arXiv subjects

Wei He

Publications and source records attributed to Wei He.

At least 19 recordsLinked to original sources

arXiv2026

Scaling Verifiable Environments for Long-horizon Work Agents

Work agents operate over digital artifacts to execute professional knowledge-intensive work, requiring training environments that support long-horizon interaction and trustworthy verification. However, hand-crafted environments incur prohibitive engineering overhead that prevents environment scaling, whereas synthesis methods sacrifice workspace complexity, realism, or grounded verifiability. To bridge this gap, we introduce WorkForge, a scalable synthesis framework for constructing verifiable work-agent environments from real-world resources. Starting from expert workflows, WorkForge first identifies the resources, decisions, and deliverables required by each workflow. It then retrieves relevant real-world files and organizes them into a workspace. WorkForge inspects the workspace to extract concrete, checkable facts about its content. These factual anchors fix which task types the workspace can support and how their outcomes can be verified. Therefore, WorkForge derives each task's instructions, solution plan, and complementary programmatic and semantic verifiers directly from these factual anchors, keeping verification traceable to observable workspace evidence. Furthermore, we construct 16.7K verifiable environments across 40 professional domains, with workspaces collectively covering 60 file types. Post-training Qwen3.5-35B-A3B-Base improves GDPVal from 45.5 to 73.6 and APEX Score from 5.0 to 21.3, while enabling Qwen3.5-27B to achieve highly competitive performance and outperform strong competitors. Our analyses confirm the efficacy of the proposed method and reveal consistent scaling behaviors across both data volume and interaction horizons.

cs.CL↗

arXiv2026

Extremal Perfect Monotone Equilibria in Bayesian Games with Strategic Complementarities

We study sharp bounds on equilibrium behavior in Bayesian games when players may make mistakes. Under strategic complementarities, the set of behavioral perfect equilibria has a largest and a smallest element, both in pure strategies. We ask when these bounds are monotone and how they depend on the way mistakes are made. For monotone supermodular games, we consider limits of equilibria under increasing perturbations, in the sense that conditional tremble distributions have full support and increase stochastically with type. We show that the largest and smallest perfect equilibria exist, which are pure and monotone. Under the general perfection notion, the extremal perfect equilibria continue to be monotone in the setting with finite actions. With infinite actions, the extremal perfect equilibria can be nonmonotone, and the set of perfect monotone equilibria can even lack extrema.

econ.TH↗

arXiv2026

Optimal Steady-State Regulation of Nonlinear Systems: A Feedback Optimization Approach

This paper studies the optimal steady-state regulation problem for a class of nonlinear systems. The regulated output is required to track a trajectory that minimizes a steady-state cost depending on both the output and the exo-system. To characterize the desired steady-state behavior, we introduce a generalized regulator equation and develop a two-layer framework. The optimization layer generates the optimal reference online through a steady-state generator combined with gradient-feedback optimization dynamics, while the control layer employs backstepping to stabilize the plant and achieve the desired optimal steady-state regulation. Under suitable immersion and minimum-phase assumptions, global asymptotic stability is established.

eess.SY↗

arXiv2026

Hidden urban exposure to energy-price shocks: spreading burdens and shifting hotspots

Cities need to anticipate where energy-price shocks could intensify hardship beyond established hotspots. How income distributions and fuel-price balances jointly shape this hidden exposure remains unclear. Here we address this gap by stress-testing nearly one million properties in six English local authorities, combining required-energy costs for existing housing with modelled local income distributions. Under gas-led crisis prices, 41% of neighbourhoods in the highest local exposure quarter lie outside the highest official fuel-poverty quarter. Electricity-led stress changes this hidden geography: when electricity unit prices alone triple from 2021, comparable to the unmitigated late-2022 cap increase, 38% of the resulting hidden-exposure group are absent from the crisis-based group. Entering neighbourhoods have newer, more efficient housing, more flats and greater electricity dependence than retained areas. Crisis preparedness should extend household protection beyond current hardship hotspots, using fuel-specific stress tests to anticipate support needs as fuel prices and heating systems change.

physics.soc-ph↗

arXiv2026

Fabric retrofit scenario analysis under uncertainty for an urban housing stock using machine learning

Fabric retrofit is central to decarbonising housing, and planning it depends on expected savings and their uncertainty. This study asks which upgrade saves most per dwelling and across a housing portfolio, how uncertain the savings are, and whether upgrade order matters. We propose an uncertainty-aware framework that converts predicted changes in energy performance certificate scores after wall, roof and window upgrades into standardised energy-cost savings and emission reductions. Six machine-learning models predict the 1-100 Standard Assessment Procedure (SAP) energy-cost and environmental impact (EI) emission scores from 17 certificate descriptors. Each model's errors on similar dwellings give calibrated prediction intervals. A global-best strategy applies the most accurate model throughout; a sample-specific strategy chooses each dwelling's model using these local errors. The analysis covers 103,414 officially registered dwellings in Westminster, London. CatBoost was the global-best model, with mean absolute errors of 3.487 SAP and 4.379 EI points on held-out dwellings. Both strategies projected similar savings; under global-best, roof insulation saved most per dwelling, GBP 203 annually (uncertainty range 176-229), and wall insulation most across the portfolio, GBP 6.09 million annually (uncertainty range 5.69-6.72). For 5,382 dwellings needing all three upgrades, every order delivered the same final package, yet installing roofs, walls, then windows a year apart accumulated 54% more projected cost savings over three years than the reverse order. Emission reductions followed the same pattern. Using only certificate data, the framework helps local authorities rank measures for dwellings and portfolios and sequence upgrades to bring savings forward.

physics.soc-ph↗

arXiv2026

Where Natural Language Ends in Agents: Executable Skill Composition with MetaSkill

Agent Skills express open-ended task knowledge in natural language, but composing them requires persistent execution order and upstream-output handoffs. We propose selective formalization at Skill boundaries: coordination relations with stable references and machine-checkable constraints become executable, while substantive task judgments remain within Skills. We introduce MetaSkill, a hybrid representation that preserves complete Skill packages and local instructions while making invocation dependencies and predecessor-text bindings executable through a directed acyclic graph (DAG) runtime. This interface supports explicit coordination without prescribing how each Skill performs its local task. Across 20 multi-Skill artifact tasks and two LLMs, we compare natural-language procedures, prompt-visible graphs, and MetaSkill derived from a common frozen plan per task, using the same designated Skill packages. Additional comparisons evaluate AgentSkillOS with the same available Skill packages as MetaSkill, as well as agent systems using their configured tools. Blinded artifact judgments yield favorable preference point estimates across all reported comparisons, with 95% task-bootstrap intervals excluding parity against Prompt Graph under DeepSeek-V4-Flash and OpenClaw under GLM-5.2. In audited GLM-5.2 execution traces, all planned nodes complete and all declared precedence constraints are satisfied. Together, these findings support MetaSkill as a practical interface for composing open-ended Skills, combining local linguistic flexibility with executable dependencies, explicit predecessor-text bindings, and auditable coordination. Code is available at: https://github.com/TokenRhythm/opensquilla/tree/paper/metaskill-lp4fm-2026

cs.AI↗

arXiv2026

Two quantum spin liquid states coexisting in the kagome antiferromagnet herbertsmithite

Herbertsmithite (ZnCu$_3$(OH)$_6$Cl$_2$) is one of the most extensively studied quantum spin liquid (QSL) candidates, yet the precise nature of its ground state remains unresolved. Here, we compare the results of high-resolution inelastic neutron scattering on single crystals of herbertsmithite (Zn$_\mathrm{x}$Cu$_\mathrm{4-x}$(OD)$_6$Cl$_2$, $x\simeq 0.85$) with large-scale density-matrix renormalization group (DMRG) and dynamical DMRG simulations to benchmark the spin Hamiltonian and uncover fundamental properties of the ground state. The measured spin correlations determine an energy density for the Heisenberg model that closely matches simulations, an essential verification of the Hamiltonian. In parallel, DMRG simulations reveal two nearly degenerate QSL states. Both preserve translational symmetry, but one retains the lattice point-group symmetries whereas the other develops anisotropic spin correlations that break point-group symmetry. These two QSLs have distinct spin structure factors, and, remarkably, close agreement with the neutron spectra is achieved with an admixture of the two states. Our results establish a quantitative correspondence between neutron scattering and simulations, and suggest that herbertsmithite realizes a regime in which two proximate gapped QSL states coexist.

cond-mat.str-el↗

arXiv2026

NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness

Recursive self-improvement (RSI) requires a concrete mechanism through which an AI system observes its capabilities and converts that evidence into the next round of learning. We present NeoHorse-1, a family of agent-native models developed to explore this path through agentic post-training. Our system combines a heterogeneous model pool with intelligent routing, recording the predicted capability demand, selected service tier, and subsequent interaction for each user turn. These records are converted into training examples that preserve interleaved reasoning, tool calls, and harness context, and are admitted through structural validation, six-dimensional semantic evaluation, and subscene-level labeling. Routing signals organize supervised fine-tuning into a three-stage curriculum and extend to routing-guided on-policy distillation, where a teacher supervises student-generated responses under the same progression. Capability-guided allocation then converts evaluation feedback into the next training mixture, closing an evaluation-selection-update loop in which what the system learns to do shapes what it learns from next. Across eleven benchmarks covering harness-based agents, tool use, coding, and instruction following, post-training raises the macro-average from 58.94 to 64.87 at 4B and from 65.60 to 69.04 at 9B, substantially narrowing the aggregate gap between the post-trained 4B model and the 9B base model. We further adapt NeoHorse-1-4B into NeoHorse-Jev-4B, a prefill-only model that returns structured decisions and probabilities for agent workflows. NeoHorse-1 provides an initial prototype of this feedback-driven process and a path toward harness-mediated RSI across successive iterations.

cs.CL↗

arXiv2026

Binding energies in the $0^{-+}$ and $1^{--}$ fermion--antifermion channels from the ladder Bethe-Salpeter and Schrödinger equations

In this work, we calculate the binding energies of an equal-mass fermion--antifermion system in the $0^{-+}$ and $1^{--}$ channels from the ladder Bethe--Salpeter (BS) equation with scalar, pseudoscalar, and vector exchange. The results are compared with those of the Schrödinger equation with the corresponding leading static interaction in two regimes: pointlike massless exchange at weak coupling and finite-mass exchange with monopole form factors at stronger coupling. For the BS calculation, we use two independent numerical methods based on momentum-space integral equations and coordinate-space differential equations. The differential-equation method gives estimated relative binding-energy errors below $10^{-9}$ in both regimes. For massless vector exchange in Coulomb gauge, the numerical results are consistent with the uncrossed-ladder perturbative spectrum through the $O(mα^5)$ terms. The results for massless scalar exchange and vector exchange in Feynman gauge are consistent with the common Wick--Cutkosky expansion through order $mα^3$. For finite exchange masses, the differences from the leading static Schrödinger spectra at the same masses, couplings, and form factors depend strongly on the exchange and channel. Scalar exchange gives nearly degenerate channels with much weaker binding than the Schrödinger reference. In contrast, vector exchange with only the metric term retained gives stronger binding than the Schrödinger reference in $0^{-+}$ and weaker binding in $1^{--}$. Notably, retaining the full Proca numerator substantially increases the vector binding in both channels. Pseudoscalar exchange also supports bound states and favors $1^{--}$, although its leading static central interaction vanishes.

nucl-th↗

arXiv2026

Three omitted values and non-Blaschke point divisors in half-planes

We construct a real meromorphic function $F$ on $\mathbb C$ such that $F^{-1}(\{0,1,\infty\})\subset\mathbb R$, while $F$ is not of bounded type in either half-plane. More strongly, for every $a\in\widehat{\mathbb C}\setminus\{0,1,\infty\}$, the $a$-point divisor in either half-plane fails the Blaschke condition. Thus the construction provides an independent negative answer to a question going back to Nevanlinna's 1925 work that had remained open for over a century. Postcomposition gives the analogous counterexample for any prescribed triple of distinct values in the Riemann sphere. The core construction and proof were generated during an autonomous run of GPT-5.6 Sol Ultra.

math.CV↗

arXiv2026

Multiorbital Anisotropy and Magnetic Excitations in Trilayer Nickelate Pr$_4$Ni$_3$O$_{10}$

Rare-earth substitution strongly modifies the structure, metallicity, and superconducting behavior of Ruddlesden-Popper (RP) nickelates, but its impact on their magnetic interactions remains unclear. Here, we combine x-ray absorption spectroscopy (XAS) and resonant inelastic x-ray scattering (RIXS) to resolve the multiorbital electronic structure and magnetic excitations of Pr$_4$Ni$_3$O$_{10}$. The absence of an overlapping rare-earth resonance provides direct access to the Ni $L_3$ edge, while polarization-dependent measurements at the Ni $L$- and O $K$-edges resolve distinct planar and out-of-plane Ni-O hybridization channels. Momentum-resolved RIXS reveals dispersive magnetic excitations that soften toward the incommensurate spin-density-wave wave vector. An effective spin-wave analysis yields a sizable intra-trilayer exchange coupling, comparable to the leading in-plane interactions and robust against rare-earth substitution. Our results identify the multilayer architecture, rather than rare-earth substitution alone, as a key control knob of the magnetic interactions in these materials.

cond-mat.str-el↗

arXiv2026

ME-Brain-1.0: Memory, Cognition and Action for Evolving Embodied Intelligence

Current embodied systems largely rely on pretrained capabilities that remain fixed after deployment, limiting their ability to learn from physical interaction. We introduce MachEmbodied-Brain (ME-Brain), a self-evolving embodied system organized around a closed loop of action execution, experience acquisition, experience evolution, and improved execution. Evolvable Memory consolidates multimodal trajectories into hierarchical, reusable experience; Cognitive Core transforms physical experience into transferable skills; and the Action Model combines event-driven keyframes, EventCell local-world prediction, and action-conditioned memory modulation to focus computation on decision-critical moments, regions, and historical evidence. Together, these modules shift embodied intelligence from train-and-freeze to deploy-and-evolve without model retraining. Cognitive Core outperforms the strongest comparison models by 8.2 and 9.6 points on embodied and agent benchmarks. The Action Model achieves 47.88% mean success on RoboMME, a 3.26-point improvement over the strongest baseline. On RoboDojo, it reaches a 21.51 mean Score and 16.03% success rate, exceeding $π_{0.5}$ by 10.10 and 9.12 points. On the six-task ME-RealBench, ME-Brain achieves a 69.5 mean Score and 66.7% success rate, outperforming DM0.5 by 12.8 and 11.7 points, respectively.

cs.RO↗

arXiv2026

Claw-SWE-Bench: A Benchmark for Evaluating OpenClaw-style Agent Harnesses on Coding Tasks

The software engineering capabilities of general-purpose agent harnesses remain underexplored, and existing benchmarks offer limited support for comparing these harnesses under consistent conditions. To address this gap, we introduce Claw-SWE-Bench, a unified benchmark that enables researchers to systematically assess the capabilities and efficiency of general-purpose harnesses on software engineering tasks. The benchmark contains 350 real-world GitHub issue-resolution instances across eight programming languages and 43 repositories and provides a shared adapter protocol to align task inputs, outputs, and execution environments across harnesses. Experiments show that general-purpose harnesses can effectively resolve real-world software issues and that their success rates and resource consumption vary substantially even when the underlying model is held fixed. To lower evaluation costs and support faster debugging and iteration, we also provide Claw-SWE-Bench Lite, an 80-instance subset designed to preserve the key evaluation properties of the full benchmark. We hope this benchmark will help researchers better evaluate and understand the performance of general-purpose harnesses on software engineering tasks and guide the development of more capable and efficient harnesses. The data is available at https://github.com/opensquilla/claw-swe-bench and https://huggingface.co/datasets/TokenRhythm/Claw-SWE-Bench.

cs.LG↗

arXiv2026

Delegation without Priors

A principal relies on information held by a biased agent in decision-making. Without a prior over the circumstances that may arise, she designs a delegation rule to minimize the worst-case regret. Optimal delegation combines a default action, a fixed gap above it, and a higher discretionary interval with random boundaries. This structure separates two margins: whether the decision departs from the default and how much discretion the agent receives conditional on doing so. The default and gap limit excessive adjustment when little adaptation is warranted, while the higher interval preserves flexibility when larger adjustments are needed. Randomizing the discretionary interval's floor and ceiling hedges the principal's exposure across potential worst-case states. This rule is optimal among all stochastic incentive-compatible mechanisms.

econ.TH↗

arXiv2026

Mixture-of-Top-k Attention: Efficient Attention via Scalable Fast Weights

The vanilla self-attention mechanism in Transformers can be viewed as a two-layer fast-weight MLP, whose weights are dynamically induced by inputs and whose hidden dimension is equal to the sequence length $N$. As the context extends, the expressive capacity of such an $N$-width MLP increases, but it becomes unscalable for extremely long sequences. Recently, this fast-weight perspective has motivated the Mixture-of-Experts (MoE) attention mechanism, which partitions the sequence into rigid blocks, treats them as fast-weight experts, and sparsely routes the tokens to them. In this paper, we elevate this perspective to a unifying framework for efficient attention mechanisms, interpreting them as making fast weights scalable through either routing or compression, and organizing them into a five-dimensional taxonomy. Then, we propose Mixture-of-Top-$k$ Attention (MiTA), which employs a small set of landmark queries to gather top-$k$ attended key-value pairs as query-aware and deformable routed experts, while compressing the $N$-width MLP into a narrower shared expert. Consequently, our MiTA improves the flexibility of prior MoE attention from rigid to deformable fast-weight experts, as well as the scalability of prior top-$k$ attention from query-specific set to reusable top-$k$ set. We conduct extensive experiments on vision tasks showing the superior effectiveness and efficiency of our MiTA, and also uncovering intriguing properties such as an emergent token-pruning effect and easy generalization from standard attention. Code is available at https://github.com/QishuaiWen/MiTA.

cs.LG↗

arXiv2026

MWE-ECL: Recoverable Long-Range Context Does Not Always Override Local Lexical Priors

Long-context evaluations often test whether a model can recover distant evidence, but recoverability does not guarantee behavioral influence. We test the prediction that a distant discourse anchor can remain explicitly recoverable yet fail to change the locally preferred reading of a familiar multiword expression; such failures should concentrate when the model's no-anchor default conflicts with the anchor, while prior-correct decisions remain largely preserved. We introduce Multiword Expression Effective Context Length (MWE-ECL), a bilingual diagnostic whose matched anchor-retrieval, no-anchor prior, and interpretation prompts measure explicit recoverability, model-observed defaults, and anchor-conditioned decisions, respectively. Across eight English deployment panels on a shared 0-128K grid, retrieval-control accuracy on prior-conflict items is 0.989-1.000, prior-conflict override spans 0.806-1.000 (0.809-1.000 after conditioning on correct retrieval), and preservation of prior-correct decisions remains 0.977-1.000. A same-call control querying retrieval and interpretation in one prompt reproduces the gap for DeepSeek V4 Pro (1.000 retrieval versus 0.900-0.920 interpretation), showing that separate invocations are not its sole explanation; smaller or absent gaps in the other two models bound its generality. For DeepSeek V4 Flash, separate prompt-fit tests retain perfect retrieval with lower interpretation at 512K and 1M, while foil-consistent cues shift the no-anchor prior far more than retrieval; cross-model cue effects are heterogeneous. A separately reported 10-family Chinese subset shows similar descriptive gaps, but imperfect retrieval for some models prevents an integration-only attribution. MWE-ECL therefore evaluates whether explicitly recoverable distant context changes a competing local semantic decision.

cs.CL↗

arXiv2026

Benchmarking StreamNTT with a Verilog-to-Routing Toolchain

As post-quantum cryptography algorithms move toward large-scale data center deployment, hardware acceleration of their computational bottleneck, which is the number theoretic transform (NTT), has gained increasing attention. StreamNTT, a high-level synthesis- and field-programmable gate array-based accelerator, achieves state-of-the-art throughput through various optimization techniques. However, its reliance on a commercial tool and a device makes direct comparisons difficult for researchers without access. We address this by building StreamNTT on an open-source Verilog-to-Routing toolchain, which achieves similar digital signal processing and multiplier usage. Significant differences in internal memory utilization indicate that further memory-level optimization is needed to approach commercial tool performance.

cs.AR↗

arXiv2026

Monotone Equilibrium, Admissibility and Perfection

This paper studies monotone equilibrium, admissibility, and perfection in Bayesian games. Athey (2001), McAdams (2003), and Reny (2011) prove existence of pure-strategy monotone equilibria under single-crossing and quasi-supermodularity. We construct a counterexample satisfying these assumptions but with no admissible monotone equilibrium, showing these results cannot generally be strengthened without further conditions. To address this existence problem, we introduce perfect monotone equilibrium, combining monotonicity with trembling-hand robustness. We show that stronger assumptions such as increasing differences and supermodularity ensure existence of perfect (and hence admissible) monotone equilibria. We also demonstrate the practical relevance of our results through auctions and Bertrand competitions.

econ.TH↗