Preference Dataset QA Reward Model Evaluator

AI Trainer Jobs; United States; Remote; USD 28,000 - 55,000; Part time

Benefits offered by this job

Remote work

Job summary

AuraOne is seeking a remote Preference Dataset QA Reward Model Evaluator to review prompts and responses against our quality rubric. You will compare paired outputs, label edge cases, and write structured feedback for the modeling team to retrain.

As a contractor, you work independently on a remote track eligible for US-based candidates, with hourly compensation assessed after the interview. You will produce rankings, tag issues, and help calibrate human judgment for post-training pipelines.

Qualifications

• Prior evaluation, annotation, or human-rater experience in preference dataset QA/reward model evaluation.
• Ability to apply rubrics consistently across long batches.
• Clear written reasoning that names the issue and the rubric clause being applied.
• Strong attention to detail and the ability to flag when a prompt itself is the problem.
• Reliable async availability for at least 10 hours per week.
• Experience with inter-rater agreement metrics and calibration cycles.

Responsibilities

• Evaluate outputs against a versioned rubric and assign severity tags.
• Compare paired responses and justify the stronger answer with rationale.
• Label hallucinations, instruction-following failures, and unsafe content with structured tags.
• Capture ambiguous prompts and route them back to the program team for rubric updates.
• Audit a batch for rubric consistency and report drift to the program lead.

Skills

Model output evaluation

Rubric-based annotation

Severity tagging

Inter-rater calibration

RLHF

Rater calibration

Preference dataset

Preference ranking

Job description

Preference Dataset QA Reward Model Evaluator is a remote evaluation track for reviewing preference dataset qa reward model evaluation prompts and responses against AuraOne's quality rubric. Reviewers compare paired outputs, label edge cases, and write the kind of structured feedback the modeling team can use to retrain.

Category: RLHF & Human Preference Data · Pay: Hourly rate confirmed after the interview process · Location: Remote — US-eligible · Contractor

Preference Dataset QA Reward Model Evaluator is a remote evaluation track for reviewing preference dataset qa reward model evaluation prompts and responses against AuraOne's quality rubric.

About the role

Preference Dataset QA Reward Model Evaluator is a remote evaluation track for reviewing preference dataset qa reward model evaluation prompts and responses against AuraOne's quality rubric. Reviewers compare paired outputs, label edge cases, and write the kind of structured feedback the modeling team can use to retrain.
AI data reviewers help turn preference dataset qa reward model evaluation outputs into auditable labels, rationales, and regression cases for AuraOne Human Data.
Produce preference rankings, reward-model feedback, and calibrated human judgment for post-training pipelines.

Responsibilities

• Evaluate preference dataset qa reward model evaluation model outputs against a versioned rubric and assign severity tags for Preference Dataset QA Reward Model Evaluator assignments.
• Compare paired responses and pick the stronger answer with a written rationale.
• Label hallucinations, instruction-following failures, and unsafe content with structured tags.
• Capture ambiguous prompts and route them back to the program team for rubric updates.

Role details

Track Evaluation & annotation Work model Remote · Independent specialist contractor Compensation Hourly rate confirmed after the interview process. Eligible from US

What you should bring

• Prior evaluation, annotation, or human-rater experience on preference dataset qa reward model evaluation or adjacent content for Preference Dataset QA Reward Model Evaluator work.
• Comfort applying multi-page rubrics consistently across long batches.
• Clear written reasoning that names the issue and the rubric clause being applied.
• Strong attention to detail and the ability to flag when a prompt itself is the problem.
• Reliable async availability for at least 10 hours per week.

Role signals

Example tasks

• Compare two preference dataset qa reward model evaluation model responses to the same prompt and pick the stronger one with rationale.
• Tag an unsafe response with the correct policy category and severity.
• Audit a 50-row batch for rubric consistency and report drift to the program lead.
• Propose a rubric clarification after spotting a recurring failure mode.

Useful experience

• Background in linguistics, content moderation, or trust & safety review.
• Experience with inter-rater agreement metrics and calibration cycles.
• Domain expertise that lets you spot subject-matter errors automated checks miss.

Compensation and schedule

Hourly rate confirmed after the interview process.
Expected arrangement: contractor , with program-defined task volume and review pacing. Placement depends on current program demand and reviewer confirmation.

Skills used in matching

• Model output evaluation
• Rubric-based annotation
• Severity tagging
• Inter-rater calibration
• Preference Dataset QA Reward Model evaluation
• Preference ranking
• RLHF
• Rater calibration
• Preference
• Dataset