ResearchTasksPredicting answer correctnessPredicting answer correctnessFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedTriviaQA (val)Baseline (Llama-3.1-8B)0.904AUROC199Jun 2, 2026TruthfulQAHybridGrad0.727AUROC48May 7, 2026GSM8K (val)attention divergence0.945AUROC3May 7, 2026HotpotQA (val)attention divergence0.804AUROC3May 7, 2026TruthfulQA (val)attention divergence0.906AUROC3May 7, 2026