Skip to content
Series
15 posts

Reading Stanford CS224V

A unit-by-unit reading of one explicitly versioned Stanford CS224V offering: understanding, dialogue management, generation, evaluation, and deployment for conversational assistants.

Stanford CS224V: Renamed to Agentic AI in 2026, but What It Teaches Is Formal Methods Against Hallucination

CS224V only became Agentic AI in the 2026–2027 catalog, and the rename changed nothing underneath: the course still translates natural language into formal semantics and constrains agents with SMT solvers and knowledge graphs instead of wiring frameworks together. Seven of the eleven mandatory readings come out of the instructor's own lab. Every slide deck is public, and the course site says outright that they are deliberately incomplete.

Stanford CS224V Lecture 1: Turning Hallucinating LLMs into Dependable Assistants

Fall 2025 opens with computational thinking: reliability comes from decomposing retrieval, formal representation, verification, and generation into testable algorithms, not from one heroic prompt.

Stanford CS224V Lecture 2: STORM, Co-STORM, and Knowledge Curation

STORM uses perspective-guided questions, simulated interviews, and outlines to broaden research; Co-STORM keeps a person in the loop so discovering unknown questions and co-editing become part of the system.

Stanford CS224V Lecture 3: Building Task-Oriented Agents with Genie Worksheets

Genie Worksheets declare task capability as a form-like specification. A contextual semantic parser updates formal dialogue state while the runtime controls queries, actions, and responses.

Stanford CS224V Lecture 4: Task-Agent Evaluation Beyond Human-Like Answers

CS224V splits task-agent evaluation into state updates and complete interaction: isolate the semantic parser, then test task completion, grounded queries, and valid actions with real users.

Stanford CS224V Lecture 5: WikiChat's Seven-Stage Defense Against Hallucination

The [WikiChat paper](https://aclanthology.org/2023.findings-emnlp.157/) expands RAG into query formulation, retrieval, filtering, generation, claim extraction, renewed retrieval and verification, and removal of unsupported content—and evaluates retrieval separately from factuality.

Stanford CS224V Lecture 6: Why Database Agents Begin with Semantic Parsing

Reliable database agents map language to executable queries, resolve schemas and enumerated values, and evaluate execution separately from answer generation; hybrid questions additionally require explicit source routing.

Stanford CS224V Lecture 7: SUQL Unifies SQL and Free-Text Retrieval

SUQL adds answer and summary functions over text to SQL. A semantic parser emits one hybrid query, while an optimizing compiler applies predicate pushdown, top-k pruning, and lazy evaluation.

Stanford CS224V Lecture 8: SLIDERS Turns Long-Document Sets into Queryable Tables

SLIDERS induces a question-specific schema, applies semantic chunking and contextualized extraction, reconciles duplicate rows, and answers with SUQL instead of feeding every long document directly to one model.

Stanford CS224V Lecture 9: Why Automated Qualitative Coding Still Needs Expert Review

Automated qualitative coding defines event types and arguments in a codebook, then separates document classification, structured extraction, and entity linking. Constrained JSON fixes form, not expert judgment.

Stanford CS224V Lecture 10: How SPINACH Explores Wikidata and Builds SPARQL

SPINACH does not guess complete SPARQL in one shot. It searches entities and properties, inspects Wikidata entries and examples, executes small queries, and composes a final query under explicit action and stopping rules.

Stanford CS224V Lecture 11: Translate Trial Criteria into SMT Instead of Asking an LLM to Decide

The lecture parses patient records and trial criteria into SMT, retrieves candidates through a weaker propositional projection, and runs a solver on the reduced set. Reasoning is inspectable, but NL-to-SMT remains the main error boundary.

Stanford CS224V Lecture 12: CHURRO Makes Multilingual Historical Documents Searchable

CHURRO represents full-page text, layout, and metadata in HDML, unifies multilingual historical data for a page-level VLM, and connects extraction to HistoryGenie for searchable, conversational archives.

Stanford CS224V Lecture 13: ReactGenie Gives Voice and Native GUIs Shared State

ReactGenie annotates React components to expose data, actions, and views, parses composite voice commands into a DSL, and renders native graphical output against shared UI context.

Stanford CS224V Lecture 14: Scaling Language Models When Data Is the Bottleneck

The final lecture is not a complete LLM-training tutorial. It studies data efficiency under fixed data and abundant compute, revisiting epochs, batches, ensembles, self-training, and conditions for synthetic continued pretraining.