skills.rest
JiusiServe
JiusiServeOFFICIAL
Explore all skills by this author →
GitHub

lvsa-reproduce-paper

Automates LVSA paper result reproduction using bundled benchmarks and scripts.

18|4|Updated May 21, 2026
Search Tags:#reproduce#benchmarks#lvsa#latency-scaling#vqeval#vbench-long#paper-results
One-click install
npx skills add https://github.com/JiusiServe/LongVideoSparseAttention --skill lvsa-reproduce-paper
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: lvsa-reproduce-paper
Source: https://github.com/JiusiServe/LongVideoSparseAttention/tree/main/skills/lvsa-reproduce-paper
Command: npx skills add https://github.com/JiusiServe/LongVideoSparseAttention --skill lvsa-reproduce-paper

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Reproduce LVSA paper results using bundled benchmarks and scripts, enabling verification of SotA comparisons and latency scaling without reproducing all resources manually.

Core Features & Use Cases

  • Reproduce headline LVSA results as shown in the paper, including SotA benchmarks, latency scaling, and quality scoring using VQeval and VBench-Long.
  • Validate results across Wan 2.1 1.3B and HunyuanVideo setups using the bundled benchmarks and scripts.
  • Use the provided workflows to regenerate figures in the README and compare against baselines.

Quick Start

Run the bundled benchmarks and scripts to reproduce the LVSA paper results.

Frequently Asked Questions about lvsa-reproduce-paper

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reproduce LVSA paper results for SotA benchmarks and latency scaling?▼

Reproduce LVSA paper results by running bundled benchmarks and scripts that automate SotA comparisons, latency scaling, and scoring across Wan 1.3B and HunyuanVideo setups without manual resource recreation.

What benchmarks are included for verifying LVSA video quality scoring?▼

The LVSA reproduction package bundles VQeval and VBench-Long benchmarks to validate video quality scoring and SotA comparison results shown in the original paper.

Do I need Python 3.12 and external model weights to run the LVSA reproduction scripts?▼

Yes, running the LVSA reproduction scripts requires a Python 3.12 environment with project dependencies, the bundled benchmarks/ and scripts/ directories, and external model weights downloaded separately.

Can I regenerate figures from the LVSA README using the bundled workflows?▼

Yes, the bundled LVSA workflows allow you to regenerate figures from the README and compare results against baselines across Wan 2.1 1.3B and HunyuanVideo model setups.

What is the best way to validate LVSA latency scaling without reproducing all resources manually?▼

Use the bundled LVSA benchmarks and scripts to automate latency scaling validation and SotA comparisons across supported model setups, bypassing the need to manually recreate experimental resources.

Are there limitations when reproducing LVSA results across different video generation model setups?▼

The LVSA reproduction scripts currently validate results specifically across Wan 2.1 1.3B and HunyuanVideo setups, and require separately downloading external model weights not included in the bundle.

Related Skills

Explore More Education & Research Skills→
JiusiServeJiusiServe
18

lvsa-quickstart

Install LVSA, select a backend, and generate a first long video.

Official
Intermediate
clawpod-appclawpod-app
46

incident-response

Analyze production incident metrics and output rollback recommendations.

Community
Intermediate
JiusiServeJiusiServe
18

lvsa-troubleshooting

Diagnose LVSA failure modes and guide remediation across vLLM-Omni, Wan, HunyuanVideo, and CogVideoX pipelines.

Official
Advanced
raja21068raja21068
2

paper-autoraters

Score paper drafts with citation F1 and literature-review quality JSON outputs.

Community
Intermediate