TimeLens2 claims SOTA on 7 video grounding benchmarks with 4B and 8B models
_akhaliq · x · 2026-07-21
TimeLens2 is a generalist video temporal grounding MLLM that predicts when supporting evidence appears in videos.
- It reports SOTA across 7 benchmarks.
- The 4B and 8B variants are said to outperform a 397B-parameter model.
- The figure highlights performance across grounding dimensions such as short indoor action grounding, event-level grounding, moments and highlights detection, long-video multi-span grounding, user-style grounding, interrogative grounding, and first-person egocentric grounding.
Related event: TimeLens2 Sets New SOTA in Video Temporal Grounding(2 posts)→
More from Multimodal
- Dense instance segmentation via Muse Wonder + SAM 3.1 on Meta's Model API — nikhilaravi · 2026-10-26
- LichtFeld Studio v0.5.4.3 ships community-built SteadySplats, POPSpa and Voxelization — janusch_patas · 2026-10-26
- One prompt to a 2-3 min narrated short: hands-on test of Antigravity video generation — AI_Andrew · 2026-10-26
- Creator makes 10,564-frame animated short entirely with Gemini toolchain — AI_Andrew · 2026-10-26
- FLUX.3 Klein step-distilled to 4-8 steps with CFG baked in — Fit-Structure-6489 · 2026-10-26
- ostris pre-training "Primal VAE": 1-channel latents match modern 8x VAEs via patching — ostrisai · 2026-10-26