Reinforcement learning from human feedback (RLHF)
A preference-tuning method in which human judgments of which answers are better are used to further train a model toward preferred behavior. It is a major reason two versions of the same base model can behave differently, and it is the kind of tuning that produced the divergence in the Maverick episode.
Defined in 2 GAGE programs, which carry 3 distinct definitions of it. The wording above is taught in Certified AI Governance Professional (CAIGP).
How each discipline defines it
The same term does different work depending on who is using it. These are the definitions as each program teaches them, unedited.
A preference-tuning method in which human judgments of which answers are better are used to further train a model toward preferred behavior. It is a major reason two versions of the same base model can behave differently, and it is the kind of tuning that produced the divergence in the Maverick episode.
A specific fine-tuning technique in which human reviewers rank or rate model outputs, and the model is further trained to produce outputs more like the highly rated ones; a major factor in the shift from raw base-model behavior to a consistently helpful, instruction-following deployed assistant.
A weight-layer method for tuning a model in which human ratings of its outputs are used to nudge the model toward the rated-good behaviors. A common place the alignment tax appears.
Where it is taught
The exact lessons this term appears in. The first module of every program is free with a free account.
- Fixing the model, breaking it again: why fixes are never free · Build Before You Govern, Certified AI Governance Professional (CAIGP)
- Prompts, context, and why the same model gives different companies different answers · Build Before You Govern, Certified AI Governance Professional (CAIGP)
- How AI Actually Works Under the Hood: Architectures and Limits for Policy People · Technical Credibility Deep Dive, The AI Lobbyist: Certified AI Policy Strategist
Where it shows up at work
A definition is where Reinforcement learning from human feedback (RLHF) starts. The lessons above teach it inside the competencies below, which are what job descriptions ask for and what the GAGE credential records. Every link here is drawn from the same careers map that builds a role's readiness check.
Explains training, tokens, context windows, embeddings, retrieval and fine-tuning well enough to ask an engineer a precise question and spot weak evidence.
2 lessons on this page teach it as a primary outcome.
Decides what a person affected by an AI decision must be told, how an output can be explained, and how they can challenge it.
One lesson on this page teach it as a primary outcome.
Recognizes hallucination, drift, skew, brittleness and biased outcomes, and knows how each one enters a system.
One lesson on this page teach it as a primary outcome.
Roles that hire for it
Each role page shows the full competency set, what the seat decides, and a readiness check against your own evidence.
- AI Model Validatorneeds it at depth
- AI Evaluation Specialistneeds it at depth · 6 open postings this week
- Chief Technology Officer, AI engineering editionneeds it at depth
- AI Regulatory Counselneeds working knowledge · 14 open postings this week
- AI Product Manager, responsible product editionneeds working knowledge · 3 open postings this week
- AI Security Architectneeds it at depth · 8 open postings this week
How job descriptions say it
Postings rarely use the textbook word. These are the phrases the careers map reads as asking for the competencies above.
Where the frameworks place it: EU AI Act Articles 13 and 50; NIST AI RMF trustworthiness characteristics; NIST AI RMF MEASURE 2; ISO/IEC 42001.
One interview question about it
Explain how a large language model produces an answer, at the depth a governance decision needs and no deeper.
Asked of candidates for how models work, at a governance depth. Answering it in your own words is exactly how a GAGE topic is passed, and the credential records the pass where an employer can check it.
If you come from another field
Some backgrounds already carry this competency, and the readiness check gives credit for it.
- Data analysis, stewardship and business intelligence carries it as a claim to back up: SQL and analytical fluency make the technical layer approachable.
- Model risk, credit risk and quantitative analysis carries it as work you can show: Quantitative fluency is demonstrated in your work.
- A computer science or software engineering degree carries it as work you can show: You have trained or integrated a model in coursework or a project.
Terms it appears with
Not an alphabetical neighbourhood: these are the terms taught in the same lessons, ranked by how often they appear together.