Mathematics LLM Evaluation Expert (Master's/Ph.D.)

You will help evaluate how well language models reason through advanced mathematics. The work includes creating challenging problems, solving them carefully, reviewing model answers, and explaining errors or missing arguments.

You will contribute to benchmarks from early undergraduate through doctoral-level mathematics. You will also write Python solutions with approved scientific libraries and translate problems and proofs into Lean to check that formal proofs compile.

• Design original problems involving multi-step, abstract, and proof-based reasoning.
• Solve mathematics problems independently with clear, logically structured explanations.
• Review model-generated solutions and identify mathematical errors, gaps, and incorrect reasoning.
• Provide precise feedback, annotations, corrections, and constructive explanations.
• Develop and validate Python solutions using approved scientific libraries.
• Translate mathematical problems and proofs into Lean and verify that the formal proofs compile.
• Help build evaluation benchmarks across a wide range of mathematics topics.

This is a fully remote contractor assignment open worldwide. The work suits someone who can explain difficult mathematics clearly, follow detailed project guidelines, and work independently with close attention to detail.

• At least 20 hours per week.
• Available commitment options of 20, 30, or 40 hours per week.
• At least 4 hours of work per day.
• At least 4 hours of overlap with Pacific Time.
• Relevant academic or professional expertise in mathematics, applied mathematics, statistics, or a related field.
• Strong foundations in areas such as algebra, calculus, analysis, geometry, or topology.
• Fluent in English with strong written communication skills.
• Python and scientific library experience is valuable.
• Familiarity with Lean or formal proof verification is beneficial.

Apply on OpenTrain with your resume and then complete the application on the hiring site.

AI training is the human work behind systems that learn from examples, including writing, reviewing, and rating model responses. OpenTrain helps people find AI training and data-labeling projects, build a profile, and apply through one account.