Accurate physical simulation is fundamental to science and engineering, yet conventional numerical solvers incur high costs when handling complex geometries, varying boundary and initial conditions, and diverse physical parameters. Recent deep-learning-based methods offer faster solutions, while limited flexibility and generalization on irregular meshes still hinder their practical deployment. Here we show an efficient graph-transformer operator, named PhysGTO, for learning physical dynamics through explicit manifold embeddings in both physical and latent spaces. The method aligns heterogeneous node-level conditions, constructs sparse structure-preserving connections, and integrates lightweight local message passing with global attention to capture multiscale physical dependencies. Its design scales linearly with the number of mesh points, reducing model size and computational cost while enabling efficient inference. On a benchmark of 11 datasets covering irregular meshes, time-dependent flows, and large three-dimensional geometries, PhysGTO achieves state-of-the-art accuracy with substantially lower computational cost, showing strong flexibility, scalability, and generalization across diverse physical systems.
Peng-Wei Liu, Xingyu Ren, Pengkai Wang et al.· Communications AI & Computin...· 0 citations
Language models can propose many plausible trading factors, but an autonomous research system must also allocate its evaluation budget, verify its own evidence, and preserve how each candidate was produced. We present AgonAlpha, an architecture that searches over frozen research artifacts---hypotheses, executable expressions, platform evidence, rationales, and review status---rather than formulas alone. To our knowledge, AgonAlpha is the first alpha-mining system to combine verified artifact search, a fresh-context adversarial reviewer with re-execution and veto authority, and pending-aware parallel budget allocation, together with a complete public evidence trail. Independent deployments on WorldQuant BRAIN produced SPECTACULAR-grade alphas across five users and six model backends, with Fitness reaching 9.50 and Sharpe reaching 3.48, while retaining prompt-to-expression provenance for every submission.
Weichen Ye, Youran Sun, Xingyu Ren et al.· 0 citations
Sparse attention is widely deployed in long-context serving stacks, yet no framework audits how discarding blocks changes the influence of specific content on model output. We first establish that the phenomenon is real and causal: Block Sparse Flash Attention (BSFA) route replay across four architectures changes output decisions in 13 of 16 cells, with zero identity-replay label flips. We then introduce a dense-calibrated counterfactual audit using matched probe cards---Gold (carrying the correct answer label), Poison (carrying a target wrong label), and Benign (filler only)---under six-layout position symmetry, isolating the sparsification-specific effect. Two patterns compete. Signal concentration: the selector preserves Gold and Poison blocks far above filler-matched Benign blocks (G$\approx$P$\gg$B across all model--task pairs). Integration loss: discarding blocks severs cross-block attention---confirmed by an ablation where isolating the probe block collapses its influence from 4.48 logits to zero. Compression ratio governs the balance: a full sweep from mild ($c=0.25$) to aggressive ($c=0.75$) compression across four model--task pairs reveals that three of four cells move toward stronger sparse amplification at higher compression, with two exhibiting sign reversals. Three independent arms---BSFA route replay, controlled block-top-$k$, and KV-cache eviction---converge: sparsification changes content influence in ways aggregate accuracy cannot detect. We provide an open measurement framework deployable on any model exposing block identities.
Xingyu Ren, Youran Sun, Chugang Yi et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
