ArcMathMath teaching platform
For schools, tutoring centers, and math departments

school math platform for verified classwork.

Run classes, assign PDFs or library work, collect submissions, and grade with formal verification where it applies.

School workflow

The core classroom loop is now built around roles.

ArcMath is not just a contest catalog. It is the operating layer for math classes, homework, resource delivery, submissions, grading, and progress visibility.

01Create classes · assign accounts

Admin owns the org

A school or tutoring org gets one admin account. The admin creates classes, adds teachers and students, controls membership, and can inspect any account under that org.

02PDF pages · problem ranges · library sets

Teachers assign from materials

Teachers upload PDFs, select pages and problem numbers from long books, transform the source into student-ready prompts, or assign existing ArcMath problem sets.

03Structured answers · file submissions

Students submit before due time

Students see their class assignments, work in the math workspace when the problem is structured, or attach written work for PDF/manual homework.

04Manual scores · verified verdicts · CSV

Teachers grade and review

Teachers review submissions, open attachments, add scores and feedback, and use gradebook views to spot missing, overdue, and needs-grading work.

The next layer is teacher prep: concise chapter previews and difficult-problem analysis that supports instruction without handing students direct answers.

MathScout / Research Mode

A research planner for advanced math programs.

MathScout helps teachers turn open-ended exploration into a structured program: select suitable targets, pace weekly milestones, and keep formal verification separate from conjecture and experimentation.

Cohort plannerLean-aware gatesTeacher controlled
MS-1

Research targets matched to students

The planner ranks accessible theorem and exploration targets by level, topic interest, team size, and verification path.

MS-2

Weekly exploration spine

Teachers get phases, deliverables, and checkpoints so students investigate ideas without the platform simply producing final answers.

MS-3

Formal status stays honest

Experimental evidence, conjectures, proof sketches, and Lean-verified artifacts are labeled separately so research mode remains rigorous.

01 / Student input

Snap written math, convert to LaTeX, then submit.

Step photo
a2 + b2 ≥ 2ab
Tap the camera in MathLive
MathLive field
(empty - waiting for OCR)
02 / Verified grading

See exactly which engine signed the step.

STEP 1a² + b² ≥ 2ab
SymPy
symbolic: (a-b)^2 >= 0
Lean kernel
proof checked via sq_nonneg
LLM judges (2/2)
independent: confidence 0.96 / 0.96

Backends voting...

03 / Teacher-safe help

Concise nudges and prep support, not answer dumping.

Click I'm stuck and reveal one hint at a time - direction first, computation last.

1
Hint · Direction

What single algebraic identity transforms (a-b)^2 into a sum of three terms?

2
Hint · Setup

Expand (a-b)^2 = a^2 - 2ab + b^2. The inequality is the same as showing this expansion is >= 0.

3
Hint · Almost there

Any real number squared is >= 0. Apply that to (a-b)^2 and rearrange to land at a^2 + b^2 >= 2ab.

04 / Class insight

Progress data teachers can act on.

Overall accuracy
0%

18 problems graded · 2 escalated to teacher review.

Algebra
92%
Number theory
78%
Combinatorics
64%
Inequality
48%
Milestone coverage
Spotted (a-b)^2 >= 0 in step 1
Expanded correctly in step 2
Reached the SOS conclusion
Skipped justifying step 3 - review
Verified math engine

Formal verification is the grading advantage.

When work is structured, ArcMath checks student reasoning with deterministic math first. It escalates to LLM review only when symbolic or formal backends cannot decide.

01 / SymPy

Algebraic identities, equation manipulation, simple inequalities. For supported structured steps, SymPy returns a deterministic verdict without an LLM round-trip.

02 / Lean kernel

Proof-level claims can run through Lean and mathlib. When a claim type-checks, the verdict comes from the kernel rather than natural-language confidence.

03 / LLM judge

For arguments outside the formalized surface, an LLM judge can summarize or flag reasoning, with the result clearly tagged instead of mixed into a black-box score.

Built for math teams

Canvas-style class operations, with math-specific intelligence.

The platform handles the internal workflow schools actually need: rosters, materials, deadlines, submissions, grading, reports, and teacher preparation.

Class and roster control

Admins create classes, assign teacher and student accounts, and keep one controlled view of the whole organization.

Material-to-assignment flow

Teachers can assign existing ArcMath content or selected pages and problem ranges from uploaded PDFs, including long textbooks.

Teacher preparation

Preview chapters and analyze hard problems with concise key ideas for teachers, separated from student-facing answer generation.

Gradebook and reports

Teachers see completion, submissions, scores, feedback, hint usage, and verified grading traces in one place.

School workspace

Set up a school or tutoring org.

Create the admin account first. After verification, use the org dashboard to create classes, assign teachers, add students, publish materials, and review submissions.

Guest