△ Mechanism confirmed, baseline not beaten 2026

Polyconvex rotation-frame Jacobian loss

Implementation & benchmark of arXiv:2608.23072 — Polyconvexity for Cosserat nonlinear elasticity and nonlinear couple-stress theory

Usefulness6/10

Difficulty6/10

Novelty6/10

Regularization Stability Accuracy Generalization Mlp Regularization Geometry Training dynamics Convex analysis Differential geometry Lie groups Linear algebra Pde

Source paper: Polyconvexity for Cosserat nonlinear elasticity and nonlinear couple-stress theory arXiv:2608.23072 ⓘ · analyzed Aug 29, 2026

AI-generated research hypothesis, automatically tested. Not peer-reviewed.

Idea description

Replace ordinary Jacobian penalties in coordinate MLPs or deformation networks with a learned local rotation frame and a polyconvex energy of the relative stretch. Penalize $U$, its cofactor, and its determinant through a convex function, while separately smoothing the rotation field through $R^T\operatorname{Curl}R$. The intended benefit is resistance to fold formation and better conditioning than directly penalizing $\|J-I\|^2$, especially for large deformations.

Formulas

$$\overline U=\overline R^{T}{\rm D}\varphi,\qquad \overline R\in {\rm SO}(3)$$

$$\operatorname{Cof}(\overline R^{T}{\rm D}\varphi)=\overline R^{T}\operatorname{Cof}{\rm D}\varphi,\qquad \det(\overline R^{T}{\rm D}\varphi)=\det({\rm D}\varphi)$$

$$K=\overline R^{T}\operatorname{Curl}\overline R$$

$$\mathcal L_{\mathrm{pc}}=\frac{1}{B}\sum_{b=1}^{B}\left[W_\theta\!\left(U_b,\operatorname{Cof}U_b,d_b\right)+\lambda_K\|K_b\|_F^2+\lambda_{\det}\,\psi(d_b)\right],\quad U_b=R_b^TJ_b,\ d_b=\det J_b,\ \psi(d)=-\log(d+\epsilon)$$

Mathematical statement

The paper lifts the deformation gradient $D\varphi(x)\in\mathbb R^{3\times3}$ using an independent rotation $\overline R(x)\in\mathrm{SO}(3)$ and relative stretch $\overline U(x)=\overline R(x)^T D\varphi(x)$. The key identities are $\operatorname{Cof}(\overline R^T D\varphi)=\overline R^T\operatorname{Cof}(D\varphi)$ and $\det(\overline R^T D\varphi)=\det(D\varphi)$, since $\det\overline R=1$. A polyconvex energy has the form $W(\overline U,\operatorname{Cof}\overline U,\det\overline U)$, where $W$ is convex in its three displayed arguments and grows coercively in them. The curvature variable $K=\overline R^T\operatorname{Curl}\overline R$ controls first-order variation of the rotation field. In the constrained model, $\overline U\in\operatorname{Sym}^+(3)$ together with $\det D\varphi>0$ forces $\overline R=\operatorname{polar}(D\varphi)$. For neural use, $J=D\varphi$, $R$ is a predicted or differentiably estimated rotation, $U=R^TJ$, and the loss is convex in $(U,\operatorname{Cof}U,\det U)$, with a barrier for positive determinant.

Implementation notes

(1) Integrate this into a coordinate MLP that predicts a deformation $\varphi_\theta(x)\in\mathbb R^3$, either for an implicit shape deformation or for a neural warp used in registration. At sampled spatial points $x_b$, compute the Jacobian $J_b=\partial\varphi_\theta/\partial x$ with forward-mode autodiff. Add a second head producing a 9-vector reshaped into a matrix $A_b$, and project it to $R_b\in\mathrm{SO}(3)$ by SVD polar projection $R_b=UV^T$, flipping the last column of $V$ if $\det(UV^T)<0$. Compute $U_b=R_b^TJ_b$, $C_b=\operatorname{Cof}(U_b)$, and $d_b=\det(J_b)$. (2) Use this pseudocode: J=jacobian(phi(x)); A=rotation_head(x); R=svd_polar(A); U=transpose(R)@J; C=cofactor(U); d=det(J); loss_data=task_loss(phi(x)); loss_pc=convex_energy(U,C,d); loss_det=-lambda_det*log(d+eps); K=R^T@curl(R); loss=loss_data+lambda_pc*loss_pc+lambda_K*mean(K*K)+loss_det. Estimate $K$ with centered finite differences of the rotation head at neighboring points; if the application has no spatial coordinates, omit this term. (3) Implement $W_\theta$ as a small input-convex neural network whose input is the concatenated entries of $(U,C,d)$, with nonnegative hidden-to-hidden weights, or initially use explicit $a\|U-I\|_F^2+b\|C-I\|_F^2+c(d-1)^2$. The paper supplies the structural variables and rotation-frame identities; autodiff estimates $J$, finite differences estimate $\operatorname{Curl}R$, and the convex-energy parameters are learned jointly. Log the fraction of points with $d\le0$, the Jacobian condition number, and the task loss. (4) First experiment: train a 4-layer SIREN or ReLU coordinate MLP on a synthetic 3D registration task involving a twist plus radial compression, comparing ordinary $\|J-I\|_F^2$, determinant-only barriers, and this loss at matched parameter count and gradient evaluations. Success is a lower fold rate and lower Jacobian condition number at equal reconstruction error, or faster convergence to the same error. Also test robustness under twice the deformation amplitude; the hypothesis is that the polyconvex minor terms retain accuracy where the baseline develops folds.

Verification

Mechanism confirmed, baseline not beaten

Stage 1 · Toy mechanism gate: Passed ✓

Stage 2 · Mechanism transferred to benchmark: Confirmed ✓

Stage 2 · Practical benchmark result: Significantly worse ✗

Methodology: 8 paired seeds, permutation test (p<0.05) against a learning-rate-tuned baseline. How verification works

Stage 1 — Mechanism check agent confidence 7/10

Built a reproducible PyTorch MVP with SO(3) rotation frames, autodiff Jacobians, stable cofactor computation, determinant barrier, and lifted polyconvex energy. The rotation/cofactor/determinant identities were verified to about 1.4e-6 absolute error. In the toy deformation experiment both methods had zero folds, while the proposed loss improved mean Jacobian conditioning and minimum determinant at both amplitudes; it slightly improved task MSE at amplitude 1 and was nearly equal at amplitude 2. This is a promising conditioning/stability signal, but not definitive evidence of lower fold rate because the tiny task produced no folds for either method.

Agent confidence: 7/10

Baseline: amp1: MSE 0.005764, fold fraction 0, mean condition 2.553, min det 0.241; amp2: MSE 0.019627, fold fraction 0, mean condition 4.863, min det 0.0674
Idea: amp1: MSE 0.004731, fold fraction 0, mean condition 2.380, min det 0.568; amp2: MSE 0.020299, fold fraction 0, mean condition 3.796, min det 0.177

Limitations: Only a small fixed-collocation synthetic warp and one random seed were tested. Rotation smoothness K=R^T Curl R, learned convex energy parameters, larger networks, parameter/FLOP matching, and genuinely folding or noisy registration tasks were not tested; the implementation uses an exponential-map SO(3) head rather than SVD polar projection for numerical stability.

How to run: python3 experiment.py

Stage 2 — Benchmark vs. tuned baseline (arXiv:2608.23072) Sep 3, 2026

Significantly WORSE than the tuned baseline (n/a, p=0.0321, wins 1 / 8 paired seeds).

Benchmark: Deformation_registration
Model: coordinate_warp
Paired seeds: 8
Baseline mean: 0
Idea mean: 0
Effect (Δ): 0.0074 (n/a; negative = idea better)
Wins: 1 / 8 paired seeds
p-value: 0.0321 (permutation test, 20 000 shuffles)
Smallest detectable effect: ±5.6%
Mechanism: Confirmed ✓
Practical effect: Hurts
Baseline tuning: swept over 0 configs

Limitations:

The registered track is a synthetic deformation-registration task rather than a real clinical or graphics dataset. Rotation curl smoothing, learned input-convex energy parameters, larger models, and longer training were not tested. CUDA encountered a native illegal-instruction crash, so the completed protocol ran on CPU.

How to run: python3 run_bench.py

Verdict computed by deterministic test code from paired-seed statistics — not by the language model.

Stage 2 — Benchmark Sep 3, 2026 latest Failed ✗

Agent confidence: 9/10

The registered deformation_registration benchmark confirmed the predicted stability mechanism: the polyconvex rotation-frame loss reduced mean Jacobian condition number from 9.6255 to 3.6457, reduced fold fraction from 5.5% to 0%, and improved mean minimum determinant from -0.0314 to 0.1291. However, test MSE worsened by 0.00736 with permutation p=0.0321, yielding the required verdict idea worse (significant), so this is not a benchmark win.

Baseline mean test mse: 0.136754
Delta mean: 0.007355469278991222
Idea mean test mse: 0.144109
Math check: map[cofactor_max_abs:7.152557373046875e-07 det_max_abs:4.76837158203125e-07 rotation_det_max_abs:2.384185791015625e-07]
Mechanism signature: map[baseline_fold_fraction:0.05500000016763806 baseline_mean_condition:9.625483721494675 baseline_mean_min_det:-0.031383185618324205 confirmed:true idea_fold_fraction:0 idea_mean_condition:3.6457439363002777 idea_mean_min_det:0.1291193552315235]
Permutation p value: 0.0321

Benchmark result