ResearchTasksMid-Session Tool Injection (MSTI) EvaluationMid-Session Tool Injection (MSTI) EvaluationFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedWebMCP MSTI Testbed (test)AbortSignal hijack100ASR (GPT-5.4)5Jun 5, 2026