# TRACE: Evidence Grounding-Guided Multi-Video Event Understanding and Claim Generation

arXiv: https://arxiv.org/abs/2605.16740 · HTML: https://mlsanity.com/paper/2605.16740/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 33 references that are in the map.
- First posted on arXiv 16 May 2026.

## Builds on (most cited first, 20 of 20)

- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection](https://mlsanity.com/paper/2303.05499/) (2023)
- [Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection](https://mlsanity.com/paper/2310.11511/) (2023)
- [Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysis](https://mlsanity.com/paper/2405.21075/) (2024)
- [Video-LLaVA: Learning United Visual Representation by Alignment Before Projection](https://mlsanity.com/paper/2311.10122/) (2023)
- [Improving language models by retrieving from trillions of tokens](https://mlsanity.com/paper/2112.04426/) (2021)
- [HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face](https://mlsanity.com/paper/2303.17580/) (2023)
- [LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding](https://mlsanity.com/paper/2407.15754/) (2024)
- [ViperGPT: Visual Inference via Python Execution for Reasoning](https://mlsanity.com/paper/2303.08128/) (2023)
- [MM-REACT: Prompting ChatGPT for Multimodal Reasoning and Action](https://mlsanity.com/paper/2303.11381/) (2023)
- [Visual Programming: Compositional visual reasoning without training](https://mlsanity.com/paper/2211.11559/) (2022)
- [Scene Text Visual Question Answering](https://mlsanity.com/paper/1905.13648/) (2019)
- [YOLOv12: Attention-Centric Real-Time Object Detectors](https://mlsanity.com/paper/2502.12524/) (2025)
- [PP-OCR: A Practical Ultra Lightweight OCR System](https://mlsanity.com/paper/2009.09941/) (2020)
- [QVHighlights: Detecting Moments and Highlights in Videos via Natural Language Queries](https://mlsanity.com/paper/2107.09609/) (2021)
- [Artemis: Towards Referential Understanding in Complex Videos](https://mlsanity.com/paper/2406.00258/) (2024)
- [MultiVENT 2.0: A Massive Multilingual Benchmark for Event-Centric Video Retrieval](https://mlsanity.com/paper/2410.11619/) (2024)
- [ChartReformer: Natural Language-Driven Chart Image Editing](https://mlsanity.com/paper/2403.00209/) (2024)
- [WikiVideo: Article Generation from Multiple Videos](https://mlsanity.com/paper/2504.00939/) (2025)
- [Unified Multimodal Uncertain Inference](https://mlsanity.com/paper/2604.08701/) (2026)

## Built on by (most cited first, 3 of 3)

- [OSEF: One-Step Evidence Fusion for Cross-Video Scene Procedure Planning](https://mlsanity.com/paper/2607.29401/) (2026)
- [Findings of the MAGMaR 2026 Shared Task](https://mlsanity.com/paper/2606.12295/) (2026)
- [MultiVENT-Raw: A Benchmark for Retrieval and Reasoning over Raw Videos](https://mlsanity.com/paper/2609.28437/) (2026)

## Abstract (arXiv)

Multi-video event understanding demands models that can locate and attribute query-relevant evidence scattered across long, heterogeneous video corpora. Existing large vision-language models (LVLMs) often underperform in this regime because they quickly exhaust their context budget and struggle to precisely localize evidentially important segments, frequently missing dense informational cues such as broadcast graphics, subtitles, and scoreboards. We introduce TRACE, an evidence grounding-guided framework that follows a ground-before-reasoning strategy for multi-video event reasoning. Our approach first builds a structured, text-searchable timeline for each video using OCR and object detection. A text-only LLM then conducts query-aware evidence localization, selecting relevant moments prior to any downstream visual reasoning. The retrieved frames and their grounding summaries are subsequently used to steer LVLM-based claim generation and cross-video citation consolidation. Experiments on MAGMaR 2026 and WikiVideo demonstrate that structured grounding markedly boosts factual completeness and attribution fidelity. On the MAGMaR validation split, TRACE raises macro-average MiRAGE F1 from 0.705 to 0.811 compared to an unguided Qwen3-VL-30B baseline, with especially strong improvements in citation recall from 0.440 to 0.628. The method also attains state-of-the-art results on the official MAGMaR 2026 leaderboard. Code is released at https://github.com/pengyu965/TRACE.