- Rating
- 3
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
MetaDecomp learns a task-relevant state schema and organizes successful training traces into procedural memory. As the schema evolves, historical observations are reprojected and procedures are regrouped into a dependency graph. At inference time, the agent uses the learned state abstraction and procedures to revise task decompositions, while the schema and memory remain frozen. The distinguishing idea is to learn how experience should be represented and organized together.
Critical strengths
- Jointly evolving the state schema and procedural organization is a plausible advance over workflow retrieval methods such as Agent Workflow Memory. Reprojecting old experience when the abstraction changes addresses a real problem: a memory can become poorly indexed as the agent learns which state distinctions matter.
- The training/test boundary is clearly described. The stronger outer-loop teacher is not used during test execution, and the learned memory is frozen. This makes the reported inference gains easier to interpret.
Critical weaknesses
- The main ablation compares the full system with an inner loop that has empty procedural memory. This shows that learned memory helps, but does not establish the contribution of adaptive schemas, historical reprojection, or DAG dependencies individually. These are the parts that carry the novelty claim.
- The cost discussion primarily concerns execution. Outer-loop learning and initial plan refinement are excluded from the execution budget, while ALFWorld can use more execution tokens than MPO. It is difficult to judge the total cost advantage without an amortized accounting.
Questions for authors
Please refer to critical weakness.
5 comments
Response to Reviewer ZaCt (2/2)
(W2) The cost discussion primarily concerns execution. Outer-loop learning and initial plan refinement are excluded from the execution budget, while ALFWorld can use more execution tokens than MPO. It is difficult to judge the total cost advantage without an amortized accounting.
Thank you for emphasizing complete cost accounting; this helped us make the training and inference cost comparison fully transparent. Regarding your concern that the
cost discussion primarily concerns executionand thatOuter-loop learning and initial plan refinement are excluded from the execution budget, we would like to respectfully clarify that theexecution budgetin original Appendix A.1 (Lines 772–773) referred only to the 1M-token per-episode termination cap during rollout, whereasinitial plan refinementwas already included in all reported inference token totals, as noted in Line 426 of the original submitted manuscript. Below, Table R3.2 and Table R3.3 report per-task inference costs and one-time training costs separately to support anamortized accountingof thetotal cost advantage.
(Inference cost including initial plan refinement) With
initial plan refinementfully included, MetaDecomp achieves higher performance with fewer inference tokens than ReAct and AutoRefine in all five settings, and uses fewer inference tokens than MPO on ScienceWorld.
- While
ALFWorld can use more execution tokens than MPOin exchange for higher task success, on the longer-horizon ScienceWorld benchmark MetaDecomp simultaneously achieves higher task scores and lower per-task inference tokens, includinginitial plan refinement, than MPO [1], ReAct [2], AutoRefine [4], and SkillOpt [5], as shown in Table R3.2 and manuscript Table 1 and Table 2.Table R3.2. Mean inference cost per task in K input-plus-output tokens, including initial planning, initial plan refinement, online control, and actor rollouts across successful and failed episodes. The 4B/8B actors use Qwen3; 9B/27B use Qwen3.5.
Method ALFWorld 4B ALFWorld 8B ScienceWorld 8B ScienceWorld 9B ScienceWorld 27B MetaDecomp 53.80 42.60 94.90 103.30 74.30 ReAct [2] 57.22 49.40 140.58 149.82 89.67 AdaPlanner [3] 32.30 28.40 40.07 42.88 30.81 MPO 48.34 40.27 130.28 130.31 76.05 AutoRefine [4] 64.75 51.04 142.41 146.70 92.82 SkillOpt [5] 101.50 86.71 213.01 218.77 86.04 (Outer-loop learning cost and amortized total cost advantage)
Outer-loop learningis a one-time training cost that is smaller than SkillOpt across all five settings and is amortized across tasks after deployment.
- Across all five settings in Table R3.3, MetaDecomp uses substantially fewer one-time training tokens for outer-loop learning and rollouts than SkillOpt and more than AutoRefine, while ReAct, AdaPlanner, and MPO have no training run in our experiments.
- Under an
amortized accountingof thetotal cost advantage, because the learned procedural memory is frozen at deployment, no further outer-loop learning cost is incurred at test time, so MetaDecomp's per-task inference savings over ReAct, AutoRefine, SkillOpt, and on ScienceWorld over MPO accumulate directly as test volume grows.Table R3.3. Full training cost in M input-plus-output tokens, including training rollouts, initial planning/refinement, and outer-loop learning updates; MPO uses an official checkpoint with unreported upstream training cost. The 4B/8B actors use Qwen3; 9B/27B use Qwen3.5.
Method ALFWorld 4B ALFWorld 8B ScienceWorld 8B ScienceWorld 9B ScienceWorld 27B MetaDecomp 23.02 20.34 65.90 37.08 35.42 AutoRefine 5.29 4.08 11.35 14.81 6.89 SkillOpt 43.62 31.44 95.00 100.51 52.97 [1] MPO: Boosting LLM Agents with Meta Plan Optimization. Findings of EMNLP 2025.
[2] ReAct: Synergizing Reasoning and Acting in Language Models. ICLR 2023.
[3] AdaPlanner: Adaptive Planning from Feedback with Language Models. NeurIPS 2023.
[4] AutoRefine: Compiling Trajectories into Validated Typed Agent Artifacts. arXiv preprint 2026.
[5] SkillOpt: Executive Strategy for Self-Evolving Agent Skills. arXiv preprint 2026.
Thank you again for helping us strengthen the evaluation and its presentation. We would be happy to clarify any remaining questions.
We would like to hear from Reviewer ZaCt
Hi Reviewer ZaCt,
We sincerely appreciate your dedicated effort in reviewing our manuscript. We would like to follow up to see if our response addresses your concerns or if you have any further questions. We would really appreciate the opportunity to discuss this further if our response has not already addressed your concerns.
Thank you again!
Submission #3603 Authors
Response to Reviewer ZaCt (1/2)
We sincerely appreciate your thoughtful review and your recognition of the value of jointly evolving state representations and procedural knowledge. Your suggestions have helped us strengthen the evidence for individual learning mechanisms and provide a more complete comparison of training and inference costs. We address both concerns point by point below. Manuscript section, appendix, figure, and table numbers refer to the revised paper; changes are highlighted in blue. Please let us know if you have any remaining concerns or would like further clarification; we would be happy to discuss them.
(W1) The main ablation compares the full system with an inner loop that has empty procedural memory. This shows that learned memory helps, but does not establish the contribution of adaptive schemas, historical reprojection, or DAG dependencies individually. These are the parts that carry the novelty claim.