The authors built on a generative video temporal grounding model (based on a vision-language decoder) that outputs candidate intervals as timestamp sequences.
Confidence scores for generated video intervals enable controllable selection without external verification
The authors train a lightweight head that scores each generated interval from decoder states, allowing ranking, threshold-based selection, and rejection using a single decoding pass.
Chinese Tech
Jinhao Chen · Benlei Cui · Ruijian Jia · Ziheng Wang · Tianyu Wo · Pengfei Sun · +4 more
Alibaba Group · Beihang University · Fudan University
Research Digest··3 min read
Chen et al.
Why this paper
From Alibaba Group and 2 others
In one line
Interval-level confidence scores trained with offline verifiers and reinforcement learning improve candidate selection in generative video temporal grounding.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§