- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 3
Read review
Paper summary
This paper proposes LP-SMC, a diversity-preserving resampling method for diffusion LLM decoding. It combines Chopthin resampling with a weight floor to preserve promising reasoning lineages during SMC decoding.
Overall, the method is well motivated and experimentally well analyzed, but its practical quality–compute trade-off remains unclear due to the high particle-based inference cost and the lack of evaluation under smaller particle budgets.
Critical strengths
(1) The paper identifies the problem that existing SMC methods may lose potentially useful reasoning lineages, and uses Chopthin resampling and a weight floor to reduce lineage collapse caused by stochastic resampling and low weights, respectively.
(2) The method outperforms SMC on two models and multiple benchmarks, and validates the method through multiple metrics including surviving lineages, distinct outputs, oracle coverage, and final accuracy. The experimental analysis is detailed.
Critical weaknesses
(1) The overall inference cost is relatively high, and the practical quality–compute trade-off of the method is not sufficiently clear. Although LP-SMC itself does not introduce additional computational overhead compared with SMC, the whole method is built on SMC decoding with particles, which requires substantially more computation than standard decoding. The paper mainly compares with SR-SMC under the same budget, but lacks comparisons with simpler methods such as best-of-.
(2) The main experiments are all based on , so it is unclear whether the advantages of LP-SMC can stably generalize to lower-compute settings such as .
Questions for authors
(1) The experimental setup mentions confidence-threshold , which I understand as corresponding to decoding with a non-fixed number of steps. However, the setup also mentions 8 decoding steps per block. Is 8 the maximum number of steps? How are these two hyperparameters made compatible?
3 comments
Thank you for the additional experiments and clarifications. I appreciate the authors' efforts to address my concerns. The new results demonstrate that LP-SMC consistently achieves better solution quality than SR-SMC across several settings. However, when compared with the simpler Best-of-N baseline, the observed improvements appear relatively modest. Overall, I am satisfied with the current assessment and will keep my score unchanged.
Best-of-N comparison under the four selection rules
Thank you for reading the new results and for the summary.
We agree that under the reward-model selector the difference from best-of- is modest. However, we think the other selection rules are worth considering too. The table gives the same two comparisons under the four selection rules of the paper (Section 5.3), on the same saved candidates (LLaDA-1.5, LP-SMC with ). Accuracy in %.
| Selector | MATH-500: LP-SMC | MATH-500: best-of-18 | MBPP: LP-SMC | MBPP: best-of-16 |
|---|---|---|---|---|
| Reward model | 55.4 | 54.0 | 50.6 | 50.2 |
| Confidence | 43.4 | 40.6 | 40.9 | 41.1 |
| Majority vote | 52.6 | 47.4 | 41.3 | 38.8 |
| Distinct-output vote | 50.2 | 47.6 | 41.3 | 40.5 |
LP-SMC is higher in seven of the eight comparisons. On MATH-500 the difference is larger under each of the other three rules than under the reward model, and on MBPP it is larger under the two voting rules. Under majority voting the difference is on MATH-500, with 95% confidence interval , and on MBPP, with interval . On MATH-500, LP-SMC is guided by the PRM during decoding, so we compare it with best-of-18 to allow for the extra computation.
So the gain over best-of- is modest with a reward-model selector, as you say, and larger under most of the other selection rules.
Best-of- is also not a single baseline in this literature. PG-DLM and FK-Steering select the sample by reward, the November revision of SR-SMC (our baseline) selects it by log-probability, and S$^3$ defines its best-of- baseline as independent samples with a majority vote; the SR-SMC revision lists majority voting as a further compute-matched baseline (its Table 11). We therefore report the comparison under both the reward model and the vote. Under the vote, LP-SMC is higher in both settings:
| Setting (majority vote, %) | LP-SMC | Independent sampling | Difference (95% CI) |
|---|---|---|---|
| LLaDA-1.5, MATH-500, PRM-guided | 52.6 | 47.4 | |
| LLaDA-1.5, MBPP, no PRM guidance | 41.3 | 38.8 |
To our knowledge, earlier SMC decoders for diffusion language models each report one final selector; we report four on the same candidates so that this comparison can be read under each.
We will add these tables to the revision as an additional ablation and state this scope there. If these results address your concerns, we would be grateful if you would consider raising your score. Thank you again for your time and for a review that improved the paper.
Response to Reviewer 8UVy
We thank the reviewer for the review, which we found informative. The points on smaller budgets and best-of-N led to the new experiments below.
Regarding the question (confidence threshold of 0.9 and 8 decoding steps):
Yes, the number of steps is not fixed, and our setup sentence was imprecise: it should say eight blocks, not eight decoding steps per block. The value 8 is not a maximum. We kept the settings identical to the baseline’s hyperparameters, as they report them, and we use its decoder unchanged, which follows Fast-dLLM’s confidence-based parallel decoding. In that decoder, the threshold rule takes over whenever a threshold is set: in every step, it commits all masked positions of the current block whose confidence is at least 0.9, always commits at least the most confident one, and repeats until the block has no masked positions left. A block of 32 tokens therefore takes between 1 and 32 steps.
The value 8 is the
stepsargument of the baseline’s configuration, which is the generation length divided by the block size (256/32), or one scheduled step per block. It defines the fixed schedule used when no threshold is given. When the threshold is set, the decoder ignores the fixed token quota and continues until the block is complete. The two settings therefore do not conflict. We will clarify the setup description: the sequence is decoded in eight blocks of 32 tokens, and the number of steps within each block is set adaptively by the 0.9 threshold.Regarding Weakness 2:
Under our rule κ=lnN and the gate ESS<N/2, resampling cannot trigger for N≤5. In that regime, particle states evolve independently, so with the same final selector the method reduces to best-of-N independent decoding. We therefore ran the ablation at N=6 and N=8.
The following MATH-500 comparison uses LLaDA-1.5 with decoder-confidence guidance and everything else fixed:
We also evaluated N=8 on all eight model-benchmark settings under decoder-confidence guidance. Accuracy is measured after the final RM selector.
All rows in these response tables are scored with the same evaluation pipeline. Minor grading differences change the N=16 MATH-500 results from 49.0 and 52.4 in Table 1 to 49.2 and 52.8 here; we will document the exact conventions in the appendix.
On MATH-500, the measured gain is largest at the smallest tested budget, and at N=8 LP-SMC is higher in all eight model-benchmark settings. We will add both tables to the appendix.
Regarding Weakness 1:
We agree that particle decoding uses roughly N times the decoder compute of single-sample decoding. This cost is shared by SR-SMC and best-of-N, and LP-SMC adds no measurable overhead to SMC (Appendix B.6). The MATH-500 table above gives the trade-off within SMC: LP-SMC with 6 particles reaches the accuracy of SR-SMC with 16 (49.2).
We also ran best-of-N controls: independent complete rollouts of the same decoder, with the same final RM selector. We report the control on MBPP without PRM guidance and on MATH-500 with PRM guidance. PRM guidance adds one reward-model call per particle at each block boundary; each call is a single forward pass over the current prefix and generates no tokens. We give independent sampling 18 complete rollouts on MATH-500 as an allowance for the intermediate PRM evaluations. LP-SMC uses N=16 particles. On MBPP, where no intermediate PRM is used, both methods use N=16.
The point estimates favour LP-SMC in both settings, with the intervals given in the table. With majority voting as the final selector, which requires no additional reward-model call at selection, PRM-guided LP-SMC is 5.2 points above best-of-18 on MATH-500, with a 95% confidence interval of +2.2 to +8.4. Resampling raises the share of correct candidates from 38.1% to 45.7%. Under decoder-confidence guidance the rise is 1.3 points, so the benefit over best-of-N grows with how informative the guide is. We will add the best-of-N baseline and the full comparison, including systematic + floor, to the revised paper.
We thank the reviewer again and are glad to clarify anything further.