- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper identifies the inefficiency in dLLMs caused by oversizes generation canvases. To address this, it proposed BC-dLLM with two components. SGC truncates the model's own long reasoning trajectories and regenerates the remainder within a restricted canvas, then fine-tune the model. PASS estimates decoding progress from <EOS> probabilities and shrinks the visible suffix window accordingly. Experiments on LLaDA-8B-Instruct, LLaDA-1.5 and Dream-7B across four math reasoning benchmarks show that this method reduces decoding steps by about 13x and latency by about 10x.
Critical strengths
- A well-motivated perspective. Instead of targeting step-level efficiency, this paper points out trajectory-level inefficiency caused by the oversized canvas.
- Two simple components. SGC reduces the number of tokens to generate and PASS reduces the cost per step. PASS is training-free and SGC needs only about 4K training steps.
- Comprehensive experiments. This paper conducts the experiments on three models across four benchmarks. The result shows that it reduces decoding steps by roughly 13–15x and latency by roughly 10x.
Critical weaknesses
- The sources of the speedup are not disentangled. There are several separate factors contribute to the reduction: confidence-threshold parallel decoding, <EOS> early termination, SGC and PASS. It is unclear the concrete contribution for the proposed methods.
- Accuracy drops consistently, not comparable. The performance for all benchmarks is worse than the original model. This should be characterized as an accuracy-efficiency trade-off.
- The simple length-control baselines are missing. Natural and simpler alternatives to induce the verbose CoTs should be running the original model with Fast-dLLM decoding and <EOS> early stopping on a smaller canvas.
Questions for authors
- Could you report the performance for: (a) SGC only, (b) PASS only, (c) SGC+PASS applied to the original model?
- Could you give more results on out-of-domain tasks (e.g., GPQA or MBPP for code) when SGC is trained only on GSM8K and MATH-500?
- What is the width of the restricted canvas uses to regenerate the fast-thinking segment? How is the truncation point sampled?
4 comments
Q2: Provide more results on out-of-domain tasks.
Thank you for your constructive comment. Our primary goal is to improve long-CoT generation efficiency while preserving the parallel decoding capability of dLLMs, which is why our main experiments focus on GSM8K and MATH-500. Since most standard general-domain benchmarks evaluate answer-option log-likelihoods rather than free-form generation efficiency, we additionally evaluate two free-form out-of-domain tasks: TruthfulQA-Gen and CNN/DailyMail.
Table 3. Out-of-domain evaluation on free-form generation tasks (score with speedup in parentheses).
| Method | TruthfulQA-Gen | CNN/DailyMail |
|---|---|---|
| Base (Fast-dLLM) | 41.2 (1.0×) | 18.2 (1.0×) |
| Ours | 40.8 (1.1×) | 18.1 (1.4×) |
As shown in the table, our model preserves out-of-domain generation quality while achieving 1.1× and 1.4× speedups, respectively. These moderate speedups are expected because such general-purpose generation tasks naturally involve less explicit long-CoT reasoning.
Q3: What is the width of the restricted canvas used to regenerate the fast-thinking segment? How is the truncation point sampled?
As stated in line 237 of our manuscript, the truncation points are randomly selected. Consequently, the remaining canvas size after truncation ranges between 128 and 512 tokens.
We hope these responses address your concerns. Please let us know if you have any further questions.
Dear Reviewer ZjpS,
Thank you once again for your time and constructive feedback. Guided by your suggestions, we have substantially strengthened the paper by adding the component-wise ablation of SGC and PASS, refining our discussion of the accuracy-efficiency trade-off, and clarifying the canvas and truncation protocols.
With less than 24-hours left in the discussion window, we would be very grateful to know whether our rebuttal has satisfactorily addressed your concerns. If any questions remain, we would be glad to discuss them further before the discussion closes.
All the best,
Authors of Submission 11654
discussion ddl
Dear Reviewer ZjpS,
Thank you once again for your time and constructive review of our paper.
With less than one hour remaining before the discussion window closes, we wanted to kindly follow up to see if our rebuttal and additional experiments have well addressed your concerns. We would be very grateful for any final thoughts or feedback you might have before the deadline.
Thank you very much for your time and consideration.
Best regards,
Authors of Submission 11654
We thank the reviewer for the thoughtful comments and suggestions. Our responses are as follows.
Thank you for this valuable suggestion. To disentangle the sources of speedup, we evaluate (a)
SGC only, (b)PASS only(applyingPASSdirectly to the original base model), andSGC + PASSon GSM8K across different confidence thresholds (accuracy (%) with decoding steps in parentheses):Table 1. Ablation study disentangling
SGCandPASSon GSM8K across confidence thresholds.The results clearly illustrate the distinct and complementary roles of both components:
SGC onlyandPASS only:SGC onlyaccounts for the majority of the step reduction (e.g., reducing steps from 89 to 44 at a threshold of 0.9) by training the model to generate concise CoT trajectories with high parallelism. Meanwhile,PASS onlyapplied to the original base model maintains high accuracy (82.1% at 0.9) while reducing per-step computation via dynamic<MASK>pruning, though it still requires more steps (89 at 0.9) because its CoT trajectories remain uncompressed.SGC + PASScombines both benefits, further reducing decoding steps from 44 to 39 (at 0.9) and from 33 to 30 (at 0.8) alongside lower per-step FLOPs, while achieving nearly the same accuracy asSGC only(e.g., 74.5% for both at 0.7). We will include this ablation table in the revised manuscript.Thank you for your constructive comment. We agree to refine our wording in the revision to explicitly acknowledge the accuracy–efficiency trade-off relative to the Base model. We would also like to place our contribution in the context of different acceleration baselines:
In our experiments, besides the Base model, we compare against training-free acceleration frameworks such as
Fast-dLLMandKLASS, as well as the training-based methoddParallel. WhileFast-dLLMandKLASSexhibit smaller performance drops, their acceleration gains are relatively limited. Conversely, training-based approaches can achieve much greater speedups at the cost of a moderate accuracy trade-off. Notably, compared todParallel, our method achieves lower inference latency while simultaneously delivering higher accuracy. This demonstrates that our approach offers substantial practical value by advancing the overall efficiency–performance boundary, rather than representing a trivial accuracy–efficiency trade-off.Thank you for suggesting this baseline comparison. Here, we evaluate a prompt-based length-control baseline (
Fast-thinking prompt only) that incorporates a CoT-shortening instruction into the prompt to encourage more concise reasoning. The comparison on GSM8K across different confidence thresholds is presented below:Table 2. GSM8K results with LLaDA under different confidence thresholds (accuracy (%) with decoding steps in parentheses).
As the results demonstrate, although simple prompt constraints can shorten the model's CoT output, they lead to severe performance degradation compared to our method. For example, at a confidence threshold of 0.9, the prompt-only baseline achieves only 76.2% accuracy (with 56 steps), whereas our method reaches 81.0% (+4.8%) with just 39 steps.