Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Counterfactual Planning for Generalizable Agents’ Actions
Jiarun Fu, Lizhong Ding, Qiuning Wei, Yuhan Guo, Yurong Cheng, Junyu Zhang
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 State Causality Evaluator (SCE) (ablation) | Counterfactual Planning (without SCE) · Crafter (22 tasks, 1M steps) | Score (%): 16.44 -> 20.4 | Improved |
| add What-If-Not (WIN) reward (ablation) | Counterfactual Planning (without WIN) · Crafter (22 tasks, 1M steps) | Score (%): 17.82 -> 20.4 | Improved |
| add both SCE and WIN (ablation) | Qwen2.5-72B-Instruct planner · Crafter (22 tasks, 1M steps) | Score (%): 15.6 -> 20.4 | Improved |
| add SCE on top of WIN (ablation) | Counterfactual Planning · Crafter (22 tasks, 1M steps) | Score (%): 15.6 -> 17.82 | Improved |
| use dynamic causal inference (CP) instead of static Causal LLMs | Causal LLMs · Crafter (22 tasks, 1M steps) | Score (%): 18.9 -> 20.4 | Improved |
| use dynamic causal inference (CP) instead of static Causal LLMs | Causal LLMs · Crafter (22 tasks, 5M steps) | Score (%): 33.6 -> 36.1 | Improved |
| use CP instead of AdaRefiner | AdaRefiner · Crafter (22 tasks, 1M steps) | Score (%): 15.8 -> 20.4 | Improved |
| use CP instead of Achievement Distillation | Achievement Distillation · Crafter (22 tasks, 1M steps) | Score (%): 21.8 -> 20.4 | No clear effect |
| use CP instead of DreamerV3 | DreamerV3 · Crafter (22 tasks, 1M steps) | Score (%): 14.77 -> 20.4 | Improved |
| train from 1M to 5M steps | Counterfactual Planning · Crafter (22 tasks) | Score (%): 20.4 -> 36.1 | Improved |
Closest papers by meaning
- Counterfactual Effect Decomposition in Multi-Agent Sequential Decision Making (ICML 2025 · similarity 0.89)
- Should I Have Expressed a Different Intent? Counterfactual Generation for LLM-Based Autonomous Control (ICML 2026 · similarity 0.88)
- Language Agents Meet Causality -- Bridging LLMs and Causal World Models (ICLR 2025 · similarity 0.88)
- Towards Generalizable Reinforcement Learning via Causality-Guided Self-Adaptive Representations (ICLR 2025 · similarity 0.87)
- Interpreting Emergent Planning in Model-Free Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Modeling Unseen Environments with Language-guided Composable Causal Components in Reinforcement Learning (ICLR 2025 · similarity 0.87)
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.
