# PAS: A Training-Free Stabilizer for Temporal Encoding in Video LLMs

arXiv: https://arxiv.org/abs/2511.10979 · HTML: https://mlsanity.com/paper/2511.10979/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 27 references that are in the map.
- First posted on arXiv 14 Nov 2025.

## Builds on (most cited first, 17 of 17)

- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [SlowFast Networks for Video Recognition](https://mlsanity.com/paper/1812.03982/) (2018)
- [Is Space-Time Attention All You Need for Video Understanding?](https://mlsanity.com/paper/2102.05095/) (2021)
- [GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints](https://mlsanity.com/paper/2305.13245/) (2023)
- [The "something something" video database for learning and evaluating visual common sense](https://mlsanity.com/paper/1706.04261/) (2017)
- [EgoSchema: A Diagnostic Benchmark for Very Long-form Video Language Understanding](https://mlsanity.com/paper/2308.09126/) (2023)
- [Rotary Position Embedding for Vision Transformer](https://mlsanity.com/paper/2403.13298/) (2024)
- [MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding](https://mlsanity.com/paper/2406.14515/) (2024)
- [SlowFast-LLaVA: A Strong Training-Free Baseline for Video Large Language Models](https://mlsanity.com/paper/2407.15841/) (2024)
- [A Short Note on the Kinetics-700-2020 Human Action Dataset](https://mlsanity.com/paper/2010.10864/) (2020)
- [VideoRoPE: What Makes for Good Video Rotary Position Embedding?](https://mlsanity.com/paper/2502.05173/) (2025)
- [Large-scale Robustness Analysis of Video Action Recognition Models](https://mlsanity.com/paper/2207.01398/) (2022)
- [Temporal Coherent Test-Time Optimization for Robust Video Classification](https://mlsanity.com/paper/2302.14309/) (2023)
- [TS-LLaVA: Constructing Visual Tokens through Thumbnail-and-Sampling for Training-Free Video Large Language Models](https://mlsanity.com/paper/2411.11066/) (2024)
- [PosMLP-Video: Spatial and Temporal Relative Position Encoding for Efficient Video Recognition](https://mlsanity.com/paper/2407.02934/) (2024)

## Built on by (most cited first, 1 of 1)

- [Stable Scores, Unstable Answers: Frame Phase and Option Order in Video Multiple-Choice Evaluation](https://mlsanity.com/paper/2610.08649/) (2026)

## Abstract (arXiv)

Video LLMs suffer from temporal inconsistency: small shifts in frame timing can flip attention and suppress relevant frames. We trace this instability to the common extension of Rotary Position Embeddings to video through multimodal RoPE. The induced inverse Fourier time kernel exhibits frame-scale ripples that multiply adjacent frames by different factors, which perturbs attention that should otherwise be governed by the raw query key inner product. We present Phase Aggregated Smoothing (PAS), a simple, training-free mechanism that applies small opposed phase offsets across heads and then aggregates their outputs. PAS preserves the per-head spectrum magnitude, while the aggregation effectively smooths the temporal kernel and reduces phase sensitivity without changing the positional encoding structure. Our analysis shows that the RoPE rotated logit can be approximated as a content dot product scaled by a time kernel; smoothing this kernel yields Lipschitz stability of attention to small temporal shifts; multi phase averaging attenuates high frequency ripples while preserving per-head spectra under Nyquist-valid sampling. Experiments on multiple video understanding benchmarks under matched token budgets show consistent improvements with negligible computational overhead. PAS provides a plug and play upgrade for robust temporal encoding in Video LLMs.