When Should the Count Change? Learning State Maintenance for Causal Video Counting
cs.CV
Submitted: 2026-09-28
Updated: 2026-09-28
Comments: 28 pages, 7 figures. Project Page: https://PLACEHOLDER.github.io/StaMina/
Project page: https://placeholder.github.io/StaMina
License: http://creativecommons.org/licenses/by/4.0/
The gist: Continuous video counting requires distinguishing new observations from new objects or completed events.
Terminology
Abstract
Continuous video counting requires distinguishing new observations from new objects or completed events. We introduce StaMina (State Maintenance), which learns to maintain counting state through state-conditioned updates. Recurrent visual context supports recognition; learned transitions maintain visibility, persistent identities, and completed-event records. A differentiable recurrence trains event transitions over legal paths constrained by count endpoints; visibility and association objectives train the object branch. A multi-source pipeline organizes 39.8K spatial queries and complementary event annotations into counting trajectories. On SVCBench, we evaluate counting adaptation with partial video overlap and held-out groups of linked annotations. Under prefix replay (Full) and persistent streaming (Stream), 4B and 8B models reach 41.9/36.4 and 44.9/38.2 Gaussian Precision Accuracy, respectively. The 8B model gains 10.9/3.2 points over Counting-SFT on the same queries. Matched-graph comparisons isolate phase conditioning and trajectory supervision, assessing training objectives alongside hard decisions. Online video benchmarks and count-conditioned decisions assess online understanding and task eligibility. Project Page: https://PLACEHOLDER.github.io/StaMina/
Sources
- Layer Normalization
- Qwen3-VL Technical Report
- Qwen2.5-VL Technical Report
- StreamTTT: Reconciling Real-Time Perception and Long-Term Memory in Streaming VLMs
- Molmo2: Open Weights and Data for Vision-Language Models with Video Understanding and Grounding
- Gaussian Error Linear Units (GELUs)
- OpenVLA: An Open-Source Vision-Language-Action Model
- Kimi Linear: An Expressive, Efficient Attention Architecture
- Kimi K2.5: Visual Agentic Intelligence
- Video-MME-Logical: A Controlled Diagnostic Benchmark for Video Temporal-Logical Reasoning
- OVO-S-Bench: A Hierarchical Benchmark for Streaming Spatial Intelligence in Multimodal LLMs
- StreamingBench: Assessing the Gap for MLLMs to Achieve Streaming Video Understanding
- Spatial-TTT: Streaming Visual-based Spatial Intelligence with Test-Time Training
- Decoupled Weight Decay Regularization
- ClipTTT: CLIP-Guided Test-Time Training Helps LVLMs See Better
- End-to-End Test-Time Training for Long Context
- InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
- Cambrian-S: Towards Spatial Supersensing in Video
- Flash-VStream: Memory-Based Real-Time Understanding for Long Video Streams
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models