TimeLens: Rethinking Video Temporal Grounding with Multimodal LLMs

Open paper
Benchmark Model Rank Results
temporal-sentence-grounding-on-qvhighlights-timelens-validationGemini-2.5-Pro#4mIoU: 70.4R@1 IoU=0.3: 84.1R@1 IoU=0.5: 75.9R@1 IoU=0.7: 61.1
temporal-sentence-grounding-on-qvhighlights-timelens-validationTimeLens-8B (extra data)#6mIoU: 65.5R@1 IoU=0.3: 80.2R@1 IoU=0.5: 71.6R@1 IoU=0.7: 55.5
temporal-sentence-grounding-on-qvhighlights-timelens-validationQwen3-VL-235B-A22B#7mIoU: 64.6R@1 IoU=0.3: 79.6R@1 IoU=0.5: 70.2R@1 IoU=0.7: 54.5
temporal-sentence-grounding-on-qvhighlights-timelens-validationGemini-2.5-Flash#8mIoU: 64.3R@1 IoU=0.3: 78.2R@1 IoU=0.5: 69.4R@1 IoU=0.7: 55.0
temporal-sentence-grounding-on-qvhighlights-timelens-validationGemini-2.0-Flash#10mIoU: 60.8R@1 IoU=0.3: 76.2R@1 IoU=0.5: 66.4R@1 IoU=0.7: 48.3
temporal-sentence-grounding-on-qvhighlights-timelens-validationQwen3-VL-8B#11mIoU: 59.4R@1 IoU=0.3: 74.2R@1 IoU=0.5: 64.6R@1 IoU=0.7: 49.3
temporal-sentence-grounding-on-qvhighlights-timelens-validationGPT-5#13mIoU: 56.8R@1 IoU=0.3: 72.4R@1 IoU=0.5: 60.4R@1 IoU=0.7: 46.4
temporal-sentence-grounding-on-qvhighlights-timelens-validationTimeLens-7B (extra data)#14mIoU: 56.0R@1 IoU=0.3: 74.1R@1 IoU=0.5: 62.7R@1 IoU=0.7: 43.1
temporal-sentence-grounding-on-qvhighlights-timelens-validationGPT-4o#15mIoU: 52.1R@1 IoU=0.3: 69.0R@1 IoU=0.5: 54.8R@1 IoU=0.7: 38.5
temporal-sentence-grounding-on-qvhighlights-timelens-validationTime-R1-7B#16mIoU: 49.2R@1 IoU=0.3: 65.8R@1 IoU=0.5: 51.5R@1 IoU=0.7: 36.1
temporal-sentence-grounding-on-qvhighlights-timelens-validationMiMo-VL-7B-RL#17mIoU: 41.5R@1 IoU=0.3: 57.1R@1 IoU=0.5: 42.6R@1 IoU=0.7: 28.4
temporal-sentence-grounding-on-qvhighlights-timelens-validationTRACE-uni#18mIoU: 39.8R@1 IoU=0.3: 49.9R@1 IoU=0.5: 40.0R@1 IoU=0.7: 29.2
temporal-sentence-grounding-on-qvhighlights-timelens-validationTRACE#19mIoU: 39.0R@1 IoU=0.3: 49.7R@1 IoU=0.5: 39.1R@1 IoU=0.7: 28.1
temporal-sentence-grounding-on-qvhighlights-timelens-validationGrounded-VideoLLM#20mIoU: 33.4R@1 IoU=0.3: 43.7R@1 IoU=0.5: 33.8R@1 IoU=0.7: 22.5
temporal-sentence-grounding-on-qvhighlights-timelens-validationVideoChat-Flash-7B#21mIoU: 32.7R@1 IoU=0.3: 45.2R@1 IoU=0.5: 30.6R@1 IoU=0.7: 16.7
temporal-sentence-grounding-on-qvhighlights-timelens-validationQwen2.5-VL-7B#22mIoU: 31.6R@1 IoU=0.3: 41.5R@1 IoU=0.5: 27.8R@1 IoU=0.7: 15.2
temporal-sentence-grounding-on-qvhighlights-timelens-validationTimeSuite#23mIoU: 21.7R@1 IoU=0.3: 27.1R@1 IoU=0.5: 16.9R@1 IoU=0.7: 9.9
temporal-sentence-grounding-on-qvhighlights-timelens-validationVideoChat-R1-7B#24mIoU: 21.5R@1 IoU=0.3: 29.3R@1 IoU=0.5: 19.1R@1 IoU=0.7: 9.4
temporal-sentence-grounding-on-activitynet-timelensGemini-2.5-Pro#6mIoU: 58.1R@1 IoU=0.3: 72.3R@1 IoU=0.5: 64.2R@1 IoU=0.7: 47.1
temporal-sentence-grounding-on-activitynet-timelensTimeLens-8B (extra data)#8mIoU: 53.2R@1 IoU=0.3: 68.9R@1 IoU=0.5: 58.4R@1 IoU=0.7: 40.6
temporal-sentence-grounding-on-activitynet-timelensGemini-2.5-Flash#9mIoU: 52.5R@1 IoU=0.3: 66.8R@1 IoU=0.5: 57.5R@1 IoU=0.7: 41.3
temporal-sentence-grounding-on-activitynet-timelensQwen3-VL-235B-A22B#10mIoU: 52.2R@1 IoU=0.3: 69.0R@1 IoU=0.5: 57.5R@1 IoU=0.7: 39.3
temporal-sentence-grounding-on-activitynet-timelensGemini-2.0-Flash#11mIoU: 49.3R@1 IoU=0.3: 62.9R@1 IoU=0.5: 54.0R@1 IoU=0.7: 37.7
temporal-sentence-grounding-on-activitynet-timelensQwen3-VL-8B#12mIoU: 46.8R@1 IoU=0.3: 62.1R@1 IoU=0.5: 51.2R@1 IoU=0.7: 34.4
temporal-sentence-grounding-on-activitynet-timelensTimeLens-7B (extra data)#14mIoU: 46.2R@1 IoU=0.3: 62.8R@1 IoU=0.5: 51.0R@1 IoU=0.7: 32.6
temporal-sentence-grounding-on-activitynet-timelensGPT-5#15mIoU: 42.9R@1 IoU=0.3: 57.4R@1 IoU=0.5: 44.9R@1 IoU=0.7: 30.4
temporal-sentence-grounding-on-activitynet-timelensGPT-4o#16mIoU: 40.4R@1 IoU=0.3: 55.2R@1 IoU=0.5: 41.4R@1 IoU=0.7: 25.8
temporal-sentence-grounding-on-activitynet-timelensMiMo-VL-7B-RL#17mIoU: 35.5R@1 IoU=0.3: 49.3R@1 IoU=0.5: 38.7R@1 IoU=0.7: 22.4
temporal-sentence-grounding-on-activitynet-timelensTRACE-uni#18mIoU: 33.6R@1 IoU=0.3: 44.3R@1 IoU=0.5: 35.1R@1 IoU=0.7: 22.6
temporal-sentence-grounding-on-activitynet-timelensTime-R1-7B#19mIoU: 33.1R@1 IoU=0.3: 44.8R@1 IoU=0.5: 31.0R@1 IoU=0.7: 19.0
temporal-sentence-grounding-on-activitynet-timelensTRACE#20mIoU: 32.7R@1 IoU=0.3: 43.4R@1 IoU=0.5: 33.9R@1 IoU=0.7: 22.0
temporal-sentence-grounding-on-activitynet-timelensQwen2.5-VL-7B#21mIoU: 31.4R@1 IoU=0.3: 44.1R@1 IoU=0.5: 31.0R@1 IoU=0.7: 16.1
temporal-sentence-grounding-on-activitynet-timelensGrounded-VideoLLM#22mIoU: 30.0R@1 IoU=0.3: 39.2R@1 IoU=0.5: 29.6R@1 IoU=0.7: 19.5
temporal-sentence-grounding-on-activitynet-timelensVideoChat-R1-7B#23mIoU: 25.0R@1 IoU=0.3: 35.0R@1 IoU=0.5: 23.9R@1 IoU=0.7: 11.3
temporal-sentence-grounding-on-activitynet-timelensVideoChat-Flash-7B#24mIoU: 24.8R@1 IoU=0.3: 35.5R@1 IoU=0.5: 21.8R@1 IoU=0.7: 10.5
temporal-sentence-grounding-on-activitynet-timelensTimeSuite#25mIoU: 19.8R@1 IoU=0.3: 27.1R@1 IoU=0.5: 17.5R@1 IoU=0.7: 8.6
temporal-sentence-grounding-on-charades-timelensTimeLens-8B (extra data)#6mIoU: 55.2R1@0.5: 63.0R1@0.7: 35.2R1@0.3: 76.6
temporal-sentence-grounding-on-charades-timelensGemini-2.5-Pro#8mIoU: 52.8R1@0.5: 61.1R1@0.7: 34.0R1@0.3: 74.1
temporal-sentence-grounding-on-charades-timelensTimeLens-7B (extra data)#10mIoU: 48.8R1@0.5: 55.6R1@0.7: 28.4R1@0.3: 70.5
temporal-sentence-grounding-on-charades-timelensGemini-2.5-Flash#11mIoU: 48.6R1@0.5: 56.1R1@0.7: 30.6R1@0.3: 68.7
temporal-sentence-grounding-on-charades-timelensQwen3-VL-8B#12mIoU: 48.3R1@0.5: 53.4R1@0.7: 27.5R1@0.3: 69.2
temporal-sentence-grounding-on-charades-timelensQwen3-VL-235B-A22B#13mIoU: 47.8R1@0.5: 50.8R1@0.7: 24.5R1@0.3: 71.7
temporal-sentence-grounding-on-charades-timelensGemini-2.0-Flash#14mIoU: 46.7R1@0.5: 53.5R1@0.7: 27.1R1@0.3: 66.4
temporal-sentence-grounding-on-charades-timelensGPT-4o#15mIoU: 41.8R1@0.5: 44.5R1@0.7: 23.5R1@0.3: 60.6
temporal-sentence-grounding-on-charades-timelensGPT-5#16mIoU: 40.5R1@0.5: 42.0R1@0.7: 22.0R1@0.3: 59.3
temporal-sentence-grounding-on-charades-timelensVideoChat-Flash-7B#17mIoU: 39.7R1@0.5: 37.9R1@0.7: 17.8R1@0.3: 60.2
temporal-sentence-grounding-on-charades-timelensMiMo-VL-7B-RL#18mIoU: 39.6R1@0.5: 42.6R1@0.7: 20.5R1@0.3: 57.9
temporal-sentence-grounding-on-charades-timelensQwen2.5-VL-7B#19mIoU: 39.3R1@0.5: 37.8R1@0.7: 16.6R1@0.3: 59.7
temporal-sentence-grounding-on-charades-timelensTimeSuite#20mIoU: 38.1R1@0.5: 35.5R1@0.7: 18.0R1@0.3: 56.3
temporal-sentence-grounding-on-charades-timelensTime-R1-7B#21mIoU: 36.6R1@0.5: 32.0R1@0.7: 16.9R1@0.3: 57.9
temporal-sentence-grounding-on-charades-timelensVideoChat-R1-7B#22mIoU: 33.7R1@0.5: 30.8R1@0.7: 11.7R1@0.3: 51.9
temporal-sentence-grounding-on-charades-timelensGrounded-VideoLLM#23mIoU: 30.0R1@0.5: 28.7R1@0.7: 13.5R1@0.3: 43.3
temporal-sentence-grounding-on-charades-timelensTRACE-uni#24mIoU: 28.1R1@0.5: 22.9R1@0.7: 10.4R1@0.3: 38.2
temporal-sentence-grounding-on-charades-timelensTRACE#25mIoU: 27.1R1@0.5: 21.8R1@0.7: 9.6R1@0.3: 37.2
temporal-sentence-grounding-on-charades-timelensTimeLens#26R1@0.5: 63.0

Rank counts only results with a code link.