Results for "model-based"
Model-Based RL
AdvancedRL using learned or known environment models.
Model-based reinforcement learning is like having a map while exploring a new city. Instead of wandering around aimlessly, you can look at the map to plan your route and make better decisions about where to go next. In this type of learning, an AI agent first learns how the environment works—like...
A system that perceives state, selects actions, and pursues goals—often combining LLM reasoning with tools and memory.
The field of building systems that perform tasks associated with human intelligence—perception, reasoning, language, planning, and decision-making—via algori...
Decisions dependent on others’ actions.
Running new model alongside production without user impact.
Local surrogate explanation method approximating model behavior near a specific input.
Standardized documentation describing intended use, performance, limitations, data, and ethical considerations.
Inferring sensitive features of training data.
Required descriptions of model behavior and limits.
A parameterized mapping from inputs to outputs; includes architecture + learned parameters.
When a model cannot capture underlying structure, performing poorly on both training and test data.
Policies and practices for approving, monitoring, auditing, and documenting models in production.
Framework for identifying, measuring, and mitigating model risks.
Competitive advantage from proprietary models/data.
Training a smaller “student” model to mimic a larger “teacher,” often improving efficiency while retaining performance.
The learned numeric values of a model adjusted during training to minimize a loss function.
The set of tokens a model can represent; impacts efficiency, multilinguality, and handling of rare strings.
End-to-end process for model training.
Model-generated content that is fluent but unsupported by evidence or incorrect; mitigated by grounding and verification.
Applying learned patterns incorrectly.
Central catalog of deployed and experimental models.
Model trained on its own outputs degrades quality.
When a model fits noise/idiosyncrasies of training data and performs poorly on unseen data.
Separating data into training (fit), validation (tune), and test (final estimate) to avoid leakage and optimism bias.
A robust evaluation technique that trains/evaluates across multiple splits to estimate performance variability.
A high-priority instruction layer setting overarching behavior constraints for a chat model.
Studying internal mechanisms or input influence on outputs (e.g., saliency maps, SHAP, attention analysis).
Training across many devices/silos without centralizing raw data; aggregates updates, not data.
System for running consistent evaluations across tasks, versions, prompts, and model settings.
Error due to sensitivity to fluctuations in the training dataset.
Hidden behavior activated by specific triggers, causing targeted mispredictions or undesired outputs.