# VELOCITI: Benchmarking Video-Language Compositional Reasoning with Strict Entailment

arXiv: https://arxiv.org/abs/2406.10889 · HTML: https://mlsanity.com/paper/2406.10889/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 35 of its 55 references that are in the map.
- First posted on arXiv 16 Jun 2024.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)

## Builds on (most cited first, 30 of 35)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [Evaluating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2305.10355/) (2023)
- [Video-LLaVA: Learning United Visual Representation by Alignment Before Projection](https://mlsanity.com/paper/2311.10122/) (2023)
- [MVBench: A Comprehensive Multi-modal Video Understanding Benchmark](https://mlsanity.com/paper/2311.17005/) (2023)
- [Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs](https://mlsanity.com/paper/2401.06209/) (2024)
- [When and why vision-language models behave like bags-of-words, and what to do about it?](https://mlsanity.com/paper/2210.01936/) (2022)
- [Hollywood in Homes: Crowdsourcing Data Collection for Activity Understanding](https://mlsanity.com/paper/1604.01753/) (2016)
- [Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality](https://mlsanity.com/paper/2204.03162/) (2022)
- [Evaluating Text-to-Visual Generation with Image-to-Text Generation](https://mlsanity.com/paper/2404.01291/) (2024)
- [TempCompass: Do Video LLMs Really Understand Videos?](https://mlsanity.com/paper/2403.00476/) (2024)
- [PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioning](https://mlsanity.com/paper/2404.16994/) (2024)
- [Perception Test: A Diagnostic Benchmark for Multimodal Video Models](https://mlsanity.com/paper/2305.13786/) (2023)
- [SugarCrepe: Fixing Hackable Benchmarks for Vision-Language Compositionality](https://mlsanity.com/paper/2306.14610/) (2023)
- [STAR: A Benchmark for Situated Reasoning in Real-World Videos](https://mlsanity.com/paper/2405.09711/) (2024)
- [CREPE: Can Vision-Language Foundation Models Reason Compositionally?](https://mlsanity.com/paper/2212.07796/) (2022)
- [Fine-tuned CLIP Models are Efficient Video Learners](https://mlsanity.com/paper/2212.03640/) (2022)
- [Revisiting the "Video" in Video-Language Understanding](https://mlsanity.com/paper/2206.01720/) (2022)
- [AGQA: A Benchmark for Compositional Spatio-Temporal Reasoning](https://mlsanity.com/paper/2103.16002/) (2021)
- [VL-CheckList: Evaluating Pre-trained Vision-Language Models with Objects, Attributes and Relations](https://mlsanity.com/paper/2207.00221/) (2022)
- [CinePile: A Long Video Question Answering Dataset and Benchmark](https://mlsanity.com/paper/2405.08813/) (2024)
- [Does CLIP Bind Concepts? Probing Compositionality in Large Image Models](https://mlsanity.com/paper/2212.10537/) (2022)
- [SEED-Bench-2: Benchmarking Multimodal Large Language Models](https://mlsanity.com/paper/2311.17092/) (2023)
- [Visual Semantic Role Labeling for Video Understanding](https://mlsanity.com/paper/2104.00990/) (2021)
- [VITATECS: A Diagnostic Dataset for Temporal Concept Understanding of Video-Language Models](https://mlsanity.com/paper/2311.17404/) (2023)
- [COLA: A Benchmark for Compositional Text-to-image Retrieval](https://mlsanity.com/paper/2305.03689/) (2023)
- [Test of Time: Instilling Video-Language Models with a Sense of Time](https://mlsanity.com/paper/2301.02074/) (2023)

## Built on by (most cited first, 10 of 10)

- [Vinoground: Scrutinizing LMMs over Dense Temporal Reasoning with Short Videos](https://mlsanity.com/paper/2410.02763/) (2024)
- [Know-Show: Benchmarking Video-Language Models on Spatio-Temporal Grounded Reasoning](https://mlsanity.com/paper/2512.05513/) (2025)
- [AMUSE: Audio-Visual Benchmark and Alignment Framework for Agentic Multi-Speaker Understanding](https://mlsanity.com/paper/2512.16250/) (2025)
- [What Happens When: Learning Temporal Orders of Events in Videos](https://mlsanity.com/paper/2512.08979/) (2025)
- [NarrativeTrack: Evaluating Entity-Centric Reasoning for Narrative Understanding](https://mlsanity.com/paper/2601.01095/) (2026)
- [VideoKR: Towards Knowledge- and Reasoning-Intensive Video Understanding](https://mlsanity.com/paper/2606.05259/) (2026)
- [Estimating Commonsense Plausibility through Semantic Shifts](https://mlsanity.com/paper/2502.13464/) (2025)
- [MINERVA-Cultural: A Benchmark for Cultural and Multilingual Long Video Reasoning](https://mlsanity.com/paper/2601.10649/) (2026)
- [Adapting MLLMs for Nuanced Video Retrieval](https://mlsanity.com/paper/2512.13511/) (2025)
- [Cross-modal Identity Mapping: Minimizing Information Loss in Modality Conversion via Reinforcement Learning](https://mlsanity.com/paper/2603.01696/) (2026)

## Abstract (arXiv)

A fundamental aspect of compositional reasoning in a video is associating people and their actions across time. Recent years have seen great progress in general-purpose vision or video models and a move towards long-video understanding. While exciting, we take a step back and ask: are current models good at compositional reasoning on short videos? To this end, we introduce VELOCITI, a benchmark to study Video-LLMs by disentangling and assessing the comprehension of agents, actions, and their associations across multiple events. We adopt the Video-Language Entailment setup and propose StrictVLE that requires correct classification (rather than ranking) of the positive and negative caption. We evaluate several models and observe that even the best, LLaVA-OneVision (44.5%) and Gemini-1.5-Pro (49.3%), are far from human accuracy at 93.0%. Results show that action understanding lags behind agents, and negative captions created using entities appearing in the video perform worse than those obtained from pure text manipulation. We also present challenges with ClassicVLE and multiple-choice (MC) evaluation, strengthening our preference for StrictVLE. Finally, we validate that our benchmark requires visual inputs of multiple frames making it ideal to study video-language compositional reasoning.