Research map / Multi-agent systems and games / LLM agents
HAMLET: Hyperadaptive Agent-based Modeling for Live Embodied Theatrics
Shufan Jiang, Sizhou Chen, Chi Zhang, Xiao-Lei Zhang, Xuelong Li
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes HAMLET, an LLM-based agent framework for live embodied theatrical drama in which agents act with initiative and interact with the physical scene without requiring detailed user input.
Problem
Existing LLM-based drama generation methods yield agents that lack initiative, cannot interact with the physical scene, and require detailed user input to drive the story.
Approach
An agent-based modeling framework for live embodied theatrics that gives characters autonomous initiative and physical-scene interaction, reducing the need for detailed user specification. Only the abstract was available, so specifics are not stated.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the advancer agent (ablation) | HAMLET · online drama performance | task completion rate: 68.1% -> 100% | Improved |
| add the PAD module to the actors (ablation) | HAMLET · 30 topics, 10 categories | drama quality (radar chart) | Improved |
| use multi-agent HAMLET workflow instead of raw-prompted single GPT-4o (ablation) | GPT-4o · 30 topics | drama quality (radar chart) | Improved |
| add the reviewer to offline planning (ablation) | HAMLET · blueprint quality and online Narrative Quality | win-rate / Narrative Quality | Improved |
| enable reasoning (thinking) in Qwen3-32B | Qwen3-32B · HAMLET Leaderboard | overall score: 65.57 -> 73.85 | Improved |
| enable thinking in Qwen3-8B | Qwen3-8B · HAMLET Leaderboard | overall score: 52.48 -> 58.16 | Improved |
| scale Qwen3 from 32B to 235B (non-thinking) | Qwen3 · HAMLET Leaderboard | overall score: 65.57 -> 71.21 | Improved |
| use reasoning models for function-calling response strategy | strategy-selection models · PAD strategy evaluation | latency penalty: penalty 0.15 for DeepSeek-R1-0528 | Hurt |
| use trained HAMLETJudge instead of prompted frontier judges | drama evaluation · human agreement | Pearson correlation: 0.762 -> 0.791 vs Claude-4-Sonnet | Improved |
Closest papers by meaning
- VirtualEnv: A Platform for Embodied AI Research (AAAI 2026 · similarity 0.85)
- Steve-Eye: Equipping LLM-based Embodied Agents with Visual Perception in Open Worlds (ICLR 2024 · similarity 0.84)
- StoryBox: Collaborative Multi-Agent Simulation for Hybrid Bottom-Up Long-Form Story Generation Using Large Language Models (AAAI 2026 · similarity 0.84)
- MMRole: A Comprehensive Framework for Developing and Evaluating Multimodal Role-Playing Agents (ICLR 2025 · similarity 0.84)
- RoleAgent: Building, Interacting, and Benchmarking High-quality Role-Playing Agents from Scripts (NeurIPS 2024 · similarity 0.83)
- NarrativePlay: An Automated System for Crafting Visual Worlds in Novels for Role-Playing (AAAI 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
