- Rating
- 2
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper proposes MemGUI-Agent, which uses the CONACT protocol to have the same model generate GUI actions and context-management decisions while maintaining history summaries, UI factual memory, and a record of the most recent step. The authors construct MemGUI-3K, a dataset containing 2,956 trajectories, train an 8B model through supervised fine-tuning, and evaluate the method on MemGUI-Bench and MobileWorld.
Critical strengths
- Information loss during long-horizon, cross-app operations is a practical problem for mobile GUI agents. With the same 235B model and unchanged weights, CONACT improves performance on both MemGUI-Bench and MobileWorld, providing evidence for the practical value of context management.
- MemGUI-3K provides joint supervision for GUI actions, history compression, and factual memory. The 8B model's training results indicate that these behaviors can be learned from demonstrations, giving the dataset research value.
Critical weaknesses
- Limited methodological novelty. AgentFold [1], OpenGCUA [4] already implements learnable history folding, MemAct [2] incorporates memory editing into policy actions, and Mobile-Agent-v2 [3] uses cross-app factual memory. The paper mainly integrates these mechanisms and lacks a direct comparison with a GUI-adapted AgentFold augmented with factual memory. A substantive technical advance has therefore not been clearly established.
- Insufficient justification for the memory partition. The paper does not adequately explain why the state should be divided into history summaries, UI factual memory, and the most recent step record. The component ablations in Table 5 do not establish the necessity of this partition. Without matched comparisons against a unified memory store or a two-part state, the independent benefit of the three-part structure remains unclear.
[1] Rui Ye et al. AgentFold: Long-Horizon Web Agents with Proactive Context Management. arXiv preprint arXiv:2510.24699, 2025.
[2] Yuxiang Zhang, Jiangming Shu, Ye Ma, Xueyuan Lin, Shangxi Wu, and Jitao Sang. Memory as Action: Autonomous Context Curation for Long-Horizon Agentic Tasks. arXiv preprint arXiv:2510.12635, 2025.
[3] Junyang Wang et al. Mobile-Agent-v2: Mobile Device Operation Assistant with Effective Navigation via Multi-Agent Collaboration. arXiv preprint arXiv:2406.01014, 2024.
[4] OpenGCUA: Toward a Fully Open Stack for Generalist Computer Use
Questions for authors
- Please clarify CONACT's specific technical advance over AgentFold with a separate factual memory. Are there comparisons using the same backbone and inference budget that support the independent contribution of this advance?
- What justifies the three-part state partition? Are there experiments controlling the model, training conditions, and context budget that compare a unified memory store, a two-part state, and the current three-part state, separating the effect of state structure from the amount of recorded information?
5 comments
Response to Reviewer R9sP (2/2)
(3) Evidence for the GUI context-update protocol (Enhanced: ). All eight component controls share unchanged Qwen3-VL-235B-Thinking weights, 40 tasks, temperature 0, matched initial states and task-step limits, and up to three attempts. denote folding/UI memory/self-description; gives the context setups. Token use and latency vary by configuration.
P@1, P@3, and IRR are percentages. gives full CONACT minus each row’s P@1/P@3, in percentage points.
| Variant | P@1 | P@3 | MTPR | IRR | ||||
|---|---|---|---|---|---|---|---|---|
| ReAct | 0 | 0 | 0 | 5.0 | 27.5 | 0.143 | 19.5 | 35.0/35.0 |
| only‡ | 0 | 1 | 0 | 17.5 | 42.5 | 0.357 | 39.1 | 22.5/20.0 |
| only† | 1 | 0 | 0 | 22.5 | 32.5 | 0.179 | 25.7 | 17.5/30.0 |
| only | 0 | 0 | 1 | 17.5 | 37.5 | 0.110 | 32.2 | 22.5/25.0 |
| 1 | 1 | 0 | 27.5 | 52.5 | 0.380 | 37.8 | 12.5/10.0 | |
| 0 | 1 | 1 | 25.0 | 45.0 | 0.214 | 33.1 | 15.0/17.5 | |
| 1 | 0 | 1 | 27.5 | 50.0 | 0.250 | 34.8 | 12.5/12.5 | |
| Full CONACT | 1 | 1 | 1 | 40.0 | 62.5 | 0.429 | 51.0 | 0.0/0.0 |
† The folding-only component shares AgentFold's history-folding idea; AgentFold additionally retains Latest Interaction. ‡ The memory-only component edits UI facts; MemAct applies memory-as-action to working history.
The ablations support joint use of folding, persistent UI facts, and recent screen evidence. Each operation improves success with the other two enabled, and the full protocol leads all four metrics. Self-description supplies the screen facts and action intent used by subsequent folding and memory decisions.
Manuscript changes. We clarified the technical contribution in (lines 087–090) and strengthened the training motivation in (p. 5; lines 240–244). now compares mobile GUI support, policy control, and joint supervision, and explains the component controls’ relation to prior methods (pp. 14–15; lines 722–761). (p. 8) now shows each ablation’s gap to full CONACT.
W2 & Q2: Three-part state design
Response. The partition reflects different update rules: history () is compressed, UI facts () persist until explicitly edited, and the latest interaction () is refreshed each step. This lets the policy compress past actions while retaining stored facts and recent screen evidence.
Over the past few days, we have expanded to report the complete component ablation. All variants use the same zero-shot Qwen3-VL-235B-Thinking checkpoint, 40 tasks, and task-dependent step limits.
The distinct update roles are useful together in practice: full CONACT leads all four metrics. Removing folding, UI memory, or self-description lowers P@1 from 40.0% to 25.0%, 27.5%, and 27.5%, respectively.
Manuscript changes. We clarified the different update rules in (p. 3; lines 156–158). The expanded (p. 8; lines 421–431) reports all eight configurations, with their context setups detailed in (p. 38; lines 2006–2019).
References
[1] AgentFold: Long-Horizon Web Agents with Proactive Context Folding. ICLR 2026.
[2] Memory as Action: Autonomous Context Curation for Long-Horizon Agentic Tasks. Findings of ACL 2026.
[3] Mobile-Agent-v2: Mobile Device Operation Assistant with Effective Navigation via Multi-Agent Collaboration. NeurIPS 2024.
[4] OpenGCUA: Toward a Fully Open Stack for Generalist Computer Use. arXiv preprint, 2026.
Dear Reviewer R9sP,
Thank you again for your suggestions. In our rebuttal, we clarified the differences from prior work and the rationale for the three-part state, and added the complete component ablation.
We would welcome your thoughts on these clarifications and results before the discussion closes.
Best regards,
Authors of MemGUI-Agent
Response to Reviewer R9sP (1/2)
We thank Reviewer R9sP for recognizing the "practical value of context management" shown by our fixed-weight results and the "research value" of MemGUI-3K's joint supervision. We address each concern below.
W1 & Q1: Technical contribution and matched component evidence
Response. Our contribution comprises CONACT's per-step GUI context interface and MemGUI-3K's joint supervision of GUI actions and context management.
(1) Comparison with related work (New: Appendix A, Table 7).
The policy column concerns control of both interaction history and task actions. The data column requires joint supervision of mobile GUI actions and context updates.
CONACT records task-relevant screen facts in
<ui_observation>before a transition and passes them, with the action intent and action/result, to subsequent folding and memory decisions (Section 2).OpenGCUA's arXiv v1 was posted on 15 October 2026, after the 25 September ICLR 2027 submission deadline; we include it in the comparison above.
(2) MemGUI-3K provides supervision for context management. Section 3.1 (Table 1) (original submission, p. 5) reports
12.5%Pass@1 with ReAct and7.5%with zero-shot CONACT for the 8B model on MemGUI-Bench-40. The tested 2B/4B models also decline, whereas 235B-Thinking improves. These results motivate training context-management decisions.MemGUI-3K provides
2,956GUI trajectories jointly supervising actions, folding, memory editing, and self-description. For the 8B model, SFT raises memory-trigger F1 from19.9%to48.0%and format compliance from94.9%to99.9%in the gold-context offline evaluation (Table 4, original submission, p. 8). These gains show improved memory-use decisions alongside reliable structured generation.