Patterns

(LLMClaim, List<RawOutput>) -> VerifiedClaim

Filter claimed LLM milestones by checking if the exact substring exists within raw tool execution logs.

(Set<Fact>, Goal) -> Intent

Plan exploration steps by checking known objective facts against a target goal to yield unexecuted intents, eliminating redundant tool executions.

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.

Prompt + Schema -> ParsedObject

Enforce a validation schema on model inference to guarantee structured JSON outputs.

GeneratorContext -> VerifiedOutput | ReproducibleFailure

Execute a candidate generation, then evaluate it using a verifier agent that is kept completely blind to the generation agent's context and history.

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.

UserFeedback + BasePrompt -> AugmentedPrompt

Augment prompt payloads dynamically by aggregating preceding interaction feedback into the agent's context window.

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.

WebpageDOM -> AccessibilityTree<ReferenceID>

Extract a webpage's accessibility tree, mapping interactive elements to short reference IDs for AI agent consumption.

ResearchPrompt -> ComprehensiveReport

Execute a multi-step loop of web searching, link crawling, and content extraction to synthesize deep reports.

MalformedToolCall -> ToolCallResult

Catch and automatically repair malformed tool call arguments generated by an LLM before execution.

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.

RawInputData + NaturalLanguageInstruction -> TransformedData

Synthesize and execute dynamic data-transformation operations on a payload using an LLM matched with natural language instructions.