- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper extends the study of the sharpening phenomenon from coding and mathematical reasoning to agentic tasks. The authors introduce sharpening tax, a metric intended to quantify the difference in test-time scalability between base and post-trained LLMs. They further propose Posterior-Tempered Group Sampling (PTGS), which dynamically adjusts the sampling temperature based on task difficulty. The paper addresses an important problem, and the proposed method appears effective in the reported experiments. However, the presentation needs substantial improvement, and several concepts and results require clarification.
Critical strengths
- The paper investigates a novel and important problem.
- The proposed PTGS method demonstrates promising empirical effectiveness.
Critical weaknesses
- The manuscript does not consistently adhere to the conventions of formal academic writing. Frequent colloquial expressions and awkward phrasing reduce its clarity and readability. Substantial language revision would improve the presentation.
- The manuscript would benefit from a more conventional and clearly structured organization. The related work discussion appears only in the appendix, making it difficult to contextualize the contributions when reading the main text. Moreover, the “Preliminaries” section primarily describes the experimental setup rather than providing the background concepts and notation needed to understand the methodology. Much of this material would be better placed in the methodology or experimental setup sections.
- The paper attempts to convey several distinct messages without clearly establishing how they contribute to its central argument. This makes the narrative feel fragmented. For example, the connection between the analyses in Figures 2 and 3 and the central concepts of sharpening tax and PTGS is unclear. The paper would benefit from a more coherent narrative that explicitly links these analyses to its main contributions.
Questions for authors
- I understand the mathematical definition of A(K) and find it reasonable. However, the motivation for S(K) is less clear. What is its intuitive or operational interpretation? Furthermore, why is it appropriate to quantify sharpening tax using the difference in A(K) or S(K) between a base model and its post-trained counterpart? Given that pass@1 and pass@K vary across models and tasks, it is unclear why these quantities can be compared directly in this manner. In particular, a positive tax does not necessarily seem to imply a loss of test-time scalability. Please clarify the interpretation of this metric and the conditions under which it supports such a conclusion.
- Could the authors clarify the conceptual relationship between the proposed diagnostic metric and PTGS? The current presentation does not clearly explain how the sharpening-tax analysis motivates the design of the method. As I understand it, PTGS increases the sampling temperature for difficult tasks to encourage exploration. How does this mechanism specifically address the behavior captured by sharpening tax?
- In Figure 1, PTGS decreases pass@1 while increasing pass@K, whereas in Table 2, it improves both. Could the authors explain what accounts for these different effects on pass@1 and clarify whether they arise from differences in the experimental settings?
4 comments
Response to Reviewer CgeC (2/3)
Q1: What is the operational meaning of ? Why can the base–post difference in or be compared when pass@1 and pass@ vary across models? Under which conditions does a positive tax imply a loss of test-time scalability?
(1) Operational meaning of . Given that you already agree with the formulation of , we want to clarify that the calibrated scalability is just a normalized version of with respect to the total budget () and the accuracy headroom (1-pass@). This measures the average scalability (coverage gain) over all budgets up to , amplified by the value of pass@. The former operation, i.e., averaging over , allows us to bound the range of the metric to [0,1], whereas the latter operation, i.e., dividing by (1-pass@), is important for a fair comparison of "effective scalability" across models, i.e., how scalable the model is given its accuracy headroom. Note that the higher the pass@ of the policy, the less room exists for additional gain from repeated sampling.
(2) Why the base-post difference is comparable. First of all, since is just a linear summation over , the direct comparison between the of base and post is basically the same as the direct comparison of pass@ or pass@ between base and post, which is the classic evaluation regime of LLM post-training (Zhao et al. 2025, Yue et al. 2025, Shen et al. 2026). Both policies run on the same tasks with the same budget, the same metric scale, and the same decoding configurations. Therefore, the direct "difference" between base and post across tasks that you mentioned is exactly what we want to measure through our Sharpening Tax formulation.
Meanwhile, ratio-difference-based metrics like our calibrated scalability have also been widely adopted in the literature for evaluating diverse models over two quantities simultaneously under a newly proposed criterion, i.e., effective robustness (Taori et al. 2020, Miller et al. 2021, Baek et al. 2022).
(3) When a positive tax implies a loss. By definition, a positive tax always implies a loss of test-time scalability, i.e., the post-trained model gains less from additional rollouts than its base model up to the budget . This is our intentional design philosophy, as written in the Abstract and Section 4, rather than a methodological limitation. What a positive tax does not imply by itself is a loss of coverage, i.e., a lower pass@ at the end-point budget. This distinction is intended, as written in the Abstract and Section 4, and our Corollary 2 makes it explicit: , where the first term denotes the coverage (pass@) gap, , at a specific end-point budget , and the second term, , represents the pass@ difference averaged over all budgets . Since a positive tax means , it falls into one of three cases:
| Cases | Interpretation |
|---|---|
| Base achieves higher final coverage and Post gets stuck in a flat scaling curve; additional sampling from Post adds little value. | |
| Base achieves higher final coverage and Post has a quickly saturating scaling curve; additional sampling from Post adds little value. | |
| Post achieves higher final coverage, yet its scaling curve saturates quickly, so additional sampling from it adds little value. |
Here, we want to clarify that we do not claim that every post-training compromise the coverage of base models. Post-training sharpening either improves pass@K (i.e., ) or not, depending on model size, test-time sampling budget, and dataset. However, the shape of the test-time scaling curve is consistent across model sizes and datasets. Whether post-training ends with a higher coverage or not, its scaling curve always saturates quickly compared to the base model’s scaling curve. That’s why we need a new metric, Sharpening Tax, to quantify this change in the curve's shape.
We revised the Abstract, Introduction, and Section 4 to make these points clear. Thanks again!
Reference
- Zhao et al. 2025, Echo Chamber: RL Post-training Amplifies Behaviors Learned in Pretraining
- Yue et al. 2025, Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?
- Shen et al. 2026, Understanding Reasoning from Pretraining to Post-Training
- Taori et al. 2020, Measuring Robustness to Natural Distribution Shifts in Image Classification
- Miller et al. 2021, Accuracy on the Line: On the Strong Correlation Between Out-of-Distribution and In-Distribution Generalization
- Baek et al. 2022, Agreement-on-the-Line: Predicting the Performance of Neural Networks under Distribution Shift
Response to Reviewer CgeC (3/3)
Q3: In Fig. 1, PTGS decreases pass@1, whereas in Table 2 it improves both metrics. Do the settings differ?
We would like to clarify that the settings for Fig. 1 and Table 2 are identical and that PTGS actually improves pass@1 in both. In Fig. 1(c), the line style marks the RL algorithm (solid: PPO; dashed: GRPO), and the color shade marks PTGS (lighter: baseline; darker: PTGS). Against the baseline of the same algorithm, PTGS improves pass@1 and pass@128, and it also lowers the tax.
Invitation to rebuttal discussion
Dear Reviewer CgeC,
As the discussion period is coming to an end, we wanted to briefly follow up on our responses. Following your suggestions, we:
- Substantially revised the writing to remove colloquial expressions and awkward phrasing.
- Reorganized the paper by moving a Related Work section into the main text and renaming “Preliminaries” to “Preliminary Experiment Setup”.
- Strengthened the links between the analyses, the Sharpening Tax, and PTGS by revising the bridge paragraphs and adding a takeaway block at the end of each core section.
- Answered your questions on the operational meaning of S(K) and why base and post-trained models can be compared directly (Q1), and on how the tax motivates PTGS (Q2). For Q3, we clarified that Fig. 1 and Table 2 use the same setup and that PTGS improves pass@1 in both.
We highlighted major edits in red in the revised PDF. If our revisions have addressed your concerns, we would appreciate it if you could consider updating your assessment. Otherwise, we would be happy to provide further clarification!
Thank you again for your time.
Best regards, The Authors
Response to Reviewer CgeC (1/3)
Dear Reviewer CgeC, we are really grateful to you for taking the time to provide this careful and constructive review, which was crucial to improving our manuscript. We respond to each weakness (W) and question (Q) below. Please also refer to the revised PDF, which reflects your comments and marks some major edits in red.
We agree and reorganized the paper into the conventional structure that you describe. First of all, we brought a compact Related Work section into the main text. We also changed the section name "Preliminaries" (Sec. 2) to "Preliminary Experiment Setup" so that the section title matches its content. We also remakrably polished the main text of draft to rewrite colloquial expressions and awkward phrasing in a more formal style. If you have additional suggestions for improving the writing, please point out the specific items. We are happy to reflect them as well.
Thanks for pointing this out; we understand your concern. The paper flows through three parts: (1) core observations (Figs. 2-3) and deeper analyses of the underlying mechanism, distribution sharpening (Figs. 4-5), (2) a metric to quantify this sharpening effect in a single number (Sec. 4), and (3) a method to mitigate this tax, inspired by our analyses in Figs. 4-5 (Sec. 5). Specifically,
In our original draft, we tried to bridge the sections seamlessly at the beginning of each section (Sec. 4 and Sec. 5). To further strengthen the connections, we not only polished the bridging sentences but also added takeaway blocks at the end of key sections to guide easier reading.