Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Jiwanta, I Putu Andika Bagas, Purwarianti, Ayu
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2512.12246
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909959795507200
author Jiwanta, I Putu Andika Bagas
Purwarianti, Ayu
author_facet Jiwanta, I Putu Andika Bagas
Purwarianti, Ayu
contents Detecting video moments and highlights from natural-language queries have been unified by transformer-based methods. Other works use generative Multimodal LLM (MLLM) to predict moments and/or highlights as text timestamps, utilizing its reasoning capability. While effective, text-based generation cannot provide direct gradients for frame-level predictions because the model only emits language tokens. Although recent Reinforcement Learning (RL) methods attempt to address the issue, we propose a novel approach by applying segmentation objectives directly on the LLM's output tokens. The LLM is fed with a fixed number of frames alongside a prompt that enforces it to output a sequence of continuous "0" and/or "1" characters, with one character per frame. The "0"/"1" characters benefit from the LLM's inherent language capability while also acting as background and foreground probabilities, respectively. Training employs segmentation losses on the probabilities alongside a normal causal LM loss. At inference, beam search generates sequence and logits, acting as moments and saliency scores, respectively. Despite sampling only 25 frames -- less than half of comparable methods -- our method achieved strong highlight detection (56.74 HIT@1) on QVHighlights. Additionally, our efficient method scores above the baseline (35.28 MAP) for moment retrieval. Empirically, segmentation losses provide a stable complementary learning signal even when the causal LM loss plateaus.
format Preprint
id arxiv_https___arxiv_org_abs_2512_12246
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Moment and Highlight Detection via MLLM Frame Segmentation
Jiwanta, I Putu Andika Bagas
Purwarianti, Ayu
Computer Vision and Pattern Recognition
Detecting video moments and highlights from natural-language queries have been unified by transformer-based methods. Other works use generative Multimodal LLM (MLLM) to predict moments and/or highlights as text timestamps, utilizing its reasoning capability. While effective, text-based generation cannot provide direct gradients for frame-level predictions because the model only emits language tokens. Although recent Reinforcement Learning (RL) methods attempt to address the issue, we propose a novel approach by applying segmentation objectives directly on the LLM's output tokens. The LLM is fed with a fixed number of frames alongside a prompt that enforces it to output a sequence of continuous "0" and/or "1" characters, with one character per frame. The "0"/"1" characters benefit from the LLM's inherent language capability while also acting as background and foreground probabilities, respectively. Training employs segmentation losses on the probabilities alongside a normal causal LM loss. At inference, beam search generates sequence and logits, acting as moments and saliency scores, respectively. Despite sampling only 25 frames -- less than half of comparable methods -- our method achieved strong highlight detection (56.74 HIT@1) on QVHighlights. Additionally, our efficient method scores above the baseline (35.28 MAP) for moment retrieval. Empirically, segmentation losses provide a stable complementary learning signal even when the causal LM loss plateaus.
title Moment and Highlight Detection via MLLM Frame Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.12246