From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs
When LLMs support public-facing or high-stakes workflows, missed fabrications can harm users and institutions, while false alarms consume limited human-review capacity. When no trusted context or reference document is available, we study two signals accessible through black-box model APIs: semantic entropy, which measures disagreement among sampled response meanings, and uncertainty derived from token log-probabilities. Their failure modes can be complementary: semantic entropy becomes uninformative when responses form one semantic cluster, while token uncertainty can miss consistently confide
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 50%Researchers develop cost-efficient method for detecting hallucinations in large language models - Tech Xplore →
- PossiblePossibly related (embedding) · 47%Most RAG Hallucinations Are Retrieval Failures: How the Retrieval Brick Decides What the Model Can Invent - Towards Data Science →
- FuzzyOverlapping authors or contributors · 62%ultralytics/yolov5 →
“Shared author/contributor keys: martin”
- LinkedLinked via arxiv author · 85%Urja Pawar →
“From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs”
- LinkedLinked via arxiv author · 85%Rajitha Ramanayake →
“From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs”
- LinkedLinked via arxiv author · 85%Owen O'Neill →
“From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs”
- LinkedLinked via arxiv author · 85%Nabeel Kemal →
“From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs”
- LinkedLinked via arxiv author · 85%Abhishek Mandal →
“From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs”