Fine tuning is widely expected to teach a model new facts, and that is largely not what it does.
What it is good at
Format, tone, task structure and domain style.
Which are behavioural rather than factual.
What it is poor at
Reliably installing new factual knowledge.
Which retrieval handles better.
Catastrophic forgetting
Capabilities degrading in areas outside the fine tuning data.
Which is a documented risk.
Parameter efficient methods
Adjusting a small subset of weights.
Which makes fine tuning affordable.
Why the common expectation is wrong
Organisations frequently approach fine tuning expecting to teach a model their internal documentation, and it is a poor mechanism for that.
Which is because knowledge in these models is distributed across weights rather than stored as retrievable records, and a modest amount of additional training does not install facts reliably.
Retrieval, where the relevant document is supplied at the time of the question, does that job far better and at a fraction of the cost and effort.
What it is genuinely for
Consistent output format, a specific tone, and task patterns the base model handles awkwardly.
Which are all behavioural adjustments.
Data requirements
Quality mattering considerably more than quantity.
Which surprises teams expecting to need enormous datasets.
Evaluation
Testing on held-out examples and on general capability.
Which catches degradation elsewhere.
When to consider it
After retrieval and prompting have been tried properly.
Why the mechanics are worth understanding
These systems are described in language that suggests understanding, reasoning and knowledge, and the underlying operations are considerably more specific than those words imply. That gap is where most confusion, most disappointment and a substantial amount of misplaced trust originates.
Someone who knows that a language model predicts tokens, that it has no separate lookup step, that its context is a hard limit and that its confidence is unrelated to its accuracy will make better decisions about when to use one than someone working from the marketing description.
The pattern across all of this
Nearly every behaviour that people find surprising follows directly from a design choice that is documented and comprehensible. Invented citations follow from the training objective. Difficulty with spelling follows from tokenisation. Degradation in long conversations follows from context limits. Cost structures follow from the fact that inference is not free.
None of these are mysteries or failures of implementation. They are consequences, and they are predictable once the mechanism is understood.
What follows in practice
Use these systems for work where the output can be checked, where fluency is genuinely useful, and where a draft that needs revision is more valuable than a blank page. Avoid relying on them for specific facts, citations or numbers where a confident wrong answer is worse than no answer.
That is a narrower recommendation than the enthusiasm suggests and a much broader one than the dismissal allows. Both positions tend to be argued by people who have not looked closely at what the systems do.
Where the reliable information is
Research papers, model documentation, technical writing from the organisations building these systems, and independent evaluation work. Most of it is freely available and considerably more measured than either the promotional material or the criticism built on top of it.
The field moves quickly enough that any general article, including this one, dates rapidly. Anything with consequences attached is worth checking against current documentation.
On the pace of change
Capabilities have moved quickly enough that specific claims about what models can and cannot do have a short shelf life. Several limitations described confidently a few years ago have been substantially reduced, and several that were expected to fall have not.
The underlying mechanisms have changed far less than the capabilities. Next token prediction, tokenisation, context limits and the absence of a verification step have all remained, which is why the same categories of failure keep recurring in new forms even as performance improves.
A note on evaluating claims
When a capability is announced, the useful questions are: measured on what benchmark, compared against what, evaluated by whom, and reproducible by anyone outside the organisation making the claim. Those four questions dispose of a considerable amount of what circulates.
Independent replication in this field is under-resourced relative to the scale of the claims being made, which is a structural problem rather than a criticism of any particular organisation. It means that a great deal of what is asserted has never been checked by anyone with an incentive to find it wanting.
One last observation
The most useful thing anyone can know about these systems is what they are optimising for, because everything else follows from it. A model trained to produce plausible text will produce plausible text, reliably, and regardless of whether the text is true.
That is not a criticism. It is a specification, and reading it as one makes the whole field considerably easier to think about clearly.