Actor, Judge, or Teacher? Understanding Decision Models in Long-Horizon Agents
Paper · https://arxiv.org/abs/2611.17167
Decision models answer a question with a typed output, such as a choice among options with a probability for each, so one model can choose an agent's actions, evaluate actions already taken, or score every action to teach a policy with reinforcement learning. Prior work has studied them mainly as actors, judging their usefulness by how well they act. We ask whether a decision model that acts poorly can still teach well. We compare three decision models and three general LLMs as actors and judges on ALFWorld, and use JEV as a teacher on ALFWorld and WebShop. As actors, all six models are pulled away from a subgoal that must come first by an easier one. As judges that see the whole trajectory, the same models credit the steps toward that subgoal, and asking whether each action helped, rather than whether the task will succeed, separates useful from useless actions. As a teacher, JEV raises a Qwen2.5-1.5B policy over GRPO from 72.8% to 82.9% success on ALFWorld and from 56.8% to 71.4% on WebShop. These results suggest that a poor actor can still teach well, because judging an action after the episode ends reveals the consequences that acting must anticipate. Code is available at https://github.com/ZJU-REAL/decision-agents.
First seen by Oriel on .
AI summary
This paper shows that decision models performing poorly as long-horizon actors can still teach policies effectively when judging completed trajectories.
- The study compares three decision models and three general LLMs as actors and judges on ALFWorld, using JEV as a teacher on ALFWorld and WebShop.
- As actors, all six models are distracted from a necessary subgoal by an easier alternative.
- When judging complete trajectories, the models credit actions toward the subgoal, especially when asked whether each action helped rather than whether the task would succeed.
- JEV improves a Qwen2.5-1.5B policy trained with GRPO from 72.8% to 82.9% success on ALFWorld and from 56.8% to 71.4% on WebShop.
#long-horizon agents #reinforcement learning #llm evaluation #alfworld
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Points
Score reported by the source.
- Hugging Face Papers: 37, latest observation ; +8 over observed period since
Comments
Comment count reported by the source.
- Hugging Face Papers: 6, latest observation ; +4 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Papers | 2611.17167 | https://arxiv.org/abs/2611.17167 Read discussion |