable sequential knowledge acquisition without catastrophic forgetting. Memory replay methods are widely used for their practicality and effectiveness, but most rely on fixed, step-based heuristics that often misalign with the model's actual learning progress, since identical training steps can result in varying degrees of parameter change. Motivated by recent findings that LLM forgetting mirrors the Ebbinghaus human forgetting curve, we propose FOREVER (FORgEtting curVe-inspired mEmory Replay), a novel CL framework that aligns replay schedules with a model-centric notion of time. FOREVER defines model time using the magnitude of optimizer updates, allowing forgetting curve-inspired replay intervals to align with the model's internal evolution rather than raw training steps. Building on this approach, FOREVER incorporates a forgetting curve-based replay scheduler to determine when to replay and an intensity-aware regularization mechanism to adaptively control how to replay. Extensive experiments on three CL benchmarks and models ranging from 0.6B to 13B parameters demonstrate that FOREVER consistently mitigates catastrophic forgetting.

cs.LG↗

arXiv2026

A neural network approach for two-body systems with spin and isospin degrees of freedom

We propose an enhanced machine learning method to calculate the ground state of two-body systems. By extending the original method [Naito, Naito, and Hashimoto, Phys. Rev. Research 5, 033189 (2023)], the present method enables consideration of the spin and isospin degrees of freedom by employing a non-fully connected deep neural network and the unsupervised machine learning technique. The validity of this method is verified by calculating the unique bound state of the deuteron.

nucl-th↗

arXiv2026

FAST Polarization Catalog of FRB 20240114A

Polarization measurements of fast radio bursts (FRBs) probe the magnetized plasma surrounding their central engines. FRB~20240114A is an exceptionally active repeating source, with 17,356 bursts detected between 2024 January 28 and 2025 May 30 by FAST, enabling time-resolved polarimetric studies. In this work, we present a polarimetric catalog of 6,131 bright bursts (with a signal-to-noise ratio S/N $\geq$ 20, 35.3% of the total sample), including arrival time (MJD$_{\text{topo}}$), dispersion measure (DM), burst width (W$_{\text{eff}}$), bandwidth, Faraday rotation measure (RM), linear and circular polarization degrees (DOL, DOC), and intrinsic polarization angle (PA$_0$). We detect a clear temporal evolution of RM: after an initial stable phase, it decreases linearly by $\sim$200 $\rm rad\ m^{-2}$ over 200 days, forming a bimodal distribution, whereas DM remains stable at 528.9 $\rm pc\ cm^{-3}$. The linear polarization fraction is generally high, with the 3$σ$ lower bound around 76%, while circular polarization is low, with 1,157 of 17,356 bursts (6.67%) having DOC $\geq$10%. We perform a power-law fit between $|\textrm{V}|$/I and $|\textrm{RM}|$, which yields an index of $-2.98 \pm 0.80$. It is found that the combined 2D distribution of L/I versus V/I remains stable, implying that the emission mechanism is largely invariant. Our PA$_0$ measurements show a broad, non-uniform distribution, implying a complex emission geometry. These results suggest that FRB~20240114A resides in a dynamically evolving magneto-ionic environment. This catalog provides a foundation for studies of repeating FRB progenitors and their environments.

astro-ph.HE↗

arXiv2026

Predicting Video Slot Attention Queries from Random Slot-Feature Pairs

Unsupervised video Object-Centric Learning (OCL) is promising as it enables object-level scene representation and understanding as we humans do. Mainstream video OCL methods adopt a recurrent architecture: An aggregator aggregates current video frame into object features, termed slots, under some queries; A transitioner transits current slots to queries for the next frame. This is an effective architecture but all existing implementations both (\textit{i1}) neglect to incorporate next frame features, the most informative source for query prediction, and (\textit{i2}) fail to learn transition dynamics, the knowledge essential for query prediction. To address these issues, we propose Random Slot-Feature pair for learning Query prediction (RandSF.Q): (\textit{t1}) We design a new transitioner to incorporate both slots and features, which provides more information for query prediction; (\textit{t2}) We train the transitioner to predict queries from slot-feature pairs randomly sampled from available recurrences, which drives it to learn transition dynamics. Experiments on scene representation demonstrate that our method surpass existing video OCL methods significantly, e.g., up to 10 points on object discovery, setting new state-of-the-art. Such superiority also benefits downstream tasks like scene understanding. Source Code, Model Checkpoints, Training Logs: https://github.com/Genera1Z/RandSF.Q

cs.CV↗

arXiv2026

Disco-RAG: Discourse-Aware Retrieval-Augmented Generation

Retrieval-Augmented Generation (RAG) has emerged as an important means of enhancing the performance of large language models (LLMs) in knowledge-intensive tasks. However, most existing RAG strategies treat retrieved passages in a flat and unstructured way, which prevents the model from capturing structural cues and constrains its ability to synthesize knowledge from dispersed evidence across documents. To overcome these limitations, we propose Disco-RAG, a discourse-aware framework that explicitly injects discourse signals into the generation process. Our method constructs intra-chunk discourse trees to capture local hierarchies and builds inter-chunk rhetorical graphs to model cross-passage coherence. These structures are jointly integrated into a planning blueprint that conditions the generation. Experiments on question answering and long-document summarization benchmarks show the efficacy of our approach. Disco-RAG achieves state-of-the-art results on the benchmarks without fine-tuning. These findings underscore the important role of discourse structure in advancing RAG systems.

cs.CL↗

arXiv2026

POLAR: Online Learning for LoRA Adapter Caching and Routing in Edge LLM Serving

Edge deployment of large language models (LLMs) increasingly relies on libraries of lightweight LoRA adapters, yet GPU/DRAM can keep only a small resident subset at a time. Serving a request through a non-resident adapter requires paging its weights from storage, incurring measurable latency. This creates a two-timescale online control problem: on a slow timescale, the system selects which adapters remain resident in fast memory, while on a fast timescale it routes each request to an adapter whose context-dependent utility is unknown a priori. The two decisions are tightly coupled: the cache determines the cost of exploration, and the router determines which adapters receive informative feedback. We formulate this joint caching-and-routing problem as a two-timescale contextual bandit and propose POLAR (Paging and Online Learning for Adapter Routing). POLAR pairs a cache-aware LinUCB router with an epoch-based cache controller. We study two variants. A fixed-epoch version provides a robust baseline with worst-case regret guarantees under arbitrary contexts. An epoch-doubling version, POLAR+, adds forced exploration and improved cache optimization to achieve $\widetilde{\mathcal{O}}(d\sqrt{NT}+\sqrt{KT})$ sublinear regret under stochastic regularity and cacheability conditions, where $N$ is the adapter count, $K$ the cache size, $d$ the context dimension, and $T$ the horizon. The routing term matches the standard contextual-bandit rate up to logarithmic factors, showing that the memory hierarchy does not fundamentally slow routing learning. Experiments using 15 real LoRA adapters for Qwen2.5-7B together with measured GPU paging latencies show that adaptive cache control substantially outperforms non-adaptive baselines and exhibits scaling trends consistent with the theory.

cs.LG↗

arXiv2026

QuIKS: Near-Zero Latency Key Supply with Adaptive Buffering for Resource-Efficient Quantum Key Distribution Networks

Quantum key distribution (QKD) networks provide information-theoretically secure keys for distant parties, emerging as a vital alternative to classical cryptography infrastructures threatened by quantum computing. In QKD networks, the immediacy of key supply service is crucial to the security and performance of applications, as their data must be encrypted before transmission. While key buffering can enable instant key supply services, existing schemes rely on heuristic solutions that incur prohibitive key resource consumption, thus significantly hindering practical deployment. To address this issue, we propose QuIKS, an instant key supply scheme based on adaptive buffering, offering the dominant advantage of near-zero key supply latency while consuming ultra-low key resources (i.e., ultra-low buffer size). Specifically, it is built upon a novel analytical model that determines the minimum buffer size required to guarantee near-zero-latency key supply performance. Guided by this model, QuIKS introduces a lightweight two-phase control algorithm that dynamically determines key relaying requests and adjusts the buffer size by probing real-time application patterns and network conditions. Experiments on a real QKD network testbed demonstrate that QuIKS achieves near-zero key supply latency while providing a more than 10-fold reduction in key buffer size compared to state-of-the-art schemes.

quant-ph↗

arXiv2026

Micro-Macro Retrieval: Reducing Long-Form Hallucination in Large Language Models

Large Language Models (LLMs) achieve impressive performance across many tasks but remain prone to hallucination, especially in long-form generation where redundant retrieved contexts and lengthy reasoning chains amplify factual errors. Recent studies highlight a critical phenomenon: the closer key information appears to the model outputs, the higher the factual accuracy. However, existing retrieval-augmented language models (RALMs) lack effective mechanisms to ensure this proximity - external evidence is injected into reasoning via multi-turn retrieval, but this cannot ensure key information stays close to the outputs. We propose Micro-Macro Retrieval (M2R), a novel retrieve-while-generate framework to fill this gap. At the macro level, M2R retrieves coarse-grained evidence from external sources; at the micro level, it extracts essential results from a key information repository built during reasoning and reuses them while generating answers. This design directly addresses the key-information-to-output proximity bottleneck, effectively reducing hallucination in long-form tasks. M2R is trained with a curriculum learning-based reinforcement learning strategy using customized rule-based rewards, enabling stable acquisition of retrieval and grounding skills. Extensive experiments across different benchmarks demonstrate the effectiveness of M2R, especially in lengthy-context settings.

cs.CL↗

arXiv2026

Predictions of charge density distributions for nuclei with $Z \geq 8$

A deep neural network (DNN) has been developed to accurately predict nuclear charge density distributions for nuclei with proton numbers $Z \geq 8$. By incorporating essential nuclear structure features, the model achieves a significant improvement in predictive accuracy over conventional methods. The charge density distributions are analyzed using a Fourier-Bessel (FB) series expansion, and the DNN is trained on a comprehensive dataset derived from relativistic continuum Hartree-Bogoliubov (RCHB) theory calculations. The model demonstrates exceptional performance, with root-mean-square deviations of 0.0123 fm and 0.0198 fm for charge radii on the training and validation sets, respectively, remarkably surpassing the precision of the original RCHB calculations. Beyond advancing nuclear physics research, this high-precision model provides critical data for applications in atomic physics, nuclear astrophysics, and related fields.

nucl-th↗

arXiv2026

Reduced Optical Gain Threshold by Carrier Multiplication in Semiconductor Perovskite Nanocrystals

Carrier multiplication (CM) describes a strong charge-carrier interaction process in semiconductor colloidal nanocrystals (NCs), wherein two band-edge excitons are simultaneously created by an absorbed photon with at least twice the bandgap energy (2 Eg). While being fundamentally intriguing, it has been exclusively utilized to enhance the light-to-electricity conversion efficiencies in the photodetector and solar-cell devices. In this report, we have synthesized the core/shell perovskite FAPbI3/NdF3 NCs with a biexciton recombination lifetime of ~3.9 ns, and demonstrated that a CM efficiency of ~25.7% can be achieved under the ~355 nm laser excitation (~2.21 Eg). This CM occurrence leads to a two-fold reduction in the optical gain threshold, as compared to that obtained under the ~640 nm laser excitation (~1.23 Eg). When combined with the single-exciton and zero-threshold optical gain schemes previously developed for semiconductor colloidal NCs, the CM effect introduced here would further mitigate the optical-pumping requirement for the routine operation of continuous-wave lasing.

physics.optics↗

arXiv2026

A 44-minute periodic radio transient in a supernova remnant

Long-period radio transients (LPTs) are a newly discovered class of radio emitters with periods ranging from minutes to hours. The astrophysical nature remains undetermined, particularly of LPTs with no detectable companions. We report the first evidence for a plausible supernova remnant (SNR) association with an LPT (DART J1832-0911, 2656.23+-0.15 s period), which supports a neutron star origin of such objects. The dispersion measure of this LPT, SNR's CO emission and HI absorption, and low probability of chance of alignment with field pulsars are all consistent with such an association. The source displays either phase-locked circular or nearly 100\% linear polarization, indicating its strong and geometrically stable magnetic field. No detectable optical counterpart was found, even with a 10m-class telescope. The SNR association and the stable polarization suggest that DART J1832-0911 most likely originates from a young neutron star, whose spin could have been braked by supernova's fallback materials. This discovery provides critical insights into the nature of ultra-long period transients and their link to stellar remnants.

astro-ph.HE↗

arXiv2026

The rigidity of dimension estimate for holomorphic functions on Kähler manifolds

In this paper, we obtain the optimal rigidity of dimension estimate for holomorphic functions with polynomial growth on Kähler manifolds with non-negative holomorphic bisectional curvature. There is a specific gap between the largest and the second largest dimension. We also determine the optimal dimension that ensures the maximal volume growth which implies the manifold is biholomorphic to the complex Euclidean space.

math.DG↗

arXiv2026

Morphology-Consistent Humanoid Interaction through Robot-Centric Video Synthesis

Equipping humanoid robots with versatile interaction skills typically requires either extensive policy training or explicit human-to-robot motion retargeting. However, learning-based policies face prohibitive data collection costs. Meanwhile, retargeting relies on human-centric pose estimation (e.g., SMPL), introducing a morphology gap. Skeletal scale mismatches result in severe spatial misalignments when mapped to robots, compromising interaction success. In this work, we propose Dream2Act, a robot-centric framework enabling zero-shot interaction through generative video synthesis. Given a third-person image of the robot and target object, our framework leverages video generation models to envision the robot completing the task with morphology-consistent motion. We employ a high-fidelity pose extraction system to recover physically feasible, robot-native joint trajectories from these synthesized dreams, subsequently executed via a general-purpose whole-body controller. Operating strictly within the robot-native coordinate space, Dream2Act avoids retargeting errors and eliminates task-specific policy training. We evaluate Dream2Act on the Unitree G1 across four whole-body mobile interaction tasks: ball kicking, sofa sitting, bag punching, and box hugging. Dream2Act achieves a 37.5% overall success rate, compared to 0% for conventional retargeting. While retargeting fails to establish correct physical contacts due to the morphology gap (with errors compounded during locomotion), Dream2Act maintains robot-consistent spatial alignment, enabling reliable contact formation and substantially higher task completion.

cs.RO↗