On Sample-Efficient Offline Reinforcement Learning: Data Diversity, Posterior Sampling, and Beyond

We seek to understand what facilitates sample-efficient learning from historical datasets for sequential decision-making, a problem that is popularly known as offline reinforcement learning (RL). Further, we are interested in algorithms that enjoy sample efficiency while leveraging (value) function approximation. In this paper, we address these fundamental questions by (i) proposing a notion of data diversity that subsumes the previous notions of coverage measures in offline RL and (ii) using this notion to {unify} three distinct classes of offline RL algorithms based on version spaces (VS), regularized optimization (RO), and posterior sampling (PS). We establish that VS-based, RO-based, and PS-based algorithms, under standard assumptions, achieve \emph{comparable} sample efficiency, which recovers the state-of-the-art sub-optimality bounds for finite and linear model classes with the standard assumptions. This result is surprising, given that the prior work suggested an unfavorable sample complexity of the RO-based algorithm compared to the VS-based algorithm, whereas posterior sampling is rarely considered in offline RL due to its explorative nature. Notably, our proposed model-free PS-based algorithm for offline RL is {novel}, with sub-optimality bounds that are {frequentist} (i.e., worst-case) in nature.

Paper

References (55)

Scroll for more · 38 remaining

Similar papers

Peer review

Reviewer FhSe7/10 · confidence 2/52023-07-04

Summary

This paper proposes a new notion of data diversity that expands all the prior data diversity conditions in offline reinforcement learning literature. Based on this notion, the paper develops concrete value sampling-based, regularized optimization-based and posterior sampling-based algorithms with corresponding sample complexity results. The paper shows that these three algorithms can achieve comparable sample efficiency.

Strengths

1. The proposed data diversity condition and its corresponding suboptimality guarantee seem to be general and tighter than previous diversity conditions characterizing the state-of-the-art bounds for the main streams of offline reinforcement learning frameworks. This data diversity condition is a potential inspiration for future work. 2. The writing is clear and presents a fair and clean picture compared to previous results.

Weaknesses

There are some typos and inconsistencies in the notation, such as $p_{0,h}$ in $p_{0,h}(\mathcal{F}_h^{\tilde{\pi}}(\epsilon; f\_{h+1}))$ in line 176 and $f_h$ and $g_h$ in Algorithm 2, which should be $f$ and $g$, respectively. Similar inconsistencies appear in Algorithm 3.

Questions

Could you please explain how the new diversity notion is able to cover value sampling-based, regularized optimization-based, and posterior sampling-based algorithms, especially posterior sampling-based algorithms? Is it because the previous diversity conditions are not always able to cover all three types of algorithms? If the previous conditions are able to cover all three types, is the main advantage of the new notion its tightness? If the previous conditions cannot cover any of the algorithm types, could you describe the reason behind their failure and why the new notion is able to succeed?

Rating

7: Accept: Technically solid paper, with high impact on at least one sub-area, or moderate-to-high impact on more than one areas, with good-to-excellent evaluation, resources, reproducibility, and no unaddressed ethical considerations.

Confidence

2: You are willing to defend your assessment, but it is quite likely that you did not understand the central parts of the submission or that you are unfamiliar with some pieces of related work. Math/other details were not carefully checked.

Soundness

4 excellent

Presentation

4 excellent

Contribution

4 excellent

Limitations

See weakness.

Reviewer ufo55/10 · confidence 3/52023-07-05

Summary

This paper investigates the problem of sample-efficient learning from historical datasets in the context of offline reinforcement learning and explores the role of data diversity and posterior sampling in improving the efficiency of RL algorithms. The authors propose a new notion of data diversity and study three classes of offline RL algorithms based on version spaces, regularized optimization, and posterior sampling. They find that these algorithms achieve comparable sample efficiency, contrary to prior work. The paper also introduces a novel model-free posterior sampling algorithm for offline RL.

Strengths

This work studies offline RL problems within the context of general function approximation. The introduction of the novel notion of data diversity is intriguing, as it offers a fresh perspective on the problem. Additionally, the inclusion of the posterior sampling algorithm is interesting, as it appears to be a new contribution to the literature on offline RL theory.

Weaknesses

It is important to ensure the accuracy of the comparison with LCB-based algorithms, as the algorithm mentioned in [JYW21] may not be considered the SOTA algorithm when compared to improved LCB-based algorithms proposed in works such as [1] and [2], which have demonstrated better dependencies on d. Additionally, it is worth noting that there has been an update in [XCJ+21], where a new arXiv version employs $\Pi^{soft}$ instead of $\Pi^{all}$. While this minor issue does not impact my evaluation, it would be beneficial for the authors to acknowledge this update and adjust the comparisons accordingly. [1] Near-optimal offline reinforcement learning with linear representation: Leveraging variance information with pessimism [2] Nearly Minimax Optimal Offline Reinforcement Learning with Linear Function Approximation: Single-Agent MDP and Markov Game

Questions

n/a

Rating

5: Borderline accept: Technically solid paper where reasons to accept outweigh reasons to reject, e.g., limited evaluation. Please use sparingly.

Confidence

3: You are fairly confident in your assessment. It is possible that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work. Math/other details were not carefully checked.

Soundness

3 good

Presentation

3 good

Contribution

3 good

Limitations

n/a

Reviewer YpYn6/10 · confidence 3/52023-07-06

Summary

This work aims to point out what enables the sample efficiency of offline reinforcement learning. The authors first define a new notion of data diversity based on the inducing the Bellman error under one policy with the Bellman error under a different policy. The authors then propose a unified view, Generic Offline Policy Optimisation framework, to study the three offline algorithms: version spaces, regularised optimisation, and posterior sampling. Under standard assumption, the authors showed that all these three algorithms can achieve comparable upper bounds of the policies' sub-optimality.

Strengths

1. The proposed view is novel, and can unify the three different types of offline reinforcement learning algorithms. 2. The whole work is very well presented, and the structure is super clear, which raises a high comprehensibility. 3. All definitions and assumptions are explicitly and clearly given in Section 2. I personally love this practice, as it helps a lot on figuring the scope of this work. 4. The results are also well-organised and clearly listed in Section 4. 5. The connection between this work and [TJJ20] is clear, thus it's interesting to see that offline RL can be connected to transfer learning, though there are certain technical differences.

Weaknesses

### Major 1. **Practical issues of posterior sampling algorithms**: My biggest concern about this work is the performance of the posterior sampling based algorithms in practice. I appreciate that the authors remark some feasible solutions to implement an offline RL algorithm based on posterior sampling from line 241 to line 245. However, given the tractable approximations from line 243 to line 245, I kind of worry the *performance* as well as the *time complexity* of the PS implementations. I'd be happy to raise my score if the authors can discuss further details about this concern in the updated version. 2. **Discussion about the limitations**: The authors mainly focus on the missing variance information in their discussion about the limitation of the work. Given the assumptions listed in the Section 2.3 as well as the practical issues I mentioned above, the limitations can be expanded to cover all of them. ### Minor 1. $H$ in line 107 and line 110: The authors can move the notation $H$ from line 110 where it's defined to line 107 where it first appears. 2. $t$ in line 129 is not defined: Though the meaning of $t$ can be inferred given the context, it's better to be defined before its appearance in line 129. 3. Introduction section is too lengthy I appreciate the gentle and comprehensive introduction to this work in Section 1. But, it might be a bit lengthy, and as a result Section 5 is too short to further discussion about the limitation and future directions of this work.

Questions

1. Can the authors comment on the *performance* as well as the *time complexity* of possible PS implementations? (See the major issue #1 in the weakness section) 2. What are the slight refinements in line 190? The authors claimed that they have made slight refinements to the RO-based and VS-based algorithms, but they haven't mentioned these refinements anywhere else.

Rating

6: Weak Accept: Technically solid, moderate-to-high impact paper, with no major concerns with respect to evaluation, resources, reproducibility, ethical considerations.

Confidence

3: You are fairly confident in your assessment. It is possible that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work. Math/other details were not carefully checked.

Soundness

3 good

Presentation

4 excellent

Contribution

3 good

Limitations

See the major issue #2 in the weakness section.

Reviewer niYV4/10 · confidence 4/52023-07-28

Summary

This paper introduces a novel data coverage measure called "data diversity," which is more stringent than existing data coverage measures like single-policy concentrability. The authors claim that actor-critic algorithms based on VS, RO, and posterior sampling can achieve state-of-the-art sample complexity. The posterior sampling based algorithm is the first of its kind for offline RL.

Strengths

The new data coverage measure is innovative and tighter than existing measures in the literature. The authors' demonstration of good sample complexity under this measure is an important contribution. Additionally, the framework and the posterior sampling approach for offline RL are novel.

Weaknesses

However, I have some doubts about the claim that these algorithms achieve the state-of-the-art convergence rate. The policy $\hat{\pi}$ is obtained by uniformly drawing a $t$ from $[T]$ and letting $\hat{\pi}$ be $\pi^t$. This policy is not a Markov policy. According to the definition of $V^{\pi}$, $V^{\hat{\pi}}=\frac{1}{T}\sum_{i=1}^TV^{{\pi}^i}$. Is this correct? The $1/T$ term comes from the additional expectation over the uniform distribution. For the chosen $t$, $\pi^t$ can have constant variance, and as a result, the high probability bound may not hold for this chosen $\pi^t$. On the other hand, recent literature such as "Optimal conservative offline RL with general function approximation via augmented Lagrangian", "Offline Primal-Dual Reinforcement Learning for Linear MDP", and "Revisiting the linear-programming framework for offline RL with general function approximation" have policies that are Markov policies and achieve the state-of-the-art sample complexity. Therefore, the authors should compare their results with these papers. To ensure a fair comparison, the output policy should be a Markov policy. If not, the authors need to clarify this in the introduction and after the theorems. Additionally, the authors should clearly explain the source of randomness for $\hat{\pi}$ in the theorems. If the authors can address these issues clearly, I would increase the rating.

Questions

None

Rating

4: Borderline reject: Technically solid paper where reasons to reject, e.g., limited evaluation, outweigh reasons to accept, e.g., good evaluation. Please use sparingly.

Confidence

4: You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work.

Soundness

3 good

Presentation

3 good

Contribution

3 good

Limitations

None

Reviewer ufo52023-08-14

Thank you for your response. However, I would like to point out that without the explorative assumption and the variance-weighted algorithm design, LCB-based algorithms can enhance their dependency on $d$ by employing the reference-advantage decomposition [2] or data splitting scheme [JYW21]. So I would like to maintain my belief that asserting an improvement over LCB-based algorithms remains improper.

Authorsrebuttal2023-08-15

We thank the reviewer for the further comments on the comparison with LCB-based algorithms. We agree that for the enhanced dependence on $d$ alone, the LCB-based algorithm in [2] does not need the variance-weighted algorithm design. We, however, remark that, as far as we can tell, this enhanced dependence on $d$ in [2] requires the explorative assumption (their Assumption 1) (e.g., please see the “High-order Error from Correlated Advantage Function” paragraph in Section 5 of [2] and the detailed proof of Theorem 1 in Section D of [2]). To the best of our knowledge, we are unaware of any reference-advantage decomposition argument that has an enhanced dependence on $d$ without some explorative assumption such as Assumption 1 of [2]. If we’ve missed any reference otherwise, we would be happy to take any suggestion to make our comparison more accurate. We also agree that the data splitting scheme in [JYW21] also has an enhanced dependence on $d$. However, the data splitting scheme in [JYW21] only uses $K/H$ episodes (instead of using all $K$ episodes) for estimating the value functions. As a result, the improved $d$ dependence of the data splitting in [JYW21] comes at the cost of much looser dependence on $H$. As a concrete example, in the finite spectrum condition and under the well-explored dataset, the sub-optimality presented in Proposition 4.11 of [JYW21] incurs an additional factor of $\sqrt{H}$ (as also discussed by [JYW21] in the paragraph after Proposition 4.11). We appreciate your suggestion and will continue to update our revision to make accurate comparisons with LCB-based algorithms ([1][2]) based on the discussion here.

Reviewer niYV2023-08-16

Thank the authors for answering my questions. However, I do not think my concerns have been fully adreesed. The authors give high probability bound but have not clearly tell the readers what the randomness comes from. Does the randomness contains the randomness for the uniform distribution for selecting a policy from $\{\pi^1,\cdots,\pi^T\}$? I am also not convinced that the mixed policy is a Markov policy. The index is selected in the begining by uniform sampling and does not change. At any horizon, the action depends not only on the current state but also depends on the random index determined in the begining. So I do not think it is a Markov policy.

Authorsrebuttal2023-08-17

We thank the reviewer for further clarification of the questions when our initial response was not clear enough. **Randomness in the high-probability bound**. The randomness in the high probability bounds is solely from the randomness of the offline dataset, not from the policy mixture. To describe it in a context, the value sub-optimality (defined in Eq. (1)) of $\hat{\pi}$ in our theorems can be written explicitly as follows: $SubOpt_{\pi}(\hat{\pi}) = V_1^{\pi}(s_1) - V_1^{\hat{\pi}}(s_1) = V_1^{\pi}(s_1) - \frac{1}{T} \sum_{t=1}^T V_1^{\pi^t}(s_1)$. We thank the reviewer again for the good question. Even though the mixture policy is standard in the literature [ZWB21,XCJ+21,CXJA22], we will make sure to clarify the source of the randomness of our high-probability bound statements in our revised version. **Markovian of the mixture policy**. We remark that even though $\hat{\pi}$ is a mixture of several policies, i.e., $\hat{\pi}\_h(a_h|s_h) = \frac{1}{T} \sum_{t=1}^T \pi^t_h(a_h|s_h)$ for all $h \in [H]$, $\hat{\pi}_h$ is Markovian by definition since each $\pi^t_h$ is Markovian. We explain further in the following. By definition, a Markovian policy is a mapping solely from the current state to a distribution over the action space, completely independent of any previous states and actions. First, each $\pi^t_h$ is Markovian because it is fully characterized by another Markovian policy $\pi^{t-1}_h$ (by induction from the uniform policy $\pi^1_h$ which is Markovian), and the state-action value function estimate (that depends solely on the current state and action, not on any previous states and actions). For the full formulae of $\pi^t_h$, please see Line 5 of Algorithm 1 in our submission. Second, note that $\hat{\pi}\_h(a_h|s_h) = \frac{1}{T} \sum_{t=1}^T \pi^t_h(a_h|s_h)$ (please notice the same time index $h$ on both sides of the equation), the actions from $\hat{\pi}\_h$ solely depends on the current state $s_h$, not on any previous states and actions $s_{h’}, a_{h’}$ for $h’ < h$. Thus, by definition, each $\hat{\pi}_h$ is also Markovian. In other words, mixture — a simple weighting across several Markovian policies at the same time index $h$, does not break the Markovian property. As a concrete example, let’s consider the simple case that $T = 2$ and $\mathcal{A} = $ `{`$a_1, a_2$`}`, $\pi^1_h(a_1|s) = \pi^1_h(a_2|s) = 0.5$, and $\pi^2_h(a_1|s) = 0.2$, $\pi^2_h(a_2|s) = 0.8$. Then, by the definition of the mixture policy, we simply have $\hat{\pi}_h(a_1|s) = 0.5(\pi^1_h(a_1|s) + \pi^2_h(a_1|s)) = 0.5 * (0.5 + 0.2) = 0.35$ and $\hat{\pi}_h(a_2|s) = 0.5(\pi^1_h(a_2|s) + \pi^2_h(a_2|s)) = 0.5 * (0.5 + 0.8) = 0.65$. The mixture policy $\hat{\pi}_h$ is obviously Markovian (the probability action at each time step $h$ depends solely on the current state at time $h$, not on any previous states and actions). We’re happy to answer new questions if our clarification is still not clear enough.

Reviewer niYV2023-08-18

I'm grateful to the authors for replying to my questions. Nonetheless, I believe the authors have not addressed the concerns about the Markovian property and the high probability bound. Consequently, I'm modifying my score to 4. I understand that $\pi^t_h$ by itself adheres to the Markov policy. However, the mixed policy $\hat\pi$ isn't Markovian, as actions depend not only on the previous state but also on the random seed for selecting $t,h$ initially. It's important for the authors to recognize that $\hat\pi$ isn't equivalent to $\pi^t_h$. The policy is Markovian only if, at any time horizon, the agent uniformly samples $t,h$ and employs $\pi^t_h$. Regarding the high probability bound, the randomness doesn't encompass the randomness of the uniform distribution. Hence, the high probability bound doesn't imply the near optimality of the Markov policy $\pi^t_h$ and solely demonstrates accuracy for the mixed policy, which isn't Markovian. I'm open to increasing my score if the authors can demonstrate that the near optimality holds for $\pi^t_h$ considering both the randomness of the data distribution and the uniform distribution. Otherwise, I'm not convinced that a theoretical guarantee for the mixed policy holds enough significance. I can also give an example showing that the mixed policy may not be Markovian: There are two states $(s^1,s^2)$ and two actions $(a^1,a^2)$. Consider $3$-horizon case with time index t=0,1,2. Suppose we have $P(s^1|s,a^1)=P(s^2|s,a^1)=1/2$ and $P(s^2|s,a^2)=1$ for any $s\in\{s^1,s^2\}$. Also, suppose there are two policies that form a mixture policy uniformly (with probability 1/2 and 1/2), where each policy $\pi^t$ (where we are using the notation in the paper) always takes action $a^t$ for $t\in \{1,2\}$. Suppose the initial state $s_0=s^1$. We use the subscript to denote the timestep. Then, we have $P(s_2=s^2)=6/8$ and $P(a_2=a^2)=1/2$, and further, by Bayes rule, we have $P(s_2=s^2|a_2=a^2)=1$ because the action $a^2$ only leads to $s^2$. Then, we have $P(a_2=a^2|s_2=s^2)=4/6$. However, we have $P(a_2=a^2|s_2=s^2,s_1=s^1)=0$ since $s_1=s^1$ means we take $a=a^1$ in the beginning. This means the policy is not Markovian.

Authorsrebuttal2023-08-19

We thank the reviewer for the further comments. We are grateful for the reviewer's feedback which has helped improve the clarification of the presentation of our paper. We believe the root of the confusion might be Line 7 of Algorithm 1. We believe we would be better served if we wrote the output as $Uniform(\pi^1, \ldots, \pi^t)$ (there is no sampling over $h$). In other words, Algorithm 1 does not return a mixture policy, even though we may have referred to it as that in the rebuttal. Note that nowhere in the paper do we even use that word. What the Algorithm returns is one of the $T$ policy iterates uniformly sampled. Since each policy iterate is Markovian, what the algorithm returns is Markovian. We apologize for using that terminology in our earlier response. It might have caused unnecessary confusion. Regarding the bound itself, we give a high probability bound on the sub-optimality of the output of Algorithm 1. As we noted above, the output of Algorithm 1 is a Markovian policy. We should also add that the sub-optimality is defined w.r.t. any comparator policy (i.e., results hold for all $\pi \in \Pi^{all}$). We can always choose the comparator policy $\pi$ to be $\pi^*$, so we do bound the near optimality of the policy returned by the learner. The source of randomness as we said before is the offline data and any internal randomization of the algorithm — this is typical in guarantees of this form. We are not deviating from the norm here in specifying these bounds (e.g., see the bounds of similar algorithms in [ZWB21,XCJ+21,CXJA22] for the norm of presentation). For better presentation, we can take the expectation with respect to the randomness of the algorithm when writing out the bound. So, we will replace $SubOpt_{\pi} (\hat{\pi})$ in Theorems 1 and 2 by $ \mathbb{E} [SubOpt_{\pi}(\hat{\pi}) | \mathcal{D} ] = V_1^{\pi}(s_1) - \frac{1}{T} \sum_{t=1}^T V_1^{\pi^t}(s_1)$, and clarify that the expectation is taken w.r.t. uniform sampling of the iterates. The only source of randomness that would remain then is the randomness inherent in the offline dataset. Would that change be acceptable to the reviewer? We do think that will make things very clear. --- For an alternative presentation, we also propose a different approach to addressing the confusion. This approach is essentially equivalent to the first approach above but totally avoids using the term “uniform distribution” over $\pi^1, …, \pi^T$ as the output of Algorithm 1. We propose to revise Line 197 (and thus Line 7 of Algorithm 1 accordingly) to read: “After $T$ update steps, Algorithm 1 *deterministically* outputs a single policy $\hat{\pi} = (\hat{\pi}_1, \ldots, \hat{\pi}_H)$ where we define $\hat{\pi}_h(a|s) = \frac{1}{T} \sum\_{t=1}^T \pi^t_h(a|s)$ for any $h \in [H]$" (please note that $\hat{\pi}_h$ is computed *deterministically* from $\pi^1_h, …, \pi^T_h$ without defining “uniform distribution” as the output of Algorithm 1). With that, since each policy iterate $\pi^t$ is Markovian, what the algorithm returns $\hat{\pi}$ is also Markovian. Now, by linearity of the value function $V^{\pi}$ in terms of $\pi$, we have $V_1^{\hat{\pi}} (s_1)= \frac{1}{T} \sum_{t=1}^T V_1^{\pi^t} (s_1) $. This is a *deterministic* relation, without defining “uniform distribution” as the output of Algorithm 1. Consequently, we have the following *deterministic* relation: $ SubOpt\_{\pi}(\hat{\pi}) = V_1^{\pi}(s_1) - V_1^{\hat{\pi}}(s_1) = V_1^{\pi}(s_1) - \frac{1}{T} \sum_{t=1}^T V_1^{\pi^t}(s_1) = \frac{1}{T} \sum_{t=1}^T ( V_1^{\pi}(s_1) - V_1^{\pi^t}(s_1))$. Here, there is no internal randomization in the algorithm (as we do not define “uniform distribution” as the output of Algorithm 1) and the only source of randomness in our algorithm and in our high-probability bounds is the offline data. Would that change also be acceptable to the reviewer? We think either of the approaches would help address the unnecessary confusion in our submission version. Again, we apologize for the unnecessary confusion. --- We would also add that in our proof of bounding the value sub-optimality of $\hat{\pi}$, we do bound the value sub-optimality $V_1^{\pi}(s_1) - V_1^{\pi^t}(s_1)$ of each individual $\pi^t$. Specifically, using Lemma B.4 in our supplementary: $V_1^{\pi}(s_1) - V_1^{\pi^t}(s_1)= \sum_{h=1}^H \mathbb{E} [\mathcal{E}\_h^{\pi^t}(Q_h, Q_{h+1})] + Q_1(s_1, \pi^t_1) - V_1^{\pi^t}(s_1) + SubOpt_{\pi}^{M(Q, \pi^t)}(\pi^t)$ for any $t \in [T]$. From Lemma B.6: $\sum_{t=1}^T SubOpt_{\pi}^{M(Q, \pi^t)}(\pi^t) ≤ Hb \sqrt{T \ln |\mathcal{A}|}$ (it holds deterministically). The high-probability bounds come from bounding $\sum_{h=1}^H \mathbb{E} [\mathcal{E}\_h^{\pi^t}(Q\_h, Q_{h+1})] + Q_1(s_1, \pi^t_1) - V_1^{\pi^t}(s_1)$ for any $t$. In fact, we bound $\sum_{h=1}^H \mathbb{E} [\mathcal{E}\_h^{\tilde{\pi}}(Q_h, Q_{h+1})] + Q_1(s_1, \tilde{\pi}_1) - V_1^{\tilde{\pi}}(s_1)$ uniformly over all $\tilde{\pi} \in \Pi^{soft}$ as $\pi^t \in \Pi^{soft}$.

Authorsrebuttal2023-08-21

As the current discussion phase is closing, for the sake of any other discussions after this phase, we thought we would add our comment to the reviewer’s example, in case it was not clear from our previous response. We believe the example was constructed out of the confusion and does not support any conclusion (let alone some miscalculations in the example). In short, it's merely by the very definition that $\hat{\pi} = (\hat{\pi}\_1, \ldots, \hat{\pi}_H)$ where $\hat{\pi}_h(a_h|s_h) = \frac{1}{T} \sum\_{t=1}^T \pi^t_h(a_h|s_h)$ which is Markovian by the very definition. In the example, given $\pi^t_h(a^t | s) = 1$ for all $t = 1,2$ and $h = 1,2,3$ in the example, it is merely that $\hat{\pi}_h(a_h = a^t | s_h = s) = 0.5$ for all $t = 1,2$ and $h = 1,2,3$. Finally, we believe that our proposed steps for improving the presentation of our paper in our earlier responses should clearly address this unnecessary confusion. We apologize again for the unnecessary confusion and we appreciate the reviewer for taking the time to read our rebuttal.

Reviewer YpYn2023-08-19

Thanks for the responses from the authors. Since my major concern #1 as been well alleviated, I decided to raise my score.

Authorsrebuttal2023-08-19

Thank you for taking your time to read our rebuttal and for giving your feedback. We will integrate your suggestions (including the “minor” ones) in our next revision.

Reviewer FhSe2023-08-21

Thanks for the clarification. I will keep my belief as before.

Authorsrebuttal2023-08-21

Thank you for taking the time to read our rebuttal.

Program Chairsdecision2023-09-21

Decision

Accept (poster)

© 2026 NYSGPT2525 LLC