arXiv subjects

Wenbo Fu

Publications and source records attributed to Wenbo Fu.

15 recordsLinked to original sources

arXiv2026

Radial Kernel Attention for Global Feature Exchange in Polymer Machine Learning Force Fields

We introduce Radial Kernel Attention (RKA) for long-range modeling in polymer machine learning force fields. Existing approaches often rely on local neighbor graphs, which limit communication across polymer chains and their surrounding environments, whereas explicit all-pair attention incurs quadratic computational cost. RKA maps interatomic distances from the cluster centroid to learned radial features and combines their kernelized overlap with content-dependent queries and keys. Factorized aggregation and readout enable global feature exchange at linear complexity in the number of atoms for fixed feature dimensions. To ensure physically consistent predictions, we insert the invariant scalar update between EquiformerV3 local blocks and obtain conservative forces by differentiating the total energy, including its dependence on radial coordinate . Our models are trained on official OPoly26 training set for 12 epochs and evaluated directly after training on the complete official validation set. A matched 5M comparison shows that RKA reduces per-atom energy and force MAEs by 67.4% and 36.3%, respectively. The 20M RKA model attains a total-energy MAE of 15.81 meV (per-atom MAE of 0.1182 meV/atom) and a force MAE of 2.619 meV/Angstrom,improving all three metrics over UMA-S-1.2 and UMA-S-1.2.1 on the same validation structures. RKA provides a compact pathway for information exchange beyond local graph connectivity in polymer machine learning force fields.

cs.AI↗

arXiv2026

Learning to Live: A Step Toward Lifelong Embodied Intelligence

Embodied agents are usually organized around individual tasks, yet the physical world keeps changing between tasks and during execution. We introduce LIFE (Lifelong Intelligence from Experience), a framework in which a robot observes continuously, remembers selectively, and turns experience into new capabilities. RoboBeats, a family of lightweight streaming gates, tracks conditions set by the agent and discovers new events, deciding when an observation warrants a response or a memory write. Selected observations enter Plastic Memory, a flat store of images and structured records that keeps the source evidence, so new fields and indexes can be added when later questions need them. When existing tools fall short, Explore-Distill-Mount (EDM) turns solutions found during problem solving into validated, reusable tools and skills. Under matched protocols, RoboBeats substantially improves condition tracking over its base models and, at the same writing budget, retains more useful evidence for memory question answering than uniform sampling. A case study of limited scale, replaying 24 hours of continuous video recorded by a real robot, illustrates how the integrated system builds memory from a day of experience and uses it to answer everyday questions. LIFE takes a step toward robots that become more familiar with their surroundings and more capable through experience: learning not only to complete tasks, but to live in the world.

cs.RO↗

arXiv2026

Can Complementary Signals Bridge Similarity Islands? Manifold-Augmented Graph Embedding for Multimodal Recommendation

Multimodal Recommender Systems (MMRS) leverage rich item features to mitigate data sparsity, yet similarity-driven graph construction and propagation can favor semantically similar substitutes over functionally related complements. When behavioral associations connect semantically distant items, complementary signals can remain geometrically isolated, yielding fragmented "Similarity Islands" and semantic drift when standard propagation is applied. To overcome these limitations, we propose MAGE (Manifold-Augmented Graph Embedding), a framework for personalized multimodal ranking that decouples complementarity modeling into offline geometric warping and online topological propagation. The offline phase enhances item representations by extracting complement-aware residual directions via a Soft Conceptor mechanism and injecting them under a local gap-preserving constraint, incorporating co-purchase evidence while controlling the magnitude of geometric perturbation. Synergistically, the online phase employs an asymmetric dual-stream GNN that treats visual features as stable anchors and augmented textual features as flexible probes, fused through a self-calibrated cross-modal mechanism to reconcile shared semantics and modality-specific residual cues. Extensive experiments on four Amazon datasets show that MAGE yields consistent improvements over strong multimodal recommendation baselines. Together with the supporting analyses in the appendix, these results suggest that geometry-aware augmentation can alleviate the semantic mismatch that arises in complementary recommendation settings. Our code is available at https://github.com/Alex-bupt/mage.

cs.IR↗

arXiv2026

HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness

Embodied navigation requires agents to ground instructions or object goals in spatial observations and translate plans into successful execution. As multimodal large language models (MLLMs) become increasingly capable, they offer stronger support for navigation without task-specific training; however, improved semantic reasoning alone does not ensure that proposed actions remain consistent with spatial evidence, task progress, and execution outcomes. We introduce HarnessVLN, a zero-shot, training-free framework that unifies instruction-following and object-goal navigation through a shared Agent Harness. The Harness coordinates perception, memory, and execution tools through a unified interface, validating planner proposals for evidential support, geometric feasibility, and subgoal consistency before dispatch. It jointly manages hierarchical event memory and a persistent Spatiotemporal Graph to track task progress, preserve spatial evidence, and contextualize failures. Structured execution feedback updates this shared state, guiding subsequent planning, recovery, and termination. Across R2R, RxR, HM3D-v2, and HM3D-OVON, HarnessVLN achieves success rates of 59.6%, 51.4%, 76.0%, and 59.3%, respectively, outperforming prior training-free methods. Humanoid robot deployment further demonstrates its applicability to both navigation tasks in real-world environments. The project page is available at [https://agibot-harnessvln.netlify.app/].

cs.RO↗

arXiv2026

STEGNav: Spatio-Temporal Event Graph Reasoning for Multimodal Lifelong Object Navigation

Multimodal lifelong navigation requires an agent to autonomously explore unseen environments while sequentially completing navigation tasks specified by object categories, language descriptions, or reference images. Existing methods primarily accomplish these tasks by constructing state-centric semantic scene graphs. By treating scene graphs as persistent repositories of semantic observations, these methods struggle to distinguish similar instances, jointly represent semantic targets and exploration frontiers, and effectively exploit navigation memory and trajectory experience. To address these limitations, we propose Spatio-Temporal Event Graph Navigation (STEGNav), a training-free framework that extends conventional scene graphs into spatio-temporal event graphs along complementary spatial and temporal axes. The spatial axis performs query-conditioned instance grounding and jointly represents semantic targets and occupancy-aware exploration frontiers characterized by reachability, path cost, and exploration utility. The temporal axis employs trajectory-aware dual-window memory to retain recent decision--trajectory events and verified cross-subtask navigation outcomes. A VLM-based navigation agent reasons over the resulting spatio-temporal event graph and selects either a target instance or an exploration frontier as its next navigation goal. STEGNav achieves 66.3% SR and 39.7 SPL on GOAT-Bench, as well as SR scores of 64.0% and 69.4% on HM3Dv1 and HM3Dv2, respectively. Ablation studies and error analyses validate the complementary effects of the two axes, demonstrating that event-driven spatio-temporal representations improve navigation reliability and cross-subtask experience reuse.

cs.RO↗

arXiv2026

MapTab: A Diagnostic Benchmark for Long-Horizon Multi-Criteria Multimodal Reasoning on Heterogeneous Topological Graphs

Systematically evaluating Multimodal Large Language Models (MLLMs) is essential for advancing Artificial General Intelligence (AGI). Yet existing benchmarks remain inadequate for rigorously measuring their reasoning capabilities under multi-criteria constraints. To address this gap, we introduce MapTab, a multimodal benchmark designed to assess holistic multi-criteria reasoning in MLLMs through route-planning tasks. MapTab requires models to perceive and ground visual information from map images while integrating route attributes, such as Time and Price, from structured tables. It covers two scenarios: Metromap, spanning metro networks in 160 cities across 52 countries, and Travelmap, featuring 168 representative tourist attractions from 19 countries. Overall, MapTab includes 328 images, 196,800 route-planning queries, and 3,936 QA queries, incorporating four key criteria: Time, Price, Comfort, and Reliability. Extensive evaluations of 21 representative MLLMs show that current models still struggle with multicriteria multimodal reasoning. Notably, when visual perception is unreliable, multimodal reasoning can even underperform unimodal approaches. MapTab therefore offers a challenging and realistic testbed for systematically evaluating and advancing MLLMs across core perception, integration, numerical comparison, and route planning capabilities.

cs.LG↗

arXiv2026

Probing the Critical Point (CritPt) of AI Reasoning: a Frontier Physics Research Benchmark

While large language models (LLMs) with reasoning capabilities are progressing rapidly on high-school math competitions and coding, can they reason effectively through complex, open-ended challenges found in frontier physics research? And crucially, what kinds of reasoning tasks do physicists want LLMs to assist with? To address these questions, we present the CritPt (Complex Research using Integrated Thinking - Physics Test, pronounced "critical point"), the first benchmark designed to test LLMs on unpublished, research-level reasoning tasks that broadly covers modern physics research areas, including condensed matter, quantum physics, atomic, molecular & optical physics, astrophysics, high energy physics, mathematical physics, statistical physics, nuclear physics, nonlinear dynamics, fluid dynamics and biophysics. CritPt consists of 71 composite research challenges designed to simulate full-scale research projects at the entry level, which are also decomposed to 190 simpler checkpoint tasks for more fine-grained insights. All problems are newly created by 50+ active physics researchers based on their own research. Every problem is hand-curated to admit a guess-resistant and machine-verifiable answer and is evaluated by an automated grading pipeline heavily customized for advanced physics-specific output formats. We find that while current state-of-the-art LLMs show early promise on isolated checkpoints, they remain far from being able to reliably solve full research-scale challenges: the best average accuracy among base models is only 5.7%, achieved by GPT-5 (high), moderately rising to around 10% when equipped with coding tools. Through the realistic yet standardized evaluation offered by CritPt, we highlight a large disconnect between current model capabilities and realistic physics research demands, offering a foundation to guide the development of scientifically grounded AI tools.

cs.AI↗

arXiv2022

Categorical Representation Learning and RG flow operators for algorithmic classifiers

Following the earlier formalism of the categorical representation learning (arXiv:2103.14770) by the first two authors, we discuss the construction of the "RG-flow based categorifier". Borrowing ideas from theory of renormalization group flows (RG) in quantum field theory, holographic duality, and hyperbolic geometry, and mixing them with neural ODE's, we construct a new algorithmic natural language processing (NLP) architecture, called the RG-flow categorifier or for short the RG categorifier, which is capable of data classification and generation in all layers. We apply our algorithmic platform to biomedical data sets and show its performance in the field of sequence-to-function mapping. In particular we apply the RG categorifier to particular genomic sequences of flu viruses and show how our technology is capable of extracting the information from given genomic sequences, find their hidden symmetries and dominant features, classify them and use the trained data to make stochastic prediction of new plausible generated sequences associated with new set of viruses which could avoid the human immune system. The content of the current article is part of the recent US patent application submitted by first two authors (U.S. Patent Application No.: 63/313.504).

cs.LG↗

arXiv2018

$Z_2$ fractionalized phases of a solvable, disordered, $t$-$J$ model

We describe the phases of a solvable $t$-$J$ model of electrons with infinite-range, and random, hopping and exchange interactions, similar to those in the Sachdev-Ye-Kitaev models. The electron fractionalizes, as in an `orthogonal metal', into a fermion $f$ which carries both the electron spin and charge, and a boson $ϕ$. Both $f$ and $ϕ$ carry emergent $\mathbb{Z}_2$ gauge charges. The model has a phase in which the $ϕ$ bosons are gapped, and the $f$ fermions are gapless and critical, and so the electron spectral function is gapped. This phase can be considered as a toy model for the underdoped cuprates. The model also has an extended, critical, `quasi-Higgs' phase where both $ϕ$ and $f$ are gapless, and the electron operator $\sim f ϕ$ has a Fermi liquid-like $1/τ$ propagator in imaginary time, $τ$. So while the electron spectral function has a Fermi liquid form, other properties are controlled by $\mathbb{Z}_2$ fractionalization and the anomalous exponents of the $f$ and $ϕ$ excitations. This `quasi-Higgs' phase is proposed as a toy model of the overdoped cuprates. We also describe the critical state separating these two phases.

cond-mat.str-el↗

arXiv2017

Thermoelectric transport in disordered metals without quasiparticles: the SYK models and holography

We compute the thermodynamic properties of the Sachdev-Ye-Kitaev (SYK) models of fermions with a conserved fermion number, $\mathcal{Q}$. We extend a previously proposed Schwarzian effective action to include a phase field, and this describes the low temperature energy and $\mathcal{Q}$ fluctuations. We obtain higher-dimensional generalizations of the SYK models which display disordered metallic states without quasiparticle excitations, and we deduce their thermoelectric transport coefficients. We also examine the corresponding properties of Einstein-Maxwell-scalar theories on black brane geometries which interpolate from either AdS$_4$ or AdS$_5$ to an AdS$_2\times \mathbb{R}^2$ or AdS$_2\times \mathbb{R}^3$ near-horizon geometry. These provide holographic descriptions of non-quasiparticle metallic states without momentum conservation. We find a precise match between low temperature transport and thermodynamics of the SYK and holographic models. In both models the Seebeck transport coefficient is exactly equal to the $\mathcal{Q}$-derivative of the entropy. For the SYK models, quantum chaos, as characterized by the butterfly velocity and the Lyapunov rate, universally determines the thermal diffusivity, but not the charge diffusivity.

cond-mat.str-el↗

arXiv2017

Supersymmetric SYK models

We discuss a supersymmetric generalization of the Sachdev-Ye-Kitaev model. These are quantum mechanical models involving $N$ Majorana fermions. The supercharge is given by a polynomial expression in terms of the Majorana fermions with random coefficients. The Hamiltonian is the square of the supercharge. The ${\cal N}=1$ model with a single supercharge has unbroken supersymmetry at large $N$, but non-perturbatively spontaneously broken supersymmetry in the exact theory. We analyze the model by looking at the large $N$ equation, and also by performing numerical computations for small values of $N$. We also compute the large $N$ spectrum of "singlet" operators, where we find a structure qualitatively similar to the ordinary SYK model. We also discuss an ${\cal N}=2$ version. In this case, the model preserves supersymmetry in the exact theory and we can compute a suitably weighted Witten index to count the number of ground states, which agrees with the large $N$ computation of the entropy. In both cases, we discuss the supersymmetric generalizations of the Schwarzian action which give the dominant effects at low energies.

hep-th↗

arXiv2016

Numerical study of fermion and boson models with infinite-range random interactions

We present numerical studies of fermion and boson models with random all-to-all interactions (the SYK models). The high temperature expansion and exact diagonalization of the $N$-site fermion model are used to compute the entropy density: our results are consistent with the numerical solution of $N=\infty$ saddle point equations, and the presence of a non-zero entropy density in the limit of vanishing temperature. The exact diagonalization results for the fermion Green's function also appear to converge well to the $N=\infty$ solution. For the hard-core boson model, the exact diagonalization study indicates spin glass order. Some results on the entanglement entropy and the out-of-time-order correlators are also presented.

cond-mat.str-el↗

arXiv2014

Quantum quenches and competing orders: I. Time-dependent Hartree-Fock+BCS theory

We study the non-equlibrium dynamics of an electronic model of competing bond density wave order and $d$-wave superconductivity. In a time-dependent Hartree-Fock+BCS approximation, the dynamics reduces to the equations of motion of operators realizing the generators of SU(4) at each pair of momenta, $(\boldsymbol{k}, -\boldsymbol{k})$, in the Brillouin zone. We compare the results of numerical studies of our model with recent picosecond optical experiments.

cond-mat.supr-con↗

arXiv2014

Quantum quenches and competing orders

We study the nonequilibrium dynamics of an electronic model of competition between an unconventional charge density wave (a bond density wave) and $d$-wave superconductivity. In a time-dependent Hartree-Fock+BCS approximation, the dynamics reduces to the equations of motion of operators realizing the generators of SU(4) at each pair of momenta, ( $\boldsymbol{k}$, - $\boldsymbol{k}$ ), in the Brillouin zone. We also study the nonequilibrium dynamics of a quantum generalization of a O(6) nonlinear $σ$ model of competing orders in the underdoped cuprates [Hayward et al., Science $\boldsymbol{343}$, 1336 (2014)]. We obtain results, in the large $N$ limit of a O($N$) model, on the time dependence of correlation functions following a pulse disturbance. We compare our numerical studies with recent picosecond optical experiments. We find that, generically, the oscillatory responses in our models share various qualitative features with the experiments.

cond-mat.supr-con↗

arXiv2012

Separation induced resonances in quasi-one-dimensional ultracold atomic gases

We study the effective one-dimensional (1D) scattering of two distinguishable atoms confined individually by {\em separated} transverse harmonic traps. With equal trapping frequency for two s-wave interacting atoms, we find that by tuning the trap separations, the system can undergo {\em double} 1D scattering resonance, named as the separation induced resonance(SIR), when the ratio between the confinement length and s-wave scattering length is within $(0.791,1.46]$. Near SIR, the scattering property shows unique dependence on the resonance position. The universality of a many-body system on scattering branch near SIR is demonstrated by studying the interaction effect of a localized impurity coupled with a Fermi sea of light atoms in a quasi-1D trap.

cond-mat.quant-gas↗