Akashshettyonline·Jun 1110/10 Ways to Reduce Hallucinations in LLM Applications: Human-in-the-Loop (HITL) ValidationThroughout this series, we’ve explored multiple techniques to reduce hallucinations:
Akashshettyonline·Jun 119/10 Ways to Reduce Hallucinations in LLM Applications: Re-ranking Retrieved DocumentsA common assumption when building RAG systems is:
Akashshettyonline·Jun 98/10 Ways to Reduce Hallucinations in LLM Applications: Hybrid Search (Semantic + Keyword Search)When building RAG systems, many developers assume that vector search alone is enough.
Akashshettyonline·Jun 97/10 Ways to Reduce Hallucinations in LLM Applications: Query Rewriting & Intent ClarificationOne of the most overlooked causes of hallucinations isn’t the LLM itself.
Akashshettyonline·Jun 56/10 Ways to Reduce Hallucinations in LLM Applications: Source Attribution & Citation-Based…One of the biggest problems with LLM-generated content is:
Akashshettyonline·Jun 45/10 Ways to Reduce Hallucinations in LLM Applications: Context Window Optimization & Smart…When engineers first build a RAG system, a common instinct is:
Akashshettyonline·Jun 24/10 Ways to Reduce Hallucinations in LLM Applications: Confidence Scoring and Fallback ResponsesOne dangerous characteristic of LLMs is:
Akashshettyonline·Jun 14/10 Ways to Reduce Hallucinations in LLM Applications: Confidence Scoring and Fallback ResponsesOne of the biggest mistakes in LLM-powered systems is forcing the model to answer every question.
Akashshettyonline·May 313/10 Ways to Reduce Hallucinations in LLM Applications: Guardrails and Response ConstraintsOne of the biggest misconceptions about LLM systems is:
Akashshettyonline·May 312/10 Ways to Reduce Hallucinations in LLM Applications: Structured Output ValidationOne major reason LLM applications hallucinate is because responses are completely open-ended.
