TimeLens: Rethinking Video Temporal Grounding with Multimodal LLMs

TL;DR

TimeLens systematically improves video temporal grounding by addressing data quality issues and exploring algorithmic design, resulting in models that outperform existing solutions and setting new standards in open-source VTG performance.

Contribution

The paper introduces TimeLens, a comprehensive framework that enhances VTG by re-annotating benchmarks, creating high-quality training data, and developing effective algorithmic strategies.

Findings

TimeLens models achieve state-of-the-art VTG performance.

Re-annotated benchmarks reveal unreliability of previous evaluations.

High-quality training data significantly improve model accuracy.

Abstract

This paper does not introduce a novel method but instead establishes a straightforward, incremental, yet essential baseline for video temporal grounding (VTG), a core capability in video understanding. While multimodal large language models (MLLMs) excel at various video understanding tasks, the recipes for optimizing them for VTG remain under-explored. In this paper, we present TimeLens, a systematic investigation into building MLLMs with strong VTG ability, along two primary dimensions: data quality and algorithmic design. We first expose critical quality issues in existing VTG benchmarks and introduce TimeLens-Bench, comprising meticulously re-annotated versions of three popular benchmarks with strict quality criteria. Our analysis reveals dramatic model re-rankings compared to legacy benchmarks, confirming the unreliability of prior evaluation standards. We also address noisy training data through an automated re-annotation pipeline, yielding TimeLens-100K, a large-scale, high-quality training dataset. Building on our data foundation, we conduct in-depth explorations of algorithmic design principles, yielding a series of meaningful insights and effective yet efficient practices. These include interleaved textual encoding for time representation, a thinking-free reinforcement learning with verifiable rewards (RLVR) approach as the training paradigm, and carefully designed recipes for RLVR training. These efforts culminate in TimeLens models, a family of MLLMs with state-of-the-art VTG performance among open-source models and even surpass proprietary models such as GPT-5 and Gemini-2.5-Flash. All codes, data, and models will be released to facilitate future research.