Results for "allocation systems"

Attacks that infer whether specific records were in training data, or reconstruct sensitive training examples.

System design where humans validate or guide model outputs, especially for high-stakes decisions.

Methods for breaking goals into steps; can be classical (A*, STRIPS) or LLM-driven with tool calls.

A model is PAC-learnable if it can, with high probability, learn an approximately correct hypothesis from finite samples.

AI subfield dealing with understanding and generating human language, including syntax, semantics, and pragmatics.

Updating beliefs about parameters using observed evidence and prior distributions.

Stores past attention states to speed up autoregressive decoding.

Multiple agents interacting cooperatively or competitively.

Learning from data generated by a different policy.

Models evaluating and improving their own outputs.

Central catalog of deployed and experimental models.

Legal or policy requirement to explain AI decisions.

Recovering training data from gradients.

Inferring sensitive features of training data.

Neural networks that operate on graph-structured data by propagating information along edges.

Graphs containing multiple node or edge types with different semantics.

Probabilistic model for sequential data with latent states.

Aligns transcripts with audio timestamps.

Detects trigger phrases in audio streams.

Generates audio waveforms from spectrograms.

Monte Carlo method for state estimation.

Identifying abrupt changes in data generation.

Directed acyclic graph encoding causal relationships.

What would have happened under different conditions.

Using production outcomes to improve models.

System that independently pursues goals over time.

Number of steps considered in planning.

Decomposing goals into sub-tasks.

Interleaving reasoning and tool use.

Simple agent responding directly to inputs.