hour-scale-mean-88-8-min-open-ended-streaming-video-qa CLM-0576observationsingle paperReviewed by AI
On hour-scale (mean 88.8 min) open-ended streaming video QA with causal access, every streaming memory design degrades as the evidence-to-query gap grows, and they fail in different ways: recent-window systems cannot recover distant events (AURA 25.4% within 5 minutes falling to 10.5% beyond 30 minutes), text-summary systems keep temporal coverage but discard visual evidence, and model-internal compression loses fine-grained detail (StreamForest 14.4%, ThinkStream 7.5% on retrospection), while human accuracy without rewatching is 63.4%.
Capability: Reasoning about time in video
Observed on
243 YouTube videos of 60-134 minutes, 3,646 manually validated open-ended QA pairs, LLM-judge binary factual-core scoring, baselines run through their released interfaces which rec.
Sources
- supportsStreamArena: Toward Continuous, Interactive, and Long-Horizon Agentic Streaming Video UnderstandingValidated by: On StreamArena it ranked first among streaming systems on all four capabilities, improving 53.7-228.1% relative over the strongest streaming baseline per capability, and against query-triggered offline inference on the same Qwen3.5-397B-A17B backbone it cut pooled query-to-answer latency 66.2% (81.4s to 27.5s) while retaining 89.7% of pooled accuracy; the authors note baselines use smaller backbones so gains are not isolated to architecture.. [truncated]
Either way, bring evidence: a link, or what you saw.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- For streaming video question answering with a frozen video-LLM, appending retrieved historical frames as extra visual context helps less than optimizing a small set of latent memory tokens at test time and then dropping the retrieved tokens before decoding; the gain is largest on questions requiring backward tracing over past history.Reasoning about time in video · reviewed by AI
- For streaming video understanding with 7-8B multimodal backbones, question-relevant historical frames can be retrieved from the keys of only the first few language-Transformer layers (layer ~3-4 of 28-32), so continuous frames need not be prefilled through full model depth: indexing shallow and re-prefilling only the retrieved units at full depth cuts per-frame prefill up to 52x and 20-second end-to-end latency up to 15x while scoring about 1.7 points above the best same-backbone baseline on OVO-Bench and StreamingBench.Reasoning about time in video · reviewed by AI
- For long-video QA under a fixed frame budget with a frozen 7B multimodal model, organizing frames into a question-agnostic hierarchical temporal tree and searching it coarse-to-fine (answer, then issue a search statement, then expand the relevant branch) raises accuracy over uniform sampling and over agent methods that score frames independently, with the gain largest on the longest-video benchmarks.Reasoning about time in video · reviewed by AI
- On long-video (10-150 minute) audio-visual reasoning questions, multiple-choice accuracy overstates grounding: a large open model can answer about 55% of items correctly while scoring 8.5 on spatio-temporal evidence localization, whereas a frontier proprietary model at 73% accuracy scores 35.2 on the same evidence metric, so answer accuracy alone cannot distinguish deduction from prior-driven guessing.Reasoning about time in video · reviewed by AI
- For 4-8B multimodal models on long-video QA, RL training on questions authored from short scene segments but answered over the whole video improves benchmark accuracy more than RL on questions authored from the whole video, because segment-authored questions target details a whole-video view overlooks; including the segment timestamp in the training instruction adds a further gain by removing referential ambiguity, and is not needed at inference.Reasoning about time in video · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: A streaming system whose retrospection accuracy stays flat across evidence-to-query gaps from 5 to over 30 minutes on this or a comparable hour-scale open-ended benchmark, with no cost in visual-detail or proactive tasks.