Pith. sign in

REVIEW 15 cited by

SlowFast-LLaVA: A Strong Training-Free Baseline for Video Large Language Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2407.15841 v2 pith:4C3XX3TH submitted 2024-07-22 cs.CV

classification cs.CV
keywords videospatialfeaturesllmstraining-freecapturedesigndetailed
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

We propose SlowFast-LLaVA (or SF-LLaVA for short), a training-free video large language model (LLM) that can jointly capture detailed spatial semantics and long-range temporal context without exceeding the token budget of commonly used LLMs. This is realized by using a two-stream SlowFast design of inputs for Video LLMs to aggregate features from sampled frames in an effective way. Specifically, the Slow pathway extracts features at a low frame rate while keeping as much spatial detail as possible (e.g., with 12x24 tokens), and the Fast pathway operates on a high frame rate but uses a larger spatial pooling stride (e.g., downsampling 6x) to focus on the motion cues. As a result, this design allows us to adequately capture both spatial and temporal features that are beneficial for detailed video understanding. Experimental results show that SF-LLaVA outperforms existing training-free methods on a wide range of video tasks. On some benchmarks, it achieves comparable or even better performance compared to state-of-the-art Video LLMs that are fine-tuned on video datasets. Code has been made available at: https://github.com/apple/ml-slowfast-llava.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 15 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. TrajTok: Learning Trajectory Tokens enables better Video Understanding

    cs.CV 2026-02 unverdicted novelty 7.0 of 10

    TrajTok learns to tokenize video into object-trajectory tokens end-to-end, improving video CLIP, probing, and VLM performance over patch and token-merging baselines.

  2. SkyVLaM: Multimodal Large Language Model for UAV Video Understanding in Remote Sensing

    cs.CV 2026-07 conditional novelty 6.0 of 10

    SkyVLaM introduces a temporal basis perceiver and adaptive dense selection to improve language-conditioned video segmentation in UAV scenes, and contributes the SkyVid dataset.

  3. Efficient Frame Selection for Long Videos at Test Time with Attention-Based MLLM Selectors

    cs.CV 2026-07 conditional novelty 6.0 of 10

    Attention maps from a small MLLM can serve as a training-free, query-conditioned frame selector, improving long-video QA accuracy under fixed frame budgets.

  4. Strefer: Empowering Video LLMs with Space-Time Referring and Reasoning via Synthetic Instruction Data

    cs.CV 2025-09 conditional novelty 6.0 of 10

    Adding Strefer's synthetic space-time reference questions to video instruction tuning improves mask-referred description/QA, timestamp QA, and temporal reasoning over a video-LLM baseline.

  5. IPFormer-VideoLLM: Enhancing Multi-modal Video Understanding for Multi-shot Scenes

    cs.CV 2025-06 conditional novelty 6.0 of 10

    A new multi-shot video dataset and an instance-prompt video LLM report large gains, but the main benchmark is built by the same authors and the model is not released.

  6. Threading Keyframe with Narratives: MLLMs as Strong Long Video Comprehenders

    cs.CV 2025-05 conditional novelty 6.0 of 10

    Nar-KFC improves long-video question answering by interleaving query-relevant diverse keyframes with captions of non-keyframes, raising accuracy on Video-MME, LongVideoBench, and MLVU across 7B MLLMs.

  7. Clapper: Compact Learning and Video Representation in VLMs

    cs.CV 2025-05 conditional novelty 6.0 of 10

    Clapper achieves 13x visual token compression in video VLMs with maintained or improved QA accuracy using a slow-fast representation and a TimePerceiver module.

  8. UniGen: Enhanced Training & Test-Time Strategies for Unified Multimodal Understanding and Generation

    cs.CV 2025-05 conditional novelty 6.0 of 10

    UniGen shows a 1.5B model trained on open data can beat larger systems on image understanding and generation once it verifies its own outputs with chain-of-thought and Best-of-N selection.

  9. Breaking Down Video LLM Benchmarks: Knowledge, Spatial Perception, or True Temporal Understanding?

    cs.CV 2025-05 conditional novelty 6.0 of 10

    Video LLM benchmark scores are inflated by language-prior and static-frame questions; VBenchComp separates those from true temporal questions and shows a trimmed core subset preserves rankings.

  10. Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model

    cs.CV 2026-07 conditional novelty 5.0 of 10

    Codec-guided sparse patch selection plus a lightweight speak/silent gate yields a 4B streaming VLM that is competitive on static tasks, stronger on video/spatial benchmarks, and much cheaper at inference.

  11. IntentVCNet: Bridging Spatio-Temporal Gaps for Intention-Oriented Controllable Video Captioning

    cs.CV 2025-07 conditional novelty 5.0 of 10

    IntentVCNet uses per-frame object coordinates, red-box visual prompts, and a lightweight box adapter to make video captioning focus on a user-selected object, reporting 225.19 CIDEr on the IntentVC public test set.

  12. MANTA: Cross-Modal Semantic Alignment and Information-Theoretic Optimization for Long-form Multimodal Understanding

    cs.CV 2025-06 reject novelty 5.0 of 10

    A multimodal retrieval pipeline that projects video and audio into text and claims near-optimal context selection, with reported gains of up to 22.6% on Video-MME that rest on circular theory and unreleased data.

  13. MOOSE: Pay Attention to Temporal Dynamics for Video Understanding via Optical Flows

    cs.CV 2025-06 reject novelty 5.0 of 10

    MOOSE fuses frozen DINOv2 image features with optical flow features via cross-attention and causal aggregation, reporting 70.84% Kinetics-400 and 65.23% SSv2 top-1 accuracy, alongside interpretable attention maps.

  14. APB-V: Accelerating Long-Video Understanding via Sequence-Parallelism-aware Approximate Attention

    cs.CV 2026-01 conditional novelty 4.0 of 10

    SPAVA compresses each GPU's video context into essential key-values for sharing, making multi-GPU long-video prefilling 1.18x faster than APB and 12.72x faster than single-GPU FlashAttention with near-equal accuracy.

  15. Region-Aware Multimodal Large Language Model via SlowFast Tokenization and Pseudo-Mask Guidance for 3D CT Report Generation

    eess.IV 2025-06 conditional novelty 4.0 of 10

    MedRegion-CT integrates region-representative tokens, mask-driven segmentation tokens, and patient-specific attribute prompts into a multimodal LLM, reporting state-of-the-art scores on RadGenome-Chest CT report generation.

Pith tools