MemInStream: Memorizing or Merely Recognizing? Rethinking Streaming Video Memory Evaluation
Abstract
MemInStream is a benchmark that evaluates memory capabilities in streaming video understanding using evidence-controlled validation to separate memory limitations from visual understanding failures.
Streaming video understanding aims to continuously process incoming visual observations, making the retention and retrieval of relevant information increasingly important as the video history grows. Although existing benchmarks include memory-related questions, question-answering accuracy alone does not reliably measure memory capabilities. Incorrect predictions may arise from failures in visual understanding rather than forgetting, whereas correct predictions may rely on linguistic priors or recent observations without requiring historical memory. To address these ambiguities, we introduce MemInStream, a benchmark dedicated to evaluating memory capabilities in streaming video understanding. We develop an evidence-controlled validation protocol that associates each question with temporally localized evidence intervals. The protocol validates evidence solvability and memory necessity by retaining questions that models can answer from the annotated evidence but cannot resolve using recent observations, post-evidence observations, or textual input alone. This helps disentangle memory limitations from understanding failures while reducing memory-independent shortcuts. MemInStream provides a systematic evaluation of memory capabilities through four major categories, covering Episodic Retrieval, State Tracking \& Updating, Multi-Evidence Integration, and Contextual Memory across diverse video scenarios. Overall, MemInStream provides a comprehensive and reliable benchmark for systematically evaluating memory capabilities in streaming video understanding.
Community
When a video model answers a question correctly, did it remember an earlier event, or could it infer the answer from the latest frames? We introduce MemInStream to study this question. Across 3,799 questions and 17 tasks, we identify the video segments that support each answer and test whether the question can still be answered without them. The videos, annotations, and evaluation code are publicly available.
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- TwinPast: Benchmarking Historical Evidence Use in Streaming Video Understanding with Evidence Contrastive Video Pairs (2026)
- TimeTraveler: Temporal Strategy Planning with Time Dictionary for Streaming Video Understanding (2026)
- OneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video Interaction (2026)
- EgoProbe: Benchmarking Long-Range and Exact Memory in Streaming Egocentric Videos (2026)
- Retrieval-Augmented Online Textual Memory for Episodic Memory Video Question Answering (2026)
- FlashBack: Knowing When to Remember in Streaming Vision-Language Models (2026)
- Memory as Weights: Internalizing Long-Term History for Streaming Videos (2026)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend
Get this paper in your agent:
hf papers read 2611.15640 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Datasets citing this paper 1
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper