Referring video object segmentation (RVOS) commonly grounds targets in videos based on static textual cues. MeViS benchmark extends this by incorporating motion-centric expressions (referring & reasoning motion expressions) and introducing no-target queries. Extending SaSaSa2VA, where increased input frames and [SEG] tokens already strengthen the Sa2VA backbone, we adopt a simple yet effective target existence-aware verification mechanism, leading to Still Awesome SaSaSa2VA (SaSaSaSa2VA). Despite its simplicity, the method achieves a final score of 89.19 in the 5th PVUW Challenge (MeViS-Text Track), securing 2nd place. Both quantitative results and ablations suggest that this existence-aware verification strategy is sufficient to unlock strong performance on motion-centric referring tasks.
Figures & tables
Rank
Team
J&F
N-acc .
T-acc.
Final
# 1
HITsz_Dragon
78.97
96.15
97.59
90.91
# 2
SaSaSaSa2VA
71.06
100.00
96.52
89.19
# 3
tobedone
71.30
96.15
98.93
88.79
# 4
yahooo
70.38
96.15
98.40
88.31
# 5
junjie_zheng
68.37
88.46
96.79
84.54
Table 1: Leaderboard of the 5th PVUW Challenge MeViS-Text Track at CVPR 2026. Our SaSaSaSa2VA team achieves a final score of 89.19 and wins 2nd place.
Figure 1: The architecture of SaSaSa2VA [ 25 ] . Given a video of T frames, the sequence is first split into N temporally ordered clips, each consisting of c=g2+1 frames. To improve efficiency while preserving temporal context, frames within each clip are compacted via the Key Frame Compression (KFC) strategy before being fed into the MLLM. Based on the compressed visual inputs, the MLLM generates a set of N [SEG] tokens, where each token encodes segmentation cues for a specific temporal segment. For each clip, SAM2 takes the corresponding [SEG] token as a prompt, together with the original (uncompressed) frames, to decode object masks at the frame level. In this illustration, we set c=5 , corresponding to g=2 .
Figure 2: The Existence-aware verification illustration of our method. Given a video-expression pair (V,T) , Gemini 3-Flash-Preview and GPT-5.4 function as a dual-consensus jury, and an expression is categorized as ’null-target’ only under a unanimous consensus, where both models independently confirm the object’s absence. Only valid video-text pairs proceed to the SaSaSa2VA base model for inference.
Method
J&F
N-acc
T-acc
SaSaSa2VA [ 25 ]
68.04
5.26
99.77
Ours
ft.
68.55
5.26
99.77
EA
72.43
97.37
100.00
ft. + EA
72.84
97.37
100.00
Table 2: Ablation on Existence-aware Augmentation (EA). We report the J&F and N-acc scores of the 26B model on the MeViS V2 [ 12 ] valid_u split. “ft.” denotes further fine-tuning on the MeViS V2 training split.