TOHA detects LLM hallucinations via topological divergence on attention graphs, showing consistent patterns and competitive benchmark results with minimal data.
Responses with a Rouge-L score of 1 (exact match) were labeled as grounded
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Hallucination Detection in LLMs with Topological Divergence on Attention Graphs
TOHA detects LLM hallucinations via topological divergence on attention graphs, showing consistent patterns and competitive benchmark results with minimal data.