- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper proposes ThinkMid, a training-free decoding method that monitors mean token entropy of each response sentence. When a sentence is uncertain, the same model reviews its response in a temporary turn; if an error is confirmed, a correction replaces the sentence and decoding resumes. On AIME 2025, GPQA Diamond, and an HLE subset with two LRMs, ThinkMid improves over the single-solve baseline and is reported to outperform compute-matched sequential, parallel, and iterative scaling baselines.
Critical strengths
- The idea of spending extra compute inside the response, where errors arise and propagate, is well motivated and complementary to existing test-time scaling.
- The method is simple, training-free, and works across two architectures and thinking formats.
- Separating the trigger (entropy) from the decision to edit (self-review) is a sensible guard against harmful self-correction, and the per-problem compute matching is careful.
Critical weaknesses
- Source of gains is unclear. ThinkMid discards high-entropy segments, adds thinking budget, and regenerates them, but these effects are never disentangled. A simple baseline that rolls back and resamples flagged segments without review is missing. The random-placement ablation still uses review, and the Section 5 "plain continuation" keeps the flagged sentence. Fixed-interval placement also matches ThinkMid's accuracy, suggesting extra regeneration opportunities, not self-verification, may drive the gains.
- Weak statistical evidence limited by the small evaluation sets. All results are single-run without confidence intervals. Margins over the strongest baseline amount to 0–4 attempts on AIME and 1–2 questions on GPQA and HLE, well within noise.
- Narrow analysis and evaluation. Ablations and τ selection use only AIME with one model (30 problems), and τ appears tuned on the test set. Benchmarks are small and limited to math/science QA.
- Compute accounting and baselines. It is unclear whether reused context is counted for auxiliary turns. Under the matched budget, BoN has only N≈1–2, making it a weak baseline. Latency is not reported.
Questions for authors
- How does ThinkMid compare with entropy-triggered resampling at matched compute?
- Can you report multi-seed results with confidence intervals and paired significance tests?
- How are prompt tokens counted for review and correction turns relative to BoN candidates?
- What are the trigger rates, review verdict distributions, and correction acceptance rates?
5 comments
Weakness #4 and Question #3: Compute accounting, BoN baselines, and latency.
Thank you very much for your comments.
(1) Compute accounting.
In our token accounting, cached prefix tokens are not counted again when reused by auxiliary turns, since the prefix is reused without repeated prefilling. All newly processed tokens are counted, including the review and correction prompts and their generated thinking and responses. Cache reuse is described in Section 3.3, while Section 4.1 and Appendix C.3 describe token accounting; we will make the treatment of cached prefixes explicit in the revision.
(2) BoN under the matched budget.
For BoN, each candidate’s prompt, thinking, and response tokens count toward the same per-problem budget, as specified in Appendix C.4. Its limited number of candidates therefore reflects the modest additional token budget required by ThinkMid: under the evaluated candidate-generation settings, generating more candidates would require a larger budget. This highlights ThinkMid’s advantage in allocating limited additional computation to targeted review and correction, whereas BoN spends its budget generating additional complete solutions.
(3) Latency.
Following your suggestion, we also report latency results in Table A2.
Table A2. Latency comparison across models and benchmarks.
| Model | Benchmark | Vanilla | Thinking Forcing | BoN-Judge | Self-Refine | ThinkMid |
|---|---|---|---|---|---|---|
| Qwen3.6-35B-A3B | AIME 2025 | 845 | 1,479 | 1,371 | 1,332 | 1,484 |
| Qwen3.6-35B-A3B | GPQA | 667 | 834 | 780 | 770 | 864 |
| Qwen3.6-35B-A3B | HLE | 811 | 1,269 | 1,182 | 1,250 | 1,148 |
| GPT-OSS-20B | AIME 2025 | 756 | 1,284 | 1,176 | 1,264 | 1,279 |
| GPT-OSS-20B | GPQA | 597 | 961 | 936 | 860 | 774 |
| GPT-OSS-20B | HLE | 789 | 1,410 | 1,371 | 1,404 | 1,161 |
Question #4: Trigger rates, review verdict distributions, and correction acceptance rates.
Thank you very much for your valuable suggestion. Following your suggestion, we report the trigger rates, review verdict distributions, and correction acceptance rates in Table A3.
Table A3. Trigger rates, review verdict distributions, and correction acceptance rates.
| Model | Benchmark | Sentence trigger rate (%) | Review verdicts: correct / error / uncertain (%) | Correction acceptance rate (%) |
|---|---|---|---|---|
| Qwen3.6-35B-A3B | AIME 2025 | 3.1 | 72.0 / 19.5 / 8.5 | 93.8 |
| GPT-OSS-20B | AIME 2025 | 16.4 | 59.8 / 26.2 / 14.0 | 60.7 |
| Qwen3.6-35B-A3B | GPQA | 6.1 | 84.5 / 10.3 / 5.2 | 70.0 |
| GPT-OSS-20B | GPQA | 16.3 | 56.1 / 32.5 / 11.5 | 37.5 |
| Qwen3.6-35B-A3B | HLE | 8.1 | 70.1 / 18.9 / 11.0 | 96.2 |
| GPT-OSS-20B | HLE | 25.6 | 52.6 / 32.8 / 14.6 | 57.0 |
Thank you for the additional experiments. The rebuttal addresses several concerns but does not fully resolve the limitations. I therefore maintain my original rating.
We sincerely appreciate your careful review and constructive comments. Please see our responses to your comments one by one below.
Thank you very much for your valuable suggestions. We would like to clarify that, as highlighted in the Critical Strengths identified by Reviewer 4Aag, ThinkMid separates uncertainty from correctness: high entropy triggers review rather than automatically discarding or rewriting a segment. The Review stage determines whether to request correction, while the Correction stage determines whether and how to revise the segment.
Following your suggestion, we added four ablations, including entropy-triggered resampling, to examine the contributions of mid-response intervention, the verification gate, and targeted local correction. ThinkMid achieves 83.3% accuracy, compared with 73.3% for entropy-triggered resampling. Please refer to our response to Reviewer 4Aag Weakness #3 for the experimental details and complete results.
We appreciate your valuable suggestions.
(1) Multi-seed evaluation and statistical uncertainty.
We would like to clarify that our AIME 2025 results aggregate runs with four distinct random seeds, rather than a single run. GPQA and HLE were evaluated with one attempt per question because their evaluation sets contain more questions, following common practice in related work.
Following your suggestion, we additionally report multi-seed results with confidence intervals and paired significance tests. Please refer to our response to Reviewer UnNa Weakness #2 for the detailed results. ThinkMid improves over Vanilla under every tested seed, supporting the consistency of its gains in this setting.
(2) Overall performance across settings.
We acknowledge that gains over the strongest baseline in individual settings can be modest. However, the strongest baseline varies across settings, and ThinkMid shows larger overall gains when compared against any fixed baseline across the six settings. We believe ThinkMid performs better overall than the compute-matched baselines: it ranks first or second in all six settings and consistently improves over Vanilla, whereas some baselines reduce accuracy in certain settings.
Thank you very much for your valuable suggestions.
(1) Additional ablations and sensitivity analysis.
Following your suggestion, we added ablation experiments and entropy-threshold sensitivity analysis on HLE for Qwen3.6-35B-A3B, with results reported in Table A1. At our default threshold, entropy-guided review achieves higher accuracy than both fixed-interval and random placement, while consuming fewer tokens than fixed-interval review. This supports more effective allocation of review computation. Moreover, all three tested thresholds yield the same accuracy, demonstrating stable accuracy across the tested threshold range.
Table A1. Ablation and entropy-threshold sensitivity results on HLE with Qwen3.6-35B-A3B.
(2) Threshold selection.
Regarding threshold selection, we clarify that we did not tune τ separately for individual models or benchmarks. We selected a generally effective threshold during implementation and used the same value across all settings. We expect that tailoring τ to individual settings could yield further improvements.
(3) Benchmark coverage and further evaluation.
Regarding evaluation scope, our experiments cover three commonly used reasoning benchmarks and multiple subject categories. GPQA and HLE, together with repeated evaluations on AIME 2025, are widely used in related reasoning-model research. These benchmarks were deliberately selected to assess tasks requiring reasoning: our claim concerns improvements on reasoning tasks, rather than universal improvements across all task types.
Given the limited rebuttal period, we could not complete further experiments. If the paper is accepted, we commit to including ablation studies and hyperparameter sensitivity analyses across all model–benchmark settings, together with evaluations on larger benchmarks covering additional task categories, in the camera-ready version.