ResearchDatasetsVLM prefillVLM prefillFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsInference LatencyVLM prefill 1024 tokens92.7Latency (ms)2Inference LatencyVLM prefill 512 tokens59.4Prefill Latency (ms)2