Skip to content

Releases: sii-research/SocioVerse2

Release list

SocioVerse2 v0.2.0

SocioVerse2 v0.2.0 Latest
Latest

Choose a tag to compare

View all tags
@Lishi905 Lishi905 released this 24 Sep 22:26
v0.2.0
87aff95

SocioVerse2 v0.2.0 is the first public release of the SocioVerse2 longitudinal social-simulation framework.

What is included

  • Runtime (socioverse): a fixed population with persistent agent ids acts step by step in an environment that keeps changing, and every step records one panel row per agent in DuckDB. Interventions (scheduled events and audience-scoped broadcasts) fire at chosen steps; a counterfactual branch replays its parent's earlier steps exactly and pays only for the steps after the fork.
  • Workflow skills (/sv-init, /sv-build-model, /sv-build-environment, /sv-build-population, /sv-run, /sv-report, /sv-iterate, plus /sv-lit and /sv-paper) for Claude Code, with a Codex mirror (experimental).
  • Local research dashboard: overview, population, environment, dataset, experiments, literature and paper tabs, with version and branch history.
  • Reference studies: from-scratch templates (opinion diffusion, campus dining choice), real-data cases (German car market, US consumer-confidence backtest, hybrid opinion dynamics), and wrappers for the 11 classic ABMs and the Chicago segregation model from the companion repository SocioVerse-ABM.

Get started

git clone https://github.com/sii-research/SocioVerse2.git
cd SocioVerse2
python -m pip install -e ".[dev,viz]"
python -m pytest -q
python scripts/demo_offline.py     # a full study end to end, no API key needed

The core library alone is also on PyPI: pip install socioverse2 (import name socioverse).

See the README for real-LLM runs (any OpenAI-compatible endpoint), the dashboard, and the Claude Code workflow.

Links

SocioVerse2 v0.2.0 是 SocioVerse2 纵向社会模拟框架的首个公开版本。

  • 运行时（socioverse）：固定人群（持久 id）在持续变化的环境中逐步行动，每一步为每个智能体记录一行面板数据（DuckDB）。干预（定时事件与定向广播）在指定步触发；反事实分支精确重放父版本分叉前的步骤，只为分叉后的步骤付费。
  • 工作流 skills：/sv-init 到 /sv-report、/sv-iterate，以及 /sv-lit、/sv-paper，支持 Claude Code；Codex 镜像为实验性。
  • 本地研究仪表盘：总览、人群、环境、数据、实验、文献、论文等页，带版本与分支历史。
  • 参考研究：从零构建的模板、真实数据案例，以及对配套仓库 SocioVerse-ABM 中 11 个经典 ABM 与芝加哥隔离模型的封装。

核心库也已发布到 PyPI：pip install socioverse2（导入名 socioverse）。快速开始见上方命令与 README。技术报告：https://arxiv.org/abs/2609.24911 · 主页：https://socioverse.fudan-disc.com/ · 使用手册：https://socioverse.fudan-disc.com/docs/ · 挑战赛：https://socioverse.fudan-disc.com/challenge/

Assets 4
Loading