Research map / Multi-agent systems and games / LLM agents
ProAct: Agentic Lookahead in Interactive Environments
Yangbin Yu, Mingyu Yang, junyou li, Yiming Gao, Feiyu Liu, Yijun Yang, Zichuan Lin, Jiafei Lyu, Zhicong Lu, Deheng Ye, Jie Jiang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add MC-Critic RL after GLAD SFT (ablation) | GLAD-trained Qwen3-4B-Instruct · 2048 4x4 | score: 3335.3 -> 4503.8 | Improved |
| add MC-Critic RL after GLAD SFT (ablation) | GLAD-trained Qwen3-4B-Instruct · Sokoban Base | avg boxes placed: 0.72 -> 0.94 | Improved |
| add MC-Critic RL after GLAD SFT (ablation) | GLAD-trained Qwen3-4B-Instruct · 2048 3072 | score: 4565.7 -> 6013.7 | Improved |
| add GLAD distillation SFT (ablation) | Qwen3-4B-Instruct · 2048 4x4 | score: 721.3 -> 3335.3 | Improved |
| add GLAD distillation SFT (ablation) | Qwen3-4B-Instruct · Sokoban Base | avg boxes placed: 0.39 -> 0.72 | Improved |
| add GLAD distillation SFT (ablation) | Qwen3-4B-Instruct · Sokoban Symbol | avg boxes placed: 0.56 -> 0.67 | Improved |
| add MC-Critic to Step-PPO, from scratch (ablation) | Step-PPO · 2048 3072 (from scratch) | score: 1912.90 -> 2229.1 | Improved |
| add MC-Critic to Step-GRPO, from scratch (ablation) | Step-GRPO · Sokoban Action (from scratch) | avg boxes placed: 0.53 -> 0.80 | Improved |
| add MC-Critic to Step-GRPO, from scratch (ablation) | Step-GRPO · 2048 3x3 (from scratch) | score: 188.0 -> 194.2 | Mixed |
| add MC-Critic to Step-PPO on GLAD init (ablation) | Step-PPO · 2048 3072 (GLAD-init RL) | score: 6248.7 -> 5818.1 | Hurt |
| add MC-Critic to Step-PPO on GLAD init (ablation) | Step-PPO · Sokoban Extra (GLAD-init RL) | avg boxes placed: 0.90 -> 1.18 | Improved |
| use M=10 rollouts instead of M=0 (Step-GRPO) in MC-Critic (ablation) | MC-GRPO · 2048 | score | Hurt |
Closest papers by meaning
- ProAct: A Benchmark and Multimodal Framework for Structure-Aware Proactive Response (ICML 2026 · similarity 0.87)
- Proactive Agent: Shifting LLM Agents from Reactive Responses to Active Assistance (ICLR 2025 · similarity 0.87)
- Reinforcement Learning with Lookahead Information (NeurIPS 2024 · similarity 0.86)
- Reason for Future, Act for Now: A Principled Architecture for Autonomous LLM Agents (ICML 2024 · similarity 0.85)
- Selective Critique for Cost-Aware LLM Agents in Long-Horizon Decision Making (NeurIPS 2026 · similarity 0.85)
- Pushing Forward Pareto Frontiers of Proactive Agents with Behavioral Agentic Optimization (ICML 2026 · similarity 0.85)
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.
