{
 "artifacts": null,
 "category": "regularization",
 "description": "Replace an unconstrained positive higher-order gating function by a factorially normalized homogeneous polynomial whose coefficients are trained toward the Lorentzian Hessian signature. This gives multiplicative routing or attention interactions a controlled curvature pattern instead of allowing arbitrary unstable higher-order amplification.",
 "formulas_latex": [
  "$$f=\\sum_{|\\alpha|=d}c_{\\alpha}x^{\\alpha},\\qquad x^{[\\alpha]}:=\\frac{x^{\\alpha}}{\\alpha!},\\qquad\\mathrm{N}(f):=\\sum_{|\\alpha|=d}c_{\\alpha}x^{[\\alpha]},\\qquad\\alpha!:=\\prod_{i}\\alpha_{i}!.$$",
  "$$H_{u_1,\\ldots,u_{d-2}}(x):=\\nabla_x^2\\left[(u_1\\cdot\\nabla_x)\\cdots(u_{d-2}\\cdot\\nabla_x)\\mathrm{N}(f)(x)\\right],\\qquad \\lambda_2\\left(H_{u_1,\\ldots,u_{d-2}}(x)\\right)\\le 0.$$",
  "$$\\mathcal{L}_{\\mathrm{Lor}}=\\frac{1}{B}\\sum_{b=1}^{B}\\max\\left(0,\\lambda_2(H_b)\\right)^2.$$",
  "$$c_{\\alpha}=\\operatorname{softplus}(\\theta_{\\alpha}),\\qquad g(z)=\\log\\left(\\varepsilon+\\mathrm{N}(f)(\\operatorname{softplus}(z))\\right).$$"
 ],
 "id": 3089,
 "implementation": "Integrate the module into a higher-order MoE router or low-rank attention bias. Let each token representation h produce z=softplus(Wh+b) in R^n. Store trainable logits theta_alpha for all multi-indices alpha with |alpha|=d, set c_alpha=softplus(theta_alpha), and compute the interaction score g(z)=log(epsilon+sum_alpha c_alpha z^alpha/alpha!). Add g(z) either to every expert logit through an expert-specific coefficient tensor or use it as a shared load-aware router bias. During each training step, sample B feature points z_b from the current minibatch and nonnegative directions u_{b,1},...,u_{b,d-2}; use normalized softplus random projections or fixed coordinate mixtures. Compute the directional derivative polynomial with automatic differentiation, form its n-by-n Hessian H_b, and add lambda_L times max(0,lambda_2(H_b))^2 to the task loss. For d=3, only one directional derivative is needed; for d=4, use two. The paper's mathematics supplies the factorial representation and the one-positive-eigenvalue criterion. Estimate lambda_2 with a symmetric eigensolver for n\u003c=32, or approximate it using a differentiable eigensolver/power iteration. Test n=16,d=3 in a small Switch-style MoE on CIFAR-100 or WikiText-103, comparing a standard softmax router, an unconstrained positive polynomial router, and the Lorentzian-regularized router at equal FLOPs. Track validation loss, expert-load coefficient of variation, router entropy, gradient norm spikes, and training time. Success is lower load variance and fewer instability spikes with equal or better validation loss; reject it if the curvature penalty costs over 10% training time without improving stability.",
 "math_summary": "The paper defines a homogeneous polynomial f(x)=sum_{|alpha|=d} c_alpha x^alpha with c_alpha\u003e=0, the divided-power monomial x^[alpha]=x^alpha/alpha!, and the normalized polynomial N(f)=sum_{|alpha|=d} c_alpha x^[alpha]. Here alpha=(alpha_1,...,alpha_n), |alpha|=sum_i alpha_i=d, and alpha!=product_i alpha_i!. The Lorentzian property means that after applying any d-2 directional derivatives in nonnegative directions, the resulting degree-2 polynomial has a Hessian with at most one positive eigenvalue. For sampled nonnegative directions u_1,...,u_{d-2}, define H_u(x)=nabla_x^2[(u_1 dot nabla)...(u_{d-2} dot nabla)N(f)(x)]. The neural regularizer penalizes the second-largest eigenvalue lambda_2(H_u), so the target condition is lambda_2(H_u)\u003c=0. The factorial normalization matters because it makes coefficients correspond directly to mixed intersection-volume coefficients and avoids degree-dependent multiplicity factors.",
 "math_tags": [
  "convex-analysis",
  "linear-algebra",
  "geometry"
 ],
 "ml_areas": [
  "moe-routing",
  "attention",
  "regularization"
 ],
 "paper": {
  "arxiv_id": "2609.02850",
  "arxiv_url": "https://arxiv.org/abs/2609.02850",
  "summary_what_math_gives_to_ml": "The paper gives constructive realizable-volume models for factorially normalized homogeneous polynomials, connecting their coefficients to Lorentzian Hessian signatures and discrete-convex supports. The transferable assets are explicit curvature inequalities and exchange-closed support structures, rather than the specific Schubert or Grothendieck applications. A practical neural adaptation is to constrain positive higher-order routing or interaction polynomials toward Lorentzian curvature, while using polymatroid-style exchange rules for structured pruning. These are most plausible in higher-order MoE routers, tensorized MLPs, and polynomial attention biases, not as replacements for standard dense Transformer blocks.",
  "title": "Canonical-row Chern flow on Bott--Samelson towers: realizable-volume models for Schubert, Grothendieck, and Lascoux polynomials",
  "year": "2026"
 },
 "ratings": {
  "difficulty": 6,
  "novelty": 8,
  "usefulness": 6
 },
 "solves": [
  "stability",
  "generalization",
  "accuracy"
 ],
 "title": "Lorentzian interaction router",
 "url": "https://synthcore.org/idea/3089/lorentzian-interaction-router",
 "verification": {
  "peer_reviewed": false,
  "status": "unverified",
  "status_label": "Unverified",
  "verdict_source": "deterministic test code (paired-seed permutation statistics)",
  "verification_axes": {
   "benchmark_mechanism": {
    "confirmed": null,
    "tested": false
   },
   "practical_benchmark": {
    "beats_baseline": null,
    "tested": false
   },
   "toy_mechanism_gate": {
    "confirmed": null,
    "tested": false
   }
  }
 }
}