Frontier Physics Researcher: Golden Reference Solutions

Obsidian; Greater London; Remote; GBP 82,000 - 138,000; Full time

Job summary

Obsidian is seeking expert physics researchers to author and verify golden reference solutions for the CritPt benchmark, a frontier research-level physics task. You will produce 100%-human-verified data used to evaluate language models on frontier physics reasoning.

You may solve challenges end-to-end with clear derivations, audit others' solutions, or adjudicate between attempts, producing transparent documentation, code, and verification tests for robust evaluation.

Qualifications

• PhD or postdoc in the relevant subfield.
• Experience with two canonical methods and publications in the target subfield.
• Proficiency with LaTeX, Python, Jupyter, and SymPy.

Responsibilities

• Solve research-level physics challenges end-to-end with verifiable derivations, code, and peer-reviewed references.
• Decompose challenges into standalone checkpoint sub-problems requiring genuine physical reasoning.
• Author Python answer templates with auto-grading functions for symbolic or numerical answers.
• Audit submitted solutions for correctness, scope, and method soundness; provide actionable feedback.
• Adjudicate between parallel solver attempts and decide which solution becomes the golden reference.
• Document chain-of-thought reasoning, error tolerances, and verification test cases.

Skills

Strong written English

Analytical reasoning

Structured documentation

Education

PhD or postdoc in the relevant subfield

Postdoc or junior professor in the relevant subfield

Senior professor or industry research PI in the relevant subfield

Tools

Python

Jupyter

SymPy

LaTeX

Job description

Obsidian is seeking expert physics researchers to author and verify golden reference solutions for the CritPt benchmark, a frontier research-level physics task. You will produce 100%-human-verified data used to evaluate language models on frontier physics reasoning.

You may solve challenges end-to-end with clear derivations, audit others' solutions, or adjudicate between attempts, producing transparent documentation, code, and verification tests for robust evaluation.