# What Do Interaction Representations Actually Measure? Pre-Event Separability in Weakly-Supervised Violence Detection

arXiv: https://arxiv.org/abs/2608.27879 · HTML: https://mlsanity.com/paper/2608.27879/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 16 references that are in the map.
- First posted on arXiv 28 Aug 2026.

## Builds on (most cited first, 12 of 12)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Shortcut Learning in Deep Neural Networks](https://mlsanity.com/paper/2004.07780/) (2020)
- [Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition](https://mlsanity.com/paper/1801.07455/) (2018)
- [Real-world Anomaly Detection in Surveillance Videos](https://mlsanity.com/paper/1801.04264/) (2018)
- [Revisiting Skeleton-based Action Recognition](https://mlsanity.com/paper/2104.13586/) (2021)
- [Not only Look, but also Listen: Learning Multimodal Violence Detection under Weak Supervision](https://mlsanity.com/paper/2007.04687/) (2020)
- [Weakly-supervised Video Anomaly Detection with Robust Temporal Feature Magnitude Learning](https://mlsanity.com/paper/2101.10030/) (2021)
- [MIST: Multiple Instance Self-Training Framework for Video Anomaly Detection](https://mlsanity.com/paper/2104.01633/) (2021)
- [Hierarchical Deep Temporal Models for Group Activity Recognition](https://mlsanity.com/paper/1607.02643/) (2016)
- [Learning Actor Relation Graphs for Group Activity Recognition](https://mlsanity.com/paper/1904.10117/) (2019)
- [Unbiased Multiple Instance Learning for Weakly Supervised Video Anomaly Detection](https://mlsanity.com/paper/2303.12369/) (2023)
- [RWF-2000: An Open Large Scale Video Database for Violence Detection](https://mlsanity.com/paper/1911.05913/) (2019)

## Abstract (arXiv)

Articulated human pose provides detailed body-configuration information beyond coarse spatial relationships, but whether this detail yields greater discriminative information when the downstream pipeline is held fixed remains unclear. We examine this through early violence detection. Holding the tracker, temporal head, supervision, folds, and evaluation fixed, we compare five interaction representations spanning coarse bounding-box geometry, a matched handcrafted pose analogue, enriched pose descriptors, and a matched-capacity encoder learned from raw joints, under video-level evaluation with cluster-bootstrap intervals. No pose-based representation outperforms coarse geometry, though with fifteen anomalous videos this subset cannot rule out small effects. Extending the pipeline to frozen visual encoders, and repeating the comparison on XD-Violence (137 anomalous videos, nine times our UCF-Crime sample), person-crop appearance and whole-frame context both exceed geometry by a wide margin, yet context matches appearance on UCF-Crime and exceeds it on the larger split: cropping to the interacting people yields no advantage over encoding the whole frame. This prompts a direct test of what the benchmark measures. Scoring anomalous videos using only frames preceding the annotated onset, under a control removing sequence length as a cue, retains 39-91% of above-chance separation on both benchmarks, including for seven hand-designed geometric channels. Inspection of the tightest pre-onset windows identifies concrete provenance artifacts: editorial title cards and platform watermarks absent from the surveillance footage supplying the normal class. Video-level AUC here is thus a composite of event evidence and pre-event source cues, a shared source of discrimination that can obscure differences between representations. The diagnostic requires only annotations these benchmarks already ship.