ParaVT — security grade SAFE, quality 71/100

Security audit verdict: SAFE · quality 71/100

No red flags found in any of the 11 categories — no credential harvesting, no data exfiltration, no curl-pipe-shell installer. Scanned against the SlowMist agent-security taxonomy, refreshed every 8 hours. Full audit →

by EvolvingLMMs-Lab · Agent Tool · ★ 55

Last updated: · Indexed by AgentSkillsHub · Auto-synced every 8h

🔒 Is ParaVT safe to install? View the security audit →

About ParaVT

ParaVT Taming the Tool Prior Paradox for Parallel Tool Use in Agentic Video Reinforcement Learning Overview Long-video understanding is increasingly framed as agentic video reasoning: a large multimodal model (LMM) post-trained with reinforcement learning to invoke video-processing tools. Prior work in this line, including our earlier LongVT (CVPR 2026), dispatches tool calls sequentially — brittle to single mis-localizations, prone to multi-turn context drift, and linear in cost. <img src="assets/method.png" width="620" alt="ParaVT architecture: sequential one-tool-per-turn (left) vs parallel single-turn dispatch with weight-sharin

agentic-rlgrpolong-video-understandingmultimodal-rlreinforcement-learningtool-usevideo-llm

Quick Facts

Stars55
Forks2
LanguagePython
CategoryAgent Tool
LicenseApache-2.0
Quality Score70.7349419360204/100
Open Issues2
Last Updated2026-06-02
Created2026-04-28
Platformspython
Est. Tokens~326k

Compatible Skills

These tools work well together with ParaVT for enhanced workflows:

  • SkillZero — semantic(0.56)+complementary+same_lang+shared_platform (60%)
  • llm-rl-environments-lil-course — semantic(0.45)+complementary+rare_topics+same_lang+shared_platform (60%)
  • AgentFly — semantic(0.57)+complementary+rare_topics+same_lang+shared_platform (59%)

ParaVT alternative? Top 6 similar tools

Looking for a ParaVT alternative? If you're comparing ParaVT with other agent tool tools, these 6 projects are the closest alternatives on Agent Skills Hub — ranked by topic overlap, star count, and community traction.

  • ToolBrain by ToolBrain · ⭐ 197

    A framework for agentic tool use training with reinforcement learning

  • NEWTON by CUTEPKQ · ⭐ 143

    NEWTON: Agentic Planning for Physically Grounded Video Generation

  • llm-rl-environments-lil-course by anakin87 · ⭐ 218

    🌱 A little course on Reinforcement Learning Environments for evaluating and training Language Models

  • agentic-grpo-longhorizon by qiqihezh · ⭐ 171

    Fixing GRPO training collapse in long-horizon multi-tool agents. A lightweight PRM-Lite + LATA joint approach

  • Travel-Agent-based-on-Qwen2-RLHF by NJUxlj · ⭐ 81

    A travel agent based on Qwen2.5, fine-tuned by SFT + DPO/PPO/GRPO using traveling question-answer dataset, a m

  • OmniAgent by HarryHsing · ⭐ 70

    OmniAgent (ICML 2026): the first native omni-modal agent for active video perception — a 7B agent that beats Q

More Agent Tool Tools

Explore other popular agent tool tools:

View all Agent Tool tools →

Popular Python Agent Tools

Frequently Asked Questions

What is ParaVT?

ParaVT is ParaVT: Taming the Tool Prior Paradox for Parallel Tool Use in Agentic Video Reinforcement Learning. It is categorized as a Agent Tool with 55 GitHub stars.

What programming language is ParaVT written in?

ParaVT is primarily written in Python. It covers topics such as agentic-rl, grpo, long-video-understanding.

How do I install or use ParaVT?

You can find installation instructions and usage details in the ParaVT GitHub repository at github.com/EvolvingLMMs-Lab/ParaVT. The project has 55 stars and 2 forks, indicating an active community.

What license does ParaVT use?

ParaVT is released under the Apache-2.0 license, making it free to use and modify according to the license terms.

What are the best alternatives to ParaVT?

The top alternatives to ParaVT on Agent Skills Hub include ToolBrain, NEWTON, llm-rl-environments-lil-course. Each offers a different approach to the same problem space — compare them side-by-side by stars, quality score, and community activity.

How this security grade is produced

Grades come from a rule-based scan built on the SlowMist agent-security taxonomy, covering 11 red-flag categories including credential harvesting, data exfiltration, and curl | sh installers. It is a first-layer scan, not a manual audit — we say so rather than overstate it.

The scale of the problem is documented independently: Liu et al. (2026), in a study of 31,132 agent skills, report that 26.1% contain security vulnerabilities. Our own full-catalog census is published as a citable open dataset.

Sources & who's responsible:

View on GitHub → Browse Agent Tool tools