RLHFJobs

The human feedback behind model training

RLHFJobs: what each RLHF platform asks contributors to do

Before you sign up, it helps to know what the work is. The table holds 15 pages from 12 platforms, newest read on 2026-10-22, logging which tasks each one names: judging model answers, writing prompts or rubrics, red-teaming, and whether it tests newcomers first.

Source: RLHF work by platform: the tasks, tracks and entry rules: every row, with its source

12 on the rosterread 2026-10-22The scored table

The scored table 12 options against 5 criteria, sortable and filterable
Sort by

  1. Alignerr

    Ranks, compares or rates model answers
    not stated
    Prompt, answer or rubric writing
    not stated
    Red-teaming
    describes red-teaming or probing models for weaknesses
    Some work needs no experience
    says some work needs no experience or special skills
    Sets an entry test
    not stated

  2. Clickworker

    Ranks, compares or rates model answers
    not stated
    Prompt, answer or rubric writing
    not stated
    Red-teaming
    not stated
    Some work needs no experience
    not stated
    Sets an entry test
    not stated

  3. CrowdGen

    Ranks, compares or rates model answers
    describes ranking, comparing or rating responses
    Prompt, answer or rubric writing
    not stated
    Red-teaming
    not stated
    Some work needs no experience
    asks for domain expertise, a degree or professional experience
    Sets an entry test
    asks an entry assessment, screening or qualification task

  4. DataAnnotation

    Ranks, compares or rates model answers
    describes ranking, comparing or rating responses
    Prompt, answer or rubric writing
    describes writing prompts, answers or rubrics, or rewriting responses
    Red-teaming
    not stated
    Some work needs no experience
    says some work needs no experience or special skills
    Sets an entry test
    asks an entry assessment, screening or qualification task

  5. Handshake AI

    Ranks, compares or rates model answers
    not stated
    Prompt, answer or rubric writing
    not stated
    Red-teaming
    not stated
    Some work needs no experience
    says some work needs no experience or special skills
    Sets an entry test
    not stated

  6. Mercor

    Ranks, compares or rates model answers
    not stated
    Prompt, answer or rubric writing
    not stated
    Red-teaming
    not stated
    Some work needs no experience
    not stated
    Sets an entry test
    asks an entry assessment, screening or qualification task

  7. Mindrift

    Ranks, compares or rates model answers
    describes ranking, comparing or rating responses
    Prompt, answer or rubric writing
    describes writing prompts, answers or rubrics, or rewriting responses
    Red-teaming
    not stated
    Some work needs no experience
    varies by document
    Sets an entry test
    asks an entry assessment, screening or qualification task

  8. OneForma

    Ranks, compares or rates model answers
    describes ranking, comparing or rating responses
    Prompt, answer or rubric writing
    not stated
    Red-teaming
    not stated
    Some work needs no experience
    not stated
    Sets an entry test
    not stated

  9. Outlier

    Ranks, compares or rates model answers
    describes ranking, comparing or rating responses
    Prompt, answer or rubric writing
    describes writing prompts, answers or rubrics, or rewriting responses
    Red-teaming
    not stated
    Some work needs no experience
    asks for domain expertise, a degree or professional experience
    Sets an entry test
    asks an entry assessment, screening or qualification task

  10. Prolific

    Ranks, compares or rates model answers
    describes ranking, comparing or rating responses
    Prompt, answer or rubric writing
    not stated
    Red-teaming
    not stated
    Some work needs no experience
    not stated
    Sets an entry test
    not stated

  11. RemoExperts

    Ranks, compares or rates model answers
    not stated
    Prompt, answer or rubric writing
    not stated
    Red-teaming
    not stated
    Some work needs no experience
    not stated
    Sets an entry test
    not stated

  12. Remotasks

    Ranks, compares or rates model answers
    not stated
    Prompt, answer or rubric writing
    not stated
    Red-teaming
    not stated
    Some work needs no experience
    says some work needs no experience or special skills
    Sets an entry test
    not stated

RLHF tasks by platform, quoted from each platform

  1. 6 of 12 platforms on the table describe ranking, comparing or rating model answers.(RLHF work by platform: the tasks, tracks and entry rules: every row, with its source)
RLHF platformJudges model answersWrites prompts, answers or rubricsRed-teamingFields it namesNo experience neededEntry testSource
Alignerrnot statednot stateddescribes red-teaming or probing models for weaknessesView all roles Audio Coding General STEMsays some work needs no experience or special skillsnot statedHome page (its own page, listed on this site's dataset page)
Clickworkernot statednot statednot statednot statednot statednot statedClickworker page
CrowdGendescribes ranking, comparing or rating responsesnot statednot statednot statedasks for domain expertise, a degree or professional experienceasks an entry assessment, screening or qualification taskHome page (its own page, listed on this site's dataset page)
DataAnnotationdescribes ranking, comparing or rating responsesdescribes writing prompts, answers or rubrics, or rewriting responsesnot statedCoding Generalist Law Mathematics Medicine Physics Finance Accounting Bilingual Chemistry…says some work needs no experience or special skillsasks an entry assessment, screening or qualification taskHome page (its own page, listed on this site's dataset page)
Handshake AInot statednot statednot statednot statedsays some work needs no experience or special skillsnot statedProgramme page
Mercornot statednot statednot statednot statednot statedasks an entry assessment, screening or qualification taskOpen roles page (its own page, listed on this site's dataset page)
Mindriftdescribes ranking, comparing or rating responsesdescribes writing prompts, answers or rubrics, or rewriting responsesnot statedWe accept experts across 90+ domains from STEM and law to creative writing and…varies by documentasks an entry assessment, screening or qualification taskHome page (its own page, listed on this site's dataset page)
OneFormadescribes ranking, comparing or rating responsesnot statednot statedMarketing, psychology, journalism, architecture, generalist. If you know it, we need it.not statednot statedHome page (its own page, listed on this site's dataset page)
Outlierdescribes ranking, comparing or rating responsesdescribes writing prompts, answers or rubrics, or rewriting responsesnot statedWe depend on subject matter experts in coding, STEM, languages, and more to improve AIasks for domain expertise, a degree or professional experienceasks an entry assessment, screening or qualification taskHome page (its own page, listed on this site's dataset page)
Prolificdescribes ranking, comparing or rating responsesnot statednot statednot statednot statednot statedParticipants page
RemoExpertsnot statednot statednot statedCinematographers Photoshop Experts After Effects Specialists Coding Developers STEM PhDs…not statednot statedHome page (its own page, listed on this site's dataset page)
Remotasksnot statednot statednot statednot statedsays some work needs no experience or special skillsnot statedHome page (its own page, listed on this site's dataset page)
One row per platform, with the tasks its own pages name, the fields it recruits for and its entry rules, newest read on 2026-10-22. Where two pages of one platform answer differently, the row says varies by document.Updated

Each platform's own home, FAQ and work-description pages were read through estate-fetch and stored whole (sha256 on every row). For every page, six questions were answered from its own words: ranks, compares or rates model responses, writes prompts, answers or rubrics, red-teams models, domain tracks it names, in its own words, says its general work needs no experience, asks an entry assessment. Every stated answer carries the sentence it rests on, checked against the stored text; 'not stated' means the page read does not address the question. A platform with several pages reads 'varies' where they disagree. 12 platforms, 15 pages.

Cite: “RLHF tasks by platform, quoted from each platform”, RLHFJobs, https://rlhfjobs.com/ (updated 2026-10-22).

How we read each RLHF platform

  1. Each row is a platform that pays people for AI training work, read on its own home, FAQ or work pages through our fetch service, newest read on 2026-10-22. Pages that only sell data to AI companies were set aside.
  2. Every answer is the platform's own sentence. Where a page is silent the row says not stated, and where two of a platform's pages disagree it says varies by document. Rows are listed by name, never ranked, and no platform pays for its row.

RLHFJobs records what platforms say about the RLHF work they offer. It hires nobody, charges contributors nothing and carries no referral link today; if a platform's programme later approves one, it will be marked beside that platform. A task named on a page is what the platform describes, not a promise that work is available to you.

Get RLHF task updates

By pressing “Send me RLHF task updates” you agree: You agree that we may email you when an RLHF platform changes the tasks or entry rules it describes. Privacy policy

Questions, answered directly

What do RLHF jobs actually involve?

Mostly judging what a model wrote. DataAnnotation's FAQ lists evaluating chatbot responses for accuracy, comparing AI outputs and flagging factual errors; Outlier's home page names writing a challenging prompt, creating grading rubrics and rating and ranking answers; Mindrift's welcome page asks you to compare two options and pick the better one. (RLHF task types board)

Can I get RLHF jobs with no experience?

On some platforms, for some work. 4 of the 12 platforms on the table say part of their work needs no experience: DataAnnotation says no coding experience is required for general projects, and Handshake AI offers hourly work no matter your background. Mindrift's pages split, its welcome page letting you start on All-access tasks while its main page says you need domain expertise. Outlier and CrowdGen ask for that expertise too. (RLHF task types board)

Do RLHF platforms test you before you can work?

5 of the 12 name a test, screening or qualification step. DataAnnotation's Starter Assessment can be taken only once, Outlier runs skill screenings during onboarding, Mindrift sets a technical assessment for coders, and CrowdGen says most of its projects have a qualification process. (RLHF task types board)

Which subject tracks do RLHF platforms recruit for?

Of the 12 platforms, 6 list fields by name. DataAnnotation's FAQ runs from coding and law to medicine, physics, finance and chemistry; Outlier names coding, STEM and languages; RemoExperts lists cinematographers and literature writers beside software engineers. (RLHF task types board)

Is red-teaming part of RLHF work?

Only 1 of the 12 platforms we read describes it: Alignerr lists an AI Red Team Tester role whose job is to probe cutting-edge models for weaknesses. The rest of the table does not mention adversarial testing on the pages we read. (RLHF task types board)

Why is there no pay column on this table?

This site keeps to the tasks themselves: what you would be asked to do, in which field, and what stands between sign-up and the first task. A rate is a different question with its own sources, so it is not mixed into this table. (RLHF work by platform: the tasks, tracks and entry rules: every row, with its source)

Go deeper

Each one cites where its numbers come from.

Sources

This site's own dataset, every figure quoted from its source and dated:

  • RLHF jobs online (12 figures): 6 of 12 platforms whose own pages we read describe ranking, comparing or rating AI model responses, the core of RLHF work, read 22 October 2026.
  1. dataannotation.tech: DataAnnotation | Future-Proof Your Career With AI Training Work (its own page, listed on this site's dataset page)
  2. dataannotation.tech: FAQs | DataAnnotation
  3. outlier.ai: Train the Next Generation of AI as a Freelancer | Outlier AI (its own page, listed on this site's dataset page)
  4. outlier.ai: Train the Next Generation of AI as a Freelancer | Outlier AI
  5. alignerr.com: Alignerr | Powered by Labelbox (its own page, listed on this site's dataset page)
  6. work.mercor.com: Explore Opportunities | Mercor (its own page, listed on this site's dataset page)
  7. mercor.com: Mercor | Organizing human intelligence to power the AI economy (its own page, listed on this site's dataset page)
  8. micro1.ai: High-paying Remote AI Training Jobs | micro1

The question this page is here to answer

What RLHF work does each platform describe, which tracks does it offer, and does it say no experience is needed?

Say which RLHF task suits you and your field. We email you when a platform on the table names new tasks, tracks or entry tests for it, and only then.

Answered by RLHFJobs, which reads every enquiry this page takes.

Email me RLHF task changes

Email me RLHF task changes