Author

aiterms

What is an LLM eval (evaluation)?

An LLM eval, or evaluation, is a test or set of tests used to measure how well a large language model behaves on a task or under a specific set of conditions.

Why it matters

If you build with LLMs, you need a way to answer practical questions like:

• Does this prompt change improve answer quality?
• Is the model following instructions more reliably?
• Did a new model version get worse on safety, latency, or factuality?

An eval gives you a repeatable way to compare models, prompts, retrieval setups, or fine-tunes. In practice, teams use evals before shipping changes, during prompt iteration, and as regression tests so quality does not quietly drift.

How it works

An eval usually starts with a defined task and a dataset or test set. That might be a list of question-answer pairs, conversation snippets, coding tasks, or policy/safety cases.

Then you run the model on those cases and score the outputs against some criterion. The score can be:

• exact match or task-specific correctness
• human judgment
• a rubric-based rating
• an automatic metric such as accuracy, BLEU, or pass@k, depending on the task

For many LLM use cases, the evaluation is not just one number. Teams often track several dimensions, such as helpfulness, factuality, refusal behavior, latency, and cost. That is because a model can improve on one axis and regress on another.

The important idea is repeatability: the same test should let you compare two versions fairly, even if the model itself is non-deterministic.

Tiny concrete example

Suppose you are testing a customer-support assistant.

Test case:
User: “I was charged twice. What should I do?”

Eval rubric:

• Gives a clear next step
• Does not invent account access it does not have
• Advises contacting support with transaction details

You run the model and check whether its answer matches the rubric. If version B starts saying “I’ve already refunded you,” the eval catches a regression.

Common pitfalls / when NOT to use it

• Using a single metric for everything. One score rarely captures quality, safety, and usefulness at the same time.
• Overfitting to the test set. If you keep tuning until the model memorizes your eval, the score stops reflecting real-world performance.
• Bad or narrow test cases. An eval is only as good as the scenarios it covers. If your cases are too easy or too repetitive, they miss real failures.
• Treating automatic metrics as ground truth. For open-ended generation, automated scoring can be useful but often misses nuance; human review is still common.
• Using evals as a substitute for product judgment. A model can score well and still be wrong for your UX, policy, or cost constraints.

Related terms

• model benchmark
• test set
• golden dataset
• rubric
• human evaluation

Related terms

• What is LLM-as-a-judge?
• What is a golden dataset / eval set?
• What is a guardrail?
• What is hallucination detection?
• What is regression testing for prompts?
• What is observability / tracing for LLM apps?

同じ著者の記事

What is LLM-as-a-judge?

LLM-as-a-judge is the practice of using a large language model to evaluate or rank outputs from another model, system, or workflow, instead of relying only on human reviewers or fixed metrics. It solves a common problem: many AI outputs are hard to score with simple automated checks. If you are judging a summary for usefulness, a chatbot answer for helpfulness, or two candidate responses for preference, exact-match metrics often miss the point. Teams reach for LLM-as-a-judge when they need: **Fa

papoo.work

What is a hallucination?

A hallucination is when an AI system generates output that sounds plausible but is wrong, unsupported, or made up. Hallucinations matter because they break trust. If you use a model for search, summarization, coding help, or customer support, a confident but false answer can mislead users, create bugs, or spread bad information. In practice, teams care about hallucinations whenever they need factual accuracy, source grounding, or reliable behavior. For creative writing, they may be less of a pro

papoo.work

What is the KV cache?

The KV cache, or key-value cache, is a memory buffer that stores past attention keys and values so a transformer can generate the next token without recomputing everything from scratch. In autoregressive generation, each new token depends on all previous tokens. Without a KV cache, the model would repeatedly re-run attention over the entire prefix at every step, which is much slower than necessary. You reach for a KV cache when you want faster text generation, lower latency, or higher throughput

papoo.work

What is a parameter (model size)?

A parameter is a learned number inside a model, and “model size” usually means how many of those learned numbers the model has. Parameters are the part of a neural network that gets updated during training, so they determine how much capacity the model has to fit patterns in data. You’ll care about parameter count when you want to compare models, estimate training cost, understand memory use, or get a rough sense of how much hardware a model needs. In practice, “bigger” often means more capable,

papoo.work

What is a mixture-of-experts (MoE) model?

A mixture-of-experts (MoE) model is a neural network that routes each input, or part of an input, to only a small subset of specialized submodels (“experts”) instead of using the whole model every time. MoE models are useful when you want much larger capacity without paying the full compute cost on every token or example. In practice, that means you can scale model size to improve quality while keeping inference and training more efficient than a dense model of similar total parameter count.

papoo.work

What is a token (and tokenization)?

A token is a chunk of text that a language model reads and predicts, and tokenization is the process of splitting text into those chunks. LLMs do not operate directly on raw characters or words; they operate on tokens. That matters because: Input limits are measured in tokens, not characters. Cost and latency are often tied to token count in LLM APIs. Model quality depends on how text is tokenized: the same sentence can become very different token sequences across tokenizers.

papoo.work

What is a transformer?

A transformer is a neural network architecture that processes sequences by using attention to compare each token with the others, making it a flexible foundation for modern language models and many other sequence tasks. Before transformers, many sequence models had to read text step by step, which made long-range dependencies harder to handle and training slower to parallelize. Transformers solve this by letting the model look at many positions in the input at once. In practice, you reach for a

papoo.work

What is attention (self-attention)?

Attention, or self-attention, is a way for a model to decide which other words or tokens in the same input matter most when interpreting each token. Self-attention solves a basic language problem: the meaning of a word often depends on other words nearby or far away. It lets a model build a context-aware representation of each token instead of treating tokens mostly independently or relying only on fixed-size windows. In practice, this is why Transformers work so well for text, code, and other s

papoo.work

What is temperature in an LLM?

Temperature is a sampling setting that controls how random or deterministic an LLM’s next-word choices are when it generates text. Temperature is one of the simplest ways to trade off consistency versus variety. Use a lower temperature when you want stable, repeatable answers: classification-like outputs, structured data, summaries, or code. Use a higher temperature when you want more diversity: brainstorming, creative writing, alternate phrasings, or exploratory chat. In practice, many

papoo.work

What is top-p / nucleus sampling?

Top-p sampling, also called nucleus sampling, is a way to make a language model pick the next token from only the smallest set of likely options whose combined probability reaches a chosen threshold. If you always choose the single most likely next token, model outputs can become repetitive and brittle. If you sample from all tokens, outputs can get noisy and incoherent. Top-p gives you a middle ground: it keeps the model flexible, but trims away the long tail of very unlikely tokens before samp

papoo.work