AAAI'27 将办第三届 LLM 社会模拟研讨会,11 月 20 日截稿
RSTZZZ · reddit · 2026-11-16
The 3rd Workshop on Social Simulation with LLMs (Social Sim'27) will be held at AAAI-27 in Montréal on Feb 22–23, 2027, with submissions due Nov 20 (AoE) and notifications Dec 2.
The central question is attribution: when a simulated LLM population shows polarization, norm formation, or information cascades, how do you know the population produced it — not the prompt, the base model, or memorized pretraining data? Three axes:
- Population construction: data-grounded populations, calibration, persona modeling, heterogeneity, human-AI hybrids
- Collective dynamics: separating emergence from recall, robustness across models/seeds, scaling, norms, network science and game theory methods
- Inference and consequence: validation protocols, shared testbeds, replications and negative results, simulation as pre-deployment evaluation
Papers challenging the premise are welcome. Scope requires simulating a social environment; single-agent work, agentic coding benchmarks, and multi-agent RL without social interpretation are out. Three non-archival tracks: Regular (7pp), Early-stage (4pp), Replication & resources (4pp + artifacts), double-blind per the AAAI-27 kit.
「研究」频道最新
- 新综述梳理 LLM 端侧部署:软件优化+专用硬件可行路径 — kalyan_kpl · 2026-11-16
- Lagomics 称肽药人体数据周期从 10 年缩到 4-6 个月 — niloofar_mire · 2026-11-16
- 免费开放教材《Bayes, AI and Deep Learning》:从贝叶斯讲到 Transformer 与 Agent — burkov · 2026-11-16
- Ropedia Xperience 2 发布:1589 个 3D 扫描与 4230 段第一视角录像支撑 Physical AI 数据扩展 — liuziwei7 · 2026-11-16
- CVPR 2027 AC 指南:AI 可辅助评审,元评审禁止用 AI 生成 — CSProfKGD · 2026-11-16
- 顶级研究员竞相探索非 Transformer 架构,重大突破可期 — khademinori · 2026-11-16