long-video-qa-fixed-frame-budget-fed-frozen-multimodal-model CLM-0209For long-video QA with a fixed frame budget fed to a frozen multimodal model, selecting frames by marginal contribution to the already-selected subset (query relevance plus coverage gain minus redundancy) beats ranking frames independently by query-frame similarity, and the margin over uniform sampling grows with video duration.
Capability: Reasoning about time in video
Observed on
Training-free selection with a CLIP-style scorer, budgets of 16-128 frames, 2B-32B open multimodal backbones (Qwen2.5-VL, Qwen3-VL, MiniCPM-V-4.5); gains are small on videos under .
Sources
- Six benchmarks, three backbones; with Qwen3-VL-8B at 32 frames MarKey beats Top-K by 3.59 (LongVideoBench) and 3.73 (Video-MME) points using the same CLIP relevance score, isolating the subset-aware terms. Duration split: +6.31 medium, +7.80 long, ~ -0.7 short vs uniform. Component ablation shows relevance is the dominant term and coverage/redundancy alone are weak. All numbers from the authors; no independent replication, and open-ended Video-TT is won by a competing method.
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 7-8B multimodal models answering multiple-choice long-video questions under a fixed 16-frame budget, selecting keyframes by embedding similarity to the query concatenated with each candidate answer beats query-only scoring, query-agnostic diversity selection, and uniform sampling across Qwen2-VL, LLaVA-Video and InternVL3.5, with the largest margins on the longest video splits — but per-category results still invert, including uniform sampling winning some categories.Reasoning about time in video · reviewed by AI
- For long-video QA under a fixed 32-frame budget with 7B-8B multimodal backbones, selecting frames from the structural shape of the frame-query similarity curve (peak, rising, falling, transition regions) beats ranking frames by similarity magnitude, and the margin grows with video length — a few points on short clips and roughly 8 points on hour-long videos — while on short/medium-clip benchmarks it is near-indistinguishable from other selectors.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
- In training-free video temporal grounding with frozen vision-language embedding backbones, matching a query only against frames sampled at a fixed rate leaves temporal context unused: adding clip-level (sliding-window) similarities projected onto the frame timeline raises localization accuracy even though clip-only similarity is worse than frame-only on its own.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 selection method that adds coverage and redundancy terms to the same relevance score fails to beat Top-K on the same backbone and budget, or the gap over uniform sampling does not widen on longer videos. Proposed technique, not catalogued: marginal-utility greedy keyframe selection.