{ "model_id": "MatterWaveAI/grt-30m-chat", "architecture": "GRTForCausalLM", "parameters": 29305498, "nn_linear_layers": 0, "chat_template": "ChatML (<|im_start|> / <|im_end|>)", "before_sft": { "val_loss": 5.2083, "val_ppl": 182.7854, "assistant_token_acc": 0.2469, "assistant_response_margin": 0.4397, "multiturn_val_loss": 5.348, "multiturn_val_ppl": 210.1897 }, "after_sft": { "val_loss": 0.2904, "val_ppl": 1.337, "pair_adam_100step_val_loss": 0.3129, "pair_adam_100step_val_ppl": 1.3674, "assistant_token_acc": 0.9988, "assistant_response_margin": 1.0754, "multiturn_val_loss": 1.7855, "multiturn_val_ppl": 5.9628, "samples": [ { "user": "Can you explain this briefly? What is GRT-30M?", "target": "GRT-30M is a 29.3M-parameter Givens Rotational Transformer developed by MatterWave AI that replaces every dense nn.Linear matrix with cascaded 2D Givens rotations and butterfly permutations.", "generated": "GRT-30M is a 29.3M-parameter Givens Rotational Transformer developed by MatterWave AI that replaces every dense nn.Linear matrix with cascaded 2D Givens rotations and butterfly permutations." }, { "user": "Can you explain this briefly? How does a Givens Rotational Transformer replace dense linear layers?", "target": "Instead of O(D^2) dense matrix multiplications, GRT represents hidden states as planar coordinate pairs and transforms them through butterfly-permuted 2D Givens rotations modulated by Clifford algebra features.", "generated": "Instead of O(D^2) dense matrix multiplications, GRT represents hidden states as planar coordinate pairs and transforms them through butterfly-permuted 2D Givens rotations modulated by Clifford algebra features." }, { "user": "Can you explain this briefly? What is PearlNorm in the GRT architecture?", "target": "PearlNorm is a pair-aware RMS normalization layer that divides each planar (x, y) pair by the root-mean-square of pair radii, rescaling magnitudes while preserving every pair's phase angle.", "generated": "PearlNorm is a pair-aware RMS normalization layer that divides each planar (x, y) pair by the root-mean-square of pair radii, rescaling magnitudes while preserving every pair's phase angle." }, { "user": "Can you explain this briefly? What is RadialGELU?", "target": "RadialGELU applies a GELU activation to the Euclidean radius r of each 2D coordinate pair as gelu(r) / r, introducing nonlinearity on pair magnitudes without distorting phase directions.", "generated": "RadialGELU applies a GELU activation to the Euclidean radius r of each 2D coordinate pair as gelu(r) / r, introducing nonlinearity on pair magnitudes without distorting phase directions." }, { "user": "Can you explain this briefly? How are token embeddings parameterized in GRT?", "target": "Each token in GRTEmbedding is stored as hidden_size // 2 phase angles plus a scalar log-magnitude, emitting exp(log_magnitude) * (cos(theta), sin(theta)) on the hypersphere.", "generated": "Each token in GRTEmbedding is stored as hidden_size // 2 phase angles plus a scalar log-magnitude, emitting exp(log_magnitude) * (cos(theta), sin(theta)) on the hypersphere." } ] }, "optimizer_ablations_8way": { "hessian_pair_adam": { "variant": "HessianPairAdam (Sophia-Torus Cauchy-Riemann Curvature)", "final_train_loss": 0.9501, "val_loss": 0.2904, "val_ppl": 1.337, "assistant_token_acc": 0.9988, "assistant_response_margin": 1.0754, "assistant_response_seq_margin": 32.9186, "multiturn_val_loss": 1.7855, "multiturn_val_ppl": 5.9628, "multiturn_token_acc": 0.8026, "step100_val_loss": 0.3522, "step100_val_ppl": 1.4222, "elapsed_s": 49.51 }, "cyclide_adam": { "variant": "CyclideAdam (Dupin Cyclide Conic Metric + Deviatoric Twist)", "final_train_loss": 0.9506, "val_loss": 0.2932, "val_ppl": 1.3406, "assistant_token_acc": 0.9988, "assistant_response_margin": 0.9446, "assistant_response_seq_margin": 28.0276, "multiturn_val_loss": 1.9591, "multiturn_val_ppl": 7.0933, "multiturn_token_acc": 0.7487, "step100_val_loss": 0.3522, "step100_val_ppl": 1.4222, "elapsed_s": 50.19 }, "pair_adam": { "variant": "PairAdam (Factored Toroidal Baseline)", "final_train_loss": 0.9505, "val_loss": 0.2971, "val_ppl": 1.346, "assistant_token_acc": 0.9988, "assistant_response_margin": 1.0982, "assistant_response_seq_margin": 32.9333, "multiturn_val_loss": 2.0593, "multiturn_val_ppl": 7.8407, "multiturn_token_acc": 0.7385, "step100_val_loss": 0.3639, "step100_val_ppl": 1.439, "elapsed_s": 48.57 }, "huygens_adam": { "variant": "HuygensAdam (Circular Order Parameter R_p Splay Torque)", "final_train_loss": 0.9498, "val_loss": 0.3107, "val_ppl": 1.3644, "assistant_token_acc": 1.0, "assistant_response_margin": 1.0257, "assistant_response_seq_margin": 30.9379, "multiturn_val_loss": 1.6645, "multiturn_val_ppl": 5.2829, "multiturn_token_acc": 0.8154, "step100_val_loss": 0.3784, "step100_val_ppl": 1.4599, "elapsed_s": 67.79 }, "huygens_symplectic_adam": { "variant": "HuygensSymplecticAdam (Two-Stage Huygens -> Symplectic Curriculum)", "final_train_loss": 0.9502, "val_loss": 0.3361, "val_ppl": 1.3995, "assistant_token_acc": 0.9988, "assistant_response_margin": 1.0957, "assistant_response_seq_margin": 31.8603, "multiturn_val_loss": 1.7592, "multiturn_val_ppl": 5.8075, "multiturn_token_acc": 0.8051, "step100_val_loss": 0.398, "step100_val_ppl": 1.4889, "elapsed_s": 65.69 }, "symplectic_adam": { "variant": "SymplecticAdam (Gromov-Liouville Non-Squeezing Floor + Casimir Cavity)", "final_train_loss": 0.9502, "val_loss": 0.3441, "val_ppl": 1.4108, "assistant_token_acc": 0.9938, "assistant_response_margin": 1.1464, "assistant_response_seq_margin": 34.5992, "multiturn_val_loss": 1.9487, "multiturn_val_ppl": 7.0199, "multiturn_token_acc": 0.7513, "step100_val_loss": 0.3961, "step100_val_ppl": 1.4859, "elapsed_s": 50.95 }, "gyric_adam": { "variant": "GyricAdam (Poinsot-Lande Gyroscopic Projection + Thomas Precession)", "final_train_loss": 0.9503, "val_loss": 0.3476, "val_ppl": 1.4157, "assistant_token_acc": 0.9975, "assistant_response_margin": 1.0779, "assistant_response_seq_margin": 32.9217, "multiturn_val_loss": 1.8009, "multiturn_val_ppl": 6.0552, "multiturn_token_acc": 0.7718, "step100_val_loss": 0.4194, "step100_val_ppl": 1.5211, "elapsed_s": 70.5 }, "polar_natural_adam": { "variant": "PolarNaturalAdam (Riemannian Polar Metric + Cauchy-Riemann Hessian)", "final_train_loss": 0.9521, "val_loss": 0.3883, "val_ppl": 1.4745, "assistant_token_acc": 0.9913, "assistant_response_margin": 0.9159, "assistant_response_seq_margin": 27.5035, "multiturn_val_loss": 1.8642, "multiturn_val_ppl": 6.4505, "multiturn_token_acc": 0.7692, "step100_val_loss": 0.4398, "step100_val_ppl": 1.5524, "elapsed_s": 49.4 } } }