· Xiaojing Yang · NLP and LLMs · 2 min read

中文

LLM Evaluation and Failure Modes

A practical map of LLM evaluation risks: hallucination, prompt sensitivity, bias, contamination, and brittle benchmarks.

Core idea

LLM evaluation must test behavior, not just benchmark scores.

1. Why LLM evaluation is different

LLMs are general-purpose systems. They can follow instructions, generate fluent text, retrieve-looking answers, and explain themselves even when wrong. Evaluation must therefore check behavior under realistic conditions.

LLM evaluation loop
Task
what the user needs
Prompt set
instructions and contexts
Model outputs
possibly stochastic
Judging
metric, human, or model judge
Failure analysis
patterns and severity

2. Common failure modes

Failure modeWhat to check
Hallucinationunsupported claims
Prompt sensitivityunstable answers across wording
Biassystematic disparities
Contaminationbenchmark examples seen in training
Verbosity trapfluent but low-value output
Long-context failureignores or misuses evidence

3. Evaluation practice

# Pseudocode
for prompt in prompt_suite:
    output = model.generate(prompt)
    score = judge(output, reference_or_rubric)
    log_failure_mode(output, score)

4. My research connection

This connects to RAG, long documents, multilingual evaluation, and responsible AI. For multilingual LLMs, a model may look strong in English while failing on lower-resource languages or domain-specific technical documents.

Benchmark view

How high is the score?

Research view

Which behaviors are reliable, which fail, and under what conditions?

Takeaway

LLM evaluation is not a scoreboard. It is behavioral science for model systems.

Interview pattern

My interview answer would usually be:

  1. define the concept in one sentence;
  2. explain the data flow;
  3. name the main failure mode;
  4. connect it to evaluation, multilinguality, or fine-tuning.

References

Share:
Back to Blog

Related Posts

View All Posts »
FoundationsNLP and LLMsEN

Attention Mechanism

Attention as a learned way to decide what context matters for each token.

FoundationsNLP and LLMsEN

Fine-Tuning Transformers

How pretrained language models are adapted to a task or domain with supervised data.

FoundationsNLP and LLMsEN

NLP Evaluation

Why NLP evaluation needs metrics, uncertainty, human judgment, and task-specific error analysis.