# CrossHOI-Bench: A Unified Benchmark for HOI Evaluation across Vision-Language Models and HOI-Specific Methods

arXiv: https://arxiv.org/abs/2508.18753 · HTML: https://mlsanity.com/paper/2508.18753/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in Driving VLMs and open-vocabulary detection, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 65 of its 81 references that are in the map.
- First posted on arXiv 26 Aug 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Driving VLMs and open-vocabulary detection](https://mlsanity.com/mllm/driving-vlms-and-open-vocabulary-detection/)

## Builds on (most cited first, 30 of 65)

- [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)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [End-to-End Object Detection with Transformers](https://mlsanity.com/paper/2005.12872/) (2020)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [SQuAD: 100,000+ Questions for Machine Comprehension of Text](https://mlsanity.com/paper/1606.05250/) (2016)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks](https://mlsanity.com/paper/2312.14238/) (2023)
- [SlowFast Networks for Video Recognition](https://mlsanity.com/paper/1812.03982/) (2018)
- [InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models](https://mlsanity.com/paper/2504.10479/) (2025)
- [Ego4D: Around the World in 3,000 Hours of Egocentric Video](https://mlsanity.com/paper/2110.07058/) (2021)
- [Deep High-Resolution Representation Learning for Human Pose Estimation](https://mlsanity.com/paper/1902.09212/) (2019)
- [Large Language Diffusion Models](https://mlsanity.com/paper/2502.09992/) (2025)
- [The Dawn of LMMs: Preliminary Explorations with GPT-4V(ision)](https://mlsanity.com/paper/2309.17421/) (2023)
- [Large Language Models Are Not Robust Multiple Choice Selectors](https://mlsanity.com/paper/2309.03882/) (2023)
- [AVA: A Video Dataset of Spatio-temporally Localized Atomic Visual Actions](https://mlsanity.com/paper/1705.08421/) (2017)
- [Large Language Models Sensitivity to The Order of Options in Multiple-Choice Questions](https://mlsanity.com/paper/2308.11483/) (2023)
- [Learning to Detect Human-Object Interactions](https://mlsanity.com/paper/1702.05448/) (2017)
- [Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization](https://mlsanity.com/paper/2411.10442/) (2024)
- [HOTR: End-to-End Human-Object Interaction Detection with Transformers](https://mlsanity.com/paper/2104.13682/) (2021)
- [QPIC: Query-Based Pairwise Human-Object Interaction Detection with Image-Wide Contextual Information](https://mlsanity.com/paper/2103.05399/) (2021)
- [End-to-End Human Object Interaction Detection with HOI Transformer](https://mlsanity.com/paper/2103.04503/) (2021)
- [Visual Compositional Learning for Human-Object Interaction Detection](https://mlsanity.com/paper/2007.12407/) (2020)
- [Learning Human-Object Interaction Detection using Interaction Points](https://mlsanity.com/paper/2003.14023/) (2020)
- [Efficient Reasoning Models: A Survey](https://mlsanity.com/paper/2504.10903/) (2025)
- [Detailed 2D-3D Joint Representation for Human-Object Interaction](https://mlsanity.com/paper/2004.08154/) (2020)
- [Spatially Conditioned Graphs for Detecting Human-Object Interactions](https://mlsanity.com/paper/2012.06060/) (2020)

## Built on by (most cited first, 5 of 5)

- [Hy-Embodied-VLM-1.0: Efficient Physical-World Agents](https://mlsanity.com/paper/2607.12894/) (2026)
- [SHOE: Semantic HOI Open-Vocabulary Evaluation Metric](https://mlsanity.com/paper/2604.01586/) (2026)
- [A Study of Failure Modes in Two-Stage Human-Object Interaction Detection](https://mlsanity.com/paper/2604.13448/) (2026)
- [Token-Based Affordance Grounding with Large Vision-Language Models](https://mlsanity.com/paper/2607.03595/) (2026)
- [HEIR: Learning Human-Entity Interactions with Functional Roles](https://mlsanity.com/paper/2609.35955/) (2026)

## Abstract (arXiv)

HOI detection has long been dominated by task-specific models, sometimes with early vision-language backbones such as CLIP. With the rise of large generative VLMs, a key question is whether standalone VLMs can perform HOI detection competitively against specialized HOI methods. Existing benchmarks such as HICO-DET require exact label matching under incomplete annotations, so any unmatched prediction is marked wrong. This unfairly penalizes valid outputs, especially from less constrained VLMs, and makes cross-paradigm comparison unreliable. To address this limitation, we introduce CrossHOI-Bench, a multiple-choice HOI benchmark with explicit positives and curated negatives, enabling unified and reliable evaluation of both VLMs and HOI-specific models. We further focus on challenging scenarios, such as multi-person scenes and fine-grained interaction distinctions, which are crucial for revealing real differences between the two paradigms. Experiments show that large VLMs achieve competitive, sometimes superior, zero-shot performance, yet they struggle with multiple concurrent actions and with correctly assigning interactions to the target person. Conversely, HOI-specific methods remain weaker in general HOI reasoning but demonstrate stronger multi-action recognition and more reliable identification of which person performs which action. These findings expose complementary strengths and weaknesses of VLMs and HOI-specific methods, which existing benchmarks fail to reveal due to incorrect penalization.