Searcharxiv⌕ Search

arXiv subjects

Jiqiang Li

Publications and source records attributed to Jiqiang Li.

2 recordsLinked to original sources

EvoSkill-S1: Evolving Reusable Skills with Genetic-Inspired Operators

Reusable skills allow language-model agents to accumulate and reuse problem-solving experience. However, most existing methods improve a skill bank only indirectly through rollout-driven accumulation and heuristic reflection, often resulting in shallow and weakly controlled updates. We present EvoSkill-S1, an RL-driven framework that instead treats reusable skills themselves as the direct objects of evolution. Given hard-negative failures, EvoSkill-S1 selects parent skills and transforms them into candidate child skills through five genetic-inspired operators: revision, cross-revision, mutation, dual-reflection, and crossover. Downstream task rewards train the skill evolver to generate more effective child skills, while evaluator-guided write-back selectively retains beneficial updates. Complementary Skill Bank Fusion consolidates operator-specific banks to exploit their distinct recovery patterns, while a Procedural Skill Interface supports transfer to unseen problems. Across all reported math and code benchmarks, EvoSkill-S1 improves the corresponding initialized skill banks under both Qwen and GPT-family solvers, with relative gains of up to 12.39%. Compared with the strongest external baseline without direct skill evolution, the largest relative improvement reaches 20.39%.

cs.AI↗

A new multiscale modeling approach to unravel the influence of interlayer sp3 bonds on the nonlinear large-deformation and fracture behaviors of 2D carbon nanostructures under tension

To delve deeply into the nonlinear large-deformation and fracture behaviors of 2D carbon nanostructures (2D CNs), including bilayer graphene, diamane, and their transitional structures, this paper introduces a multiscale auxiliary nodes (MAN) method rooted in atomic structures and potentials. This approach simulates 2D CNs by constructing two virtual continuum sheets with high-order continuity. The moving least squares (MLS) approximation is employed to facilitate the transformation between atomic displacements and nodal displacements, thereby converting atomic potential energy into strain energy within the continuum model. Through iterative solutions of nonlinear stiffness equations, the equilibrium configuration of the system under specified loading conditions can be obtained. The flexibility in the density and arrangement of nodes allows for a smooth and seamless cross-scale transition from discrete atomic structures to a continuum model. Numerical simulations demonstrate that MAN method accurately predicts the nonlinear large-deformation and fracture behaviors of 2D CNs. The Young's modulus and shear modulus of diamane in both zigzag and armchair directions closely approach those of diamond and are notably higher than those of graphene. Furthermore, the quantity and distribution of interlayer sp3 bonds significantly influence the fracture behavior of 2D CNs, with strategic placement of these bonds effectively enhancing the tensile strength of the structures.

cond-mat.mtrl-sci↗