Patterns

TaskDescription -> SourceCode

Generate an initial runnable scaffolding file tailored to a structured task description.

SourceCode, ExecutionLogs, BenchmarkMetrics -> ImprovedSourceCode

Refine task agent source code iteratively by analyzing previous execution logs, diff histories, and execution error metrics.

RepositoryMetadata -> LLMToolPrompt

Extract function metadata from dynamic third-party package registries to compile structural prompt instructions for LLM tool selection.

Embedding & NoisyTrajectory -> ActionTrajectory

Denoise a continuous robot action trajectory sequence using a flow-matching diffusion transformer conditioned on vision-language embeddings.

Prompt -> List<AdversarialPrompt>

Synthesize variations of test prompts using semantic manipulation operators to test LLM safety guardrails.

GroupCompletions -> RelativeRewards

Compute relative policy advantages from group-level rewards across multiple outputs generated for a single prompt, eliminating the memory and compute overhead of a separate critic model.

Table -> Table

Synthesize tabular data by modeling joint distributions using Gaussian copulas on real data marginals.

HighRankLayer -> LowRankDecomposedLayers

Decompose a single high-rank weight tensor into two smaller low-rank tensors using Singular Value Decomposition.

ModelTensors -> OrthogonallyRotatedModelTensors

Apply random or structured orthogonal rotations to weights and activations to distribute outlier values uniformly before quantization.

GraphWithBatchNorm -> GraphWithFoldedBatchNorm

Fold batch normalization parameters directly into the weights and biases of preceding or succeeding convolution/linear layers.

UnquantizedWeights, CalibrationData -> QuantizedWeights

Optimize individual weight rounding decisions (up or down) using localized reconstruction loss on calibration data.

NeuralNetworkModel -> NeuralNetworkModel

Rescale weights across consecutive layers to reduce range imbalance before model quantization.

ModelMetadata + HardwareSpecs -> InferenceEngineConfiguration

Select and configure the optimal inference backend engine configuration based on target hardware specifications and model parameters.

BindingPocket -> VolumetricDescriptorGrid

Map a 3D grid over localized pocket cavities and compute spatial pharmacophore indicators and electrostatic potentials for each grid node.

HeuristicPockets -> RescoredPockets

Rescore and filter binding pocket candidates generated by a fast external geometric finder using a local-feature machine learning classifier.

ProteinStructure -> FeatureVector

Selectively drop or replace crystallographic B-factor features when processing computationally predicted structures (e.g., AlphaFold) to prevent misinterpretation of pLDDT confidence scores as thermal mobility.

ProteinStructure -> RankedList<BindingPocket>

Project points onto a protein's solvent-accessible surface, score them for ligandability using a machine learning classifier, and cluster high-scoring points to delineate binding pocket boundaries.

Tensor<RoutingLogits> -> Tensor<LossScalar>

Calculate a routing loss using auxiliary load-balancing and z-loss regularization to stabilize MoE training.

Tuple<Sequence<Nucleotide>, Document<ClinicalText>> -> Tensor<ClassificationLogits>

Combine features extracted from a genomic foundation model and a medical text encoder to perform clinical prediction.

ImagePatch -> RoutedFeatures

Route scientific image patches to specialized neural experts using a learned routing gating network.

PointCloud -> FeatureMap

Map multi-modal inputs like 3D point clouds to a shared latent space of a pre-trained visual model to leverage 2D semantic priors.

Image -> AnomalyMap

Segment anomalies in an image by computing the spatial similarity between dense visual embeddings and contrasting textual prompts representing normal vs anomalous states.