Advanced LLM Agents MOOC Spring 2025 - video 03 - 1:12:12

Informal plus formal math reasoning

A complete proof often needs informal planning before formal verification.

proof sketchesformal verificationmath agents
Informal+Formal MathReasoning by Sean Welleck

Problem-first learning

The problem this lecture is trying to solve

A complete proof often needs informal planning before formal verification.

Lowest-level failure mode

The agent must maintain correspondence between sketch steps and formal proof states.

Frontier update

The frontier is hybrid: informal reasoning for strategy, formal verification for correctness.

Transcript-grounded route

How the lecture unfolds

This is built from 1,273 caption segments. Use the timestamp buttons to jump into the original video when a term feels fuzzy.

0:00-12:03

Pass 1: That

The lecture segment repeatedly returns to that, math, actually, proof, what. Treat this part as the board-work for the mechanism, not as a definition list.

Write one line that connects the terms to the central failure mode: The agent must maintain correspondence between sketch steps and formal proof states.

12:03-24:05

Pass 2: That

The lecture segment repeatedly returns to that, step, formal, proofs, next. Treat this part as the board-work for the mechanism, not as a definition list.

Write one line that connects the terms to the central failure mode: The agent must maintain correspondence between sketch steps and formal proof states.

24:08-36:07

Pass 3: That

The lecture segment repeatedly returns to that, proof, informal, formal, actually. Treat this part as the board-work for the mechanism, not as a definition list.

Write one line that connects the terms to the central failure mode: The agent must maintain correspondence between sketch steps and formal proof states.

36:07-48:10

Pass 4: That

The lecture segment repeatedly returns to that, theorem, premises, lean, what. Treat this part as the board-work for the mechanism, not as a definition list.

Write one line that connects the terms to the central failure mode: The agent must maintain correspondence between sketch steps and formal proof states.

48:10-1:00:11

Pass 5: That

The lecture segment repeatedly returns to that, proof, just, lean, actually. Treat this part as the board-work for the mechanism, not as a definition list.

Write one line that connects the terms to the central failure mode: The agent must maintain correspondence between sketch steps and formal proof states.

1:00:11-1:12:12

Pass 6: That

The lecture segment repeatedly returns to that, just, what, lean, project. Treat this part as the board-work for the mechanism, not as a definition list.

Write one line that connects the terms to the central failure mode: The agent must maintain correspondence between sketch steps and formal proof states.

Build the mental model

What you should understand after this lecture

1. Start from the bottleneck

A complete proof often needs informal planning before formal verification. The lecture is useful because it does not treat this as a naming problem. It asks what breaks at the operational level and what design pattern removes that break.

2. Name the moving parts

The recurring vocabulary in the transcript is that, proof, what, actually, just, formal. When studying, do not memorize these as separate buzzwords. Ask what state is stored, what action is chosen, what feedback is observed, and what verifier decides whether progress happened.

3. Convert the idea into an architecture

Draft-sketch-prove decomposes proof generation. Long-context theorem proving needs relevant local context. Proof optimization can be agentic: find shorter or more robust proof paths. In exam or interview answers, this becomes a four-part answer: objective, loop, control boundary, evaluation.

4. Know the failure case

The agent must maintain correspondence between sketch steps and formal proof states. If you cannot say how the proposed system fails, the explanation is still shallow. Always include the failure it prevents and the new cost it introduces.

Concept weave

Ideas to remember

  1. Draft-sketch-prove decomposes proof generation.
  2. Long-context theorem proving needs relevant local context.
  3. Proof optimization can be agentic: find shorter or more robust proof paths.

Visual model

Agent system view

Use the graph to ask where the intelligence really lives: model, memory, tools, environment, verifier, or orchestration.

Written practice

Questions that make the idea stick

Drill 1Convert an informal proof sketch to a formal plan.
  1. List lemmas.
  2. Map each lemma to library search.
  3. Define tactic sequence.
  4. Check and repair.
Drill 2Why use sketches?
  1. They reduce search space.
  2. They preserve human-level strategy.
  3. They guide formal tactics.

Written answer pattern

How to write this under pressure

ClaimInformal plus formal math reasoning solves a concrete control problem, not just a prompt-writing problem.
MechanismState the loop: observe state, choose action/tool, get feedback, update memory or plan, stop using a verifier.
Why it worksIt makes the hidden failure mode visible: The agent must maintain correspondence between sketch steps and formal proof states.
TradeoffExtra orchestration improves reliability only if evaluation, cost, and authority boundaries are explicit.

Build skill

How to apply this in your own agent

  1. Write the concrete task and the failure mode before choosing any framework.
  2. Choose the smallest architecture that handles the failure: workflow, single agent, orchestrator-worker, or evaluator loop.
  3. Define tool schemas, memory boundaries, and a success checker.
  4. Run a small eval set with failure labels, cost, latency, and trace review.

Source route

Original course links and readings

Page generated from 1,273 YouTube captions. Raw transcript files are kept out of the public site; this page publishes study notes, timestamp routes, and paraphrased explanations.