REVIEW 5 major objections 5 minor 1 cited by
Stream Aligner: Efficient Sentence-Level Alignment via Distribution Induction
T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash
Pith's one-line read Stream Aligner claims that a small model rewriting and feeding back each sentence of a large model's output during generation aligns the response better, improves math reasoning, and lowers latency compared with one-shot whole-answer…
desk verdict Sentence-level iterative correction is a real extension of Aligner, but the QA evaluation is too inconsistent and circular to support the headline claim. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The operative mechanism is the sentence-level residual objective: training maximizes $\log A(y^2 \mid y^1, q + p)$, where $y^1$ is the upstream model's suffix sentence, $y^2$ is the corrected suffix, $q$ the query, and $p$ the accumulated prefix. At inference the same small model $A$ is placed in a generation loop: the upstream model generates a sentence, $A$ copies it if it is good or rewrites it if it is bad (rewrite the bad, improve the neutral, keep the good), and the rewritten sentence becomes part of the prefix for the next step. This turns a single full-answer correction into many small distribution shifts, letting the upstream model contribute more of its own latent knowledge while the small model only nudges each step.
What would settle it
Run the same generation loop with a corrector that is trained only to copy its input sentence; if helpfulness or math accuracy still improves over rounds, the improvement is not caused by the learned correction. Separately, re-score a random sample of 200 responses with human raters instead of GPT-4 to check whether the reported win-rate ordering survives.
Extended reading notes
Core claim
The central claim is that alignment can be performed online, one sentence at a time, by training a small model to replicate the residuals between an original suffix sentence and a human-preferred correction, conditioned on the query and the current answer prefix. During inference, the small model corrects each suffix produced by the upstream model, and the corrected suffix is appended to the prefix for the next generation step. The paper reports that Stream Aligner-2B raises the helpfulness win rate of Llama2-70B-chat to 76.1% and harmlessness to 36.0% at the best correction round, and Stream Aligner-8B adds 3.5% accuracy on MATH for Llama3-70B-Instruct, with gains that grow as correction rounds increase and then plateau.
Load-bearing premise
The claim assumes that each sentence-level correction makes the response better and that feeding corrected sentences back into the prefix does not compound errors over many rounds; if either fails, the reported gains would not be monotonic or stable.
Editorial extensions
If this is right
- A 2B-parameter corrector can improve helpfulness and harmlessness of a 70B chat model, so alignment need not scale with the base model.
- On math tasks, accuracy rises roughly monotonically with the number of correction rounds and converges, meaning longer interventions can be budgeted by allowed latency.
- The per-token inference time is claimed to be about $0.80\times$ that of Aligner under the same upstream model and equal corrector size, and first-token latency is about $10\times$ lower.
- The pipeline is model-agnostic: the same trained corrector can be plugged into different upstream models without access to their logits or weights.
- Compared with SFT and DPO on Llama3-8B-Instruct, Stream Aligner reports a larger accuracy improvement ($+5.8\%$ vs. $-0.5\%$ and $+0.3\%$).
Reading between the lines
- If sentence-level residuals are the right granularity, the same corrector should transfer across upstream models with no retraining beyond the initial preference data; the paper's reported $0.5B$ and $2B$ correctors on $8B$ and $70B$ upstream models are consistent with this but do not directly test cross-model transfer.
- The non-monotonic helpfulness win rate (peak at round 3, decline to $9.9\%$ at round 10) suggests the loop trades conciseness for safety; a length-penalty or early-stopping policy could be a simple extension.
- Because GPT-4 both produces the training corrections and judges the win rates, a human preference replication study would determine how much of the reported gain is an artifact of the evaluator.
- The continue-generation ablation, where the corrector finishes the answer and then corrects once, performs competitively on math with fewer rounds; extending this variant to longer reasoning chains may beat the per-sentence loop at lower latency.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes Stream Aligner, a sentence-level correction paradigm in which a small model A is trained to rewrite the last sentence of a response given the query and the current prefix, and at inference time the correction is appended to the prefix before the upstream model continues generating. This loop is repeated until the response ends. The paper claims that this dynamic correction improves helpfulness and harmlessness of Llama2-70B-chat and improves math reasoning of Llama3-70B-Instruct, while reducing latency and reducing dependence on the size of the additional model compared with Aligner.
Significance. If the claims were robust, the contribution would be practically significant: a 2B-parameter corrector improving a 70B model at inference time with lower latency than Aligner would be an attractive plug-and-play alignment method. The paper has commendable elements: it releases code, reports ablations across corrector sizes and generation pipelines, and includes an interpretability analysis with representation vectors. However, the central empirical claim is not currently supported. The headline numbers are internally inconsistent, the main table shows the helpfulness win rate collapsing after a few rounds, the evaluation has a circularity risk because GPT-4 serves as both annotator and judge, and no error bars or statistical tests are provided. These issues are load-bearing because the proposed method has no stopping rule and the claimed benefits depend on the iterative correction loop being stable and beneficial.
major comments (5)
- [Abstract, Section 3, Table 1] The headline results are inconsistent across the paper. The abstract reports a maximum helpfulness improvement of 41.2%, while the results section (Section 3) reports a maximum helpfulness win rate of 76.1%, and Table 1 shows 76.1% at round 3 and 41.2% at round 7. The paper never states which round corresponds to the headline number or why the abstract and conclusion give different values. This internal inconsistency prevents the reader from knowing which result is being claimed.
- [Table 1 and Algorithm 1] The round-0 baseline is implausible. In Table 1, the "Direct" helpfulness row reports 46.9% at round 0, but round 0 should be the uncorrected upstream output, which by definition has a win rate of 0 against itself. The "Continue" row correctly shows 0.0 at round 0. This suggests either the round labels are shifted, the evaluation protocol differs between rows, or the baseline is computed differently than stated. Additionally, the direct-pipeline helpfulness win rate rises to 76.1% at round 3 and then collapses to 9.9% at round 10. Since Algorithm 1 has no stopping rule, a deployed system cannot know to stop at round 3, and the same algorithm can produce a much worse answer after more rounds. This directly contradicts the claim that the pipeline improves helpfulness and that performance converges to a stable value.
- [Section 2, Eq. (3), Algorithm 1] There is a training/inference distribution shift for the corrector. Equation (3) trains the Stream Aligner A on sentence-level preference data in which the prefix p is an original prefix from the dataset. At inference, however, p is the concatenation of A's own previous corrections. The paper provides no analysis or experiment showing that these corrected prefixes remain in the training distribution, so errors from earlier corrections can compound. The non-monotonic behavior in Table 1 is consistent with such compounding, and the central claim of the paper requires this issue to be resolved.
- [Section 3, Evaluation Metrics; Appendix A] The evaluation for helpful and harmless QA is circular to a significant degree. GPT-4 is used both as the annotator to create the training corrections and as the evaluator that judges the final answers. Thus the reported win rates may reflect the corrector's ability to imitate GPT-4's preferences rather than an improvement in human-aligned behavior. Independent human evaluation, or at least evaluation by a different model with validation against human judgments, is needed to support the alignment claim.
- [Section 3, Math Task] The math result is too thin to support the paper's central claim. On Llama3-70B-Instruct, the reported improvement is 3.5% (or up to 4.1% with a 70B corrector), and no error bars, number of test samples, or significance tests are given. The text says the math win rate "monotonically increases" but the supporting figure is not included in the paper body, and the paper does not provide a per-round table for math. Without error bars and a clear description of the evaluation protocol, the 3.5% gain cannot be distinguished from noise.
minor comments (5)
- [Algorithm 1] The stopping condition contains a syntax error: "if y2_i = ∅ or |p_i ≥ max length|" mixes an assignment and an incomplete comparison. The intended condition should be written as "if y2_i = ∅ or |p_i| ≥ max length".
- [Figure 3 caption] The caption states that performance "eventually converging to a stable value," while the main text says that helpfulness begins to decline after a certain number of rounds. These statements should be reconciled.
- [Section 3, Results] The phrase "maximum win rate of over 76.1% in helpfulness, 36.0% in harmlessness, and 19.0% in math tasks" is confusing because the math result is later reported as 3.5% accuracy improvement; the distinction between win rate and accuracy improvement should be stated explicitly.
- [Table 1] Table 1 lacks any indication of the number of evaluation samples, variance, or statistical significance. Adding confidence intervals or at least sample sizes would materially improve interpretability.
- [Throughout] There are numerous typos and formatting issues, including "theStream Aligner," "repsonse," and missing spaces around citations. A careful proofread is needed.
Circularity Check
Helpful/harmless gains are measured by the same GPT-4 model that produced the training corrections, and the headline round is selected from a curve that collapses to 9.9%.
-
self definitional
[Section 3, 'Experiment Setup' (Dataset) and 'Evaluation Metrics']
"we take GPT-4 (Achiam et al. 2023), Llama3-70B-Instruct, Qwen1.5-110B-Chat (Team 2024) as annotators to refine the suffix sentences after generation. ... We then utilize strong models (in our case, GPT-4 for helpful & harmless QA and math) to evaluate the two answers and took the win rate of Stream Aligner against the upstream model as the evaluation result."
The training target y2 is a GPT-4 rewrite of the upstream suffix, and the reported helpful/harmless win rate is GPT-4 comparing the upstream answer with an answer assembled from Stream Aligner's GPT-4-derived corrections. The 'preferred' correction used as the training signal and the 'better' response chosen by the evaluator are therefore the same model's judgments. The measured improvement is largely GPT-4 self-consistency - how often the evaluator prefers its own editing style - rather than an independent human-preference measurement. The math result, being scored against external ground truth, is not subject to this same overlap.
-
fitted input called prediction
[Section 3 'Experiment Results', Figure 3 caption and Table 1; Abstract; Conclusion]
"It is demonstrated that Stream Aligner achieves excellent performances across all evaluation metrics for every task. Furthermore, the overall performance of Stream Aligner tends to increase with the number of correction rounds, eventually converging to a stable value."
Table 1's own helpfulness numbers peak at 76.1% at round 3 and fall to 9.9% at round 10, while the introduction and conclusion cite 41.2% (the round-7 value) and the abstract reports 76.1%. Algorithm 1 specifies no stopping rule, so the reported improvement is obtained by selecting the best round from the evaluation curve. The 'prediction' is therefore the maximum of the method's own test-set curve over an unstated round-count hyperparameter, not the output of a fixed procedure, making the headline gain a fitted statistic rather than a robust prediction.
full rationale
The central empirical claim for helpful and harmless QA is partially circular: the same GPT-4 model that wrote the sentence-level correction labels is used to judge whether the assembled corrected answer is better than the upstream answer. Because the corrected answer is composed of GPT-4-style rewrites, the win rate largely measures GPT-4's agreement with its own preferences, not an independent evaluation of alignment. This affects the headline 76.1% helpfulness / 36.0% harmlessness figure. The math result is less circular because it is checked against external ground truth, but the reported 3.5% improvement is small. Separately, the paper's round-selection practice is a fitted-input issue: Table 1 shows the helpfulness win rate collapsing from 76.1% at round 3 to 9.9% at round 10, the abstract and introduction report inconsistent headline numbers (76.1% vs 41.2%), and Algorithm 1 has no stopping rule, so the 'maximum improvement' is selected post hoc from the test curve. These robustness problems are not the same as definitional circularity, but they compound the evaluator-overlap concern. No load-bearing uniqueness theorem or self-citation chain is invoked, and the method itself is a straightforward extension of the authors' Aligner framework. The score of 6 reflects that the central QA claim reduces in part to a GPT-4 self-consistency check plus a peak-round selection, while the math evaluation retains independent content.
Assumptions & free parameters
assumptions (3)
- domain assumption GPT-4 annotations approximate human preferences for helpfulness and harmlessness.
- ad hoc to paper Sentence-level corrections do not accumulate errors across generation rounds.
- ad hoc to paper The upstream model's latent knowledge is elicited by the correction pipeline.
Cite this review
Pith. "Pith review of Stream Aligner: Efficient Sentence-Level Alignment via Distribution Induction." pith.science (2026). https://pith.science/paper/JFJK5Q3M
@misc{pith2026250105336,
author = {Pith},
title = {Pith review of: Stream Aligner: Efficient Sentence-Level Alignment via Distribution Induction},
year = {2026},
howpublished = {\url{https://pith.science/paper/JFJK5Q3M}},
note = {Machine review of arXiv:2501.05336}
}
read the original abstract
The rapid advancement of large language models (LLMs) has led to significant improvements in their capabilities, but also to increased concerns about their alignment with human values and intentions. Current alignment strategies, including adaptive training and inference-time methods, have demonstrated potential in this area. However, these approaches still struggle to balance deployment complexity and capability across various tasks and difficulties. In this work, we introduce the Streaming Distribution Induce Aligner (Stream Aligner), a novel alignment paradigm that combines efficiency with enhanced performance in various tasks throughout the generation process. Stream Aligner achieves dynamic sentence-level correction by using a small model to learn the preferences of the suffix sentence, iteratively correcting the suffix sentence output by the upstream model, and then using the corrected sentence to replace the suffix sentence in subsequent generations. Compared to Aligner, our experiments demonstrate that Stream Aligner reduces reliance on the capabilities of additional models, enhances the reasoning abilities of LLMs, and decreases latency during user interaction. Specifically, Stream Aligner-2B model has achieved an improvement of 76.1% in helpfulness, 36.0% in harmlessness on the tested Llama2-70B-chat model, and Stream Aligner-8B has achieved an improvement of 3.5% on the math ability of the tested Llama3-70B-Instruct model.
Figures
Figures from the paper (5 more)
Forward citations
Cited by 1 Pith paper
-
On Almost Surely Safe Alignment of Large Language Models at Inference-Time
An inference-time beam-search method with a safety-state tracker and latent critic enforces a user-supplied safety cost model, with an almost-sure guarantee only relative to that model.
Reference graph
Works this paper leans on
-
[1]
L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al
Achiam, J.; Adler, S.; Agarwal, S.; Ahmad, L.; Akkaya, I.; Aleman, F. L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774
arXiv 2023
-
[2]
Bai, Y.; Jones, A.; Ndousse, K.; Askell, A.; Chen, A.; DasSarma, N.; Drain, D.; Fort, S.; Ganguli, D.; Henighan, T.; et al. 2022 a . Training a helpful and harmless assistant with reinforcement learning from human feedback. arXiv preprint arXiv:2204.05862
arXiv 2022
-
[3]
Bai, Y.; Kadavath, S.; Kundu, S.; Askell, A.; Kernion, J.; Jones, A.; Chen, A.; Goldie, A.; Mirhoseini, A.; McKinnon, C.; et al. 2022 b . Constitutional ai: Harmlessness from ai feedback. arXiv preprint arXiv:2212.08073
arXiv 2022
-
[4]
K.; Scheurer, J.; Rando, J.; Freedman, R.; Korbak, T.; Lindner, D.; Freire, P.; et al
Casper, S.; Davies, X.; Shi, C.; Gilbert, T. K.; Scheurer, J.; Rando, J.; Freedman, R.; Korbak, T.; Lindner, D.; Freire, P.; et al. 2023. Open problems and fundamental limitations of reinforcement learning from human feedback. arXiv preprint arXiv:2307.15217
arXiv 2023
-
[5]
Chen, C.; Borgeaud, S.; Irving, G.; Lespiau, J.-B.; Sifre, L.; and Jumper, J. 2023. Accelerating large language model decoding with speculative sampling. arXiv preprint arXiv:2302.01318
arXiv 2023
-
[6]
Christiano, P.; Xu, M.; and Cotra, A. 2021. Arc’s first technical report: Eliciting latent knowledge. In AI Alignment Forum
work page 2021
-
[7]
Dai, J.; Pan, X.; Sun, R.; Ji, J.; Xu, X.; Liu, M.; Wang, Y.; and Yang, Y. 2023. Safe rlhf: Safe reinforcement learning from human feedback. arXiv preprint arXiv:2310.12773
arXiv 2023
-
[8]
Dathathri, S.; Madotto, A.; Lan, J.; Hung, J.; Frank, E.; Molino, P.; Yosinski, J.; and Liu, R. 2019. Plug and play language models: A simple approach to controlled text generation. arXiv preprint arXiv:1912.02164
arXiv 2019
Show all 38 references
-
[9]
Ding, N.; Qin, Y.; Yang, G.; Wei, F.; Yang, Z.; Su, Y.; Hu, S.; Chen, Y.; Chan, C.-M.; Chen, W.; et al. 2023. Parameter-efficient fine-tuning of large-scale pre-trained language models. Nature Machine Intelligence, 5(3): 220--235
2023
-
[10]
Hendrycks, D.; Burns, C.; Kadavath, S.; Arora, A.; Basart, S.; Tang, E.; Song, D.; and Steinhardt, J. 2021. Measuring mathematical problem solving with the math dataset. arXiv preprint arXiv:2103.03874
2021 arXiv
-
[11]
Hoffmann, J.; Borgeaud, S.; Mensch, A.; Buchatskaya, E.; Cai, T.; Rutherford, E.; Casas, D. d. L.; Hendricks, L. A.; Welbl, J.; Clark, A.; et al. 2022. Training compute-optimal large language models. arXiv preprint arXiv:2203.15556
2022 arXiv
-
[12]
M.; Maxwell, T.; Cheng, N.; et al
Hubinger, E.; Denison, C.; Mu, J.; Lambert, M.; Tong, M.; MacDiarmid, M.; Lanham, T.; Ziegler, D. M.; Maxwell, T.; Cheng, N.; et al. 2024. Sleeper agents: Training deceptive llms that persist through safety training. arXiv preprint arXiv:2401.05566
2024 arXiv
-
[13]
Ji, J.; Chen, B.; Lou, H.; Hong, D.; Zhang, B.; Pan, X.; Dai, J.; and Yang, Y. 2024 a . Aligner: Achieving efficient alignment through weak-to-strong correction. arXiv preprint arXiv:2402.02416
2024 arXiv
-
[14]
Ji, J.; Hong, D.; Zhang, B.; Chen, B.; Dai, J.; Zheng, B.; Qiu, T.; Li, B.; and Yang, Y. 2024 b . Pku-saferlhf: Towards multi-level safety alignment for llms with human preference. arXiv preprint arXiv:2406.15513
2024 arXiv
-
[15]
Ji, J.; Liu, M.; Dai, J.; Pan, X.; Zhang, C.; Bian, C.; Chen, B.; Sun, R.; Wang, Y.; and Yang, Y. 2024 c . Beavertails: Towards improved safety alignment of llm via a human-preference dataset. Advances in Neural Information Processing Systems, 36
2024
-
[16]
Ji, J.; Qiu, T.; Chen, B.; Zhang, B.; Lou, H.; Wang, K.; Duan, Y.; He, Z.; Zhou, J.; Zhang, Z.; et al. 2023. Ai alignment: A comprehensive survey. arXiv preprint arXiv:2310.19852
2023 arXiv
-
[17]
Ji, J.; Wang, K.; Qiu, T.; Chen, B.; Zhou, J.; Li, C.; Lou, H.; and Yang, Y. 2024 d . Language Models Resist Alignment. arXiv:2406.06144
2024
-
[18]
Ji, J.; Zhou, J.; Lou, H.; Chen, B.; Hong, D.; Wang, X.; Chen, W.; Wang, K.; Pan, R.; Li, J.; et al. 2024 e . Align Anything: Training All-Modality Models to Follow Instructions with Language Feedback. arXiv preprint arXiv:2412.15838
2024 arXiv
-
[19]
Li, K.; Patel, O.; Vi \'e gas, F.; Pfister, H.; and Wattenberg, M. 2024. Inference-time intervention: Eliciting truthful answers from a language model. Advances in Neural Information Processing Systems, 36
2024
-
[20]
Li, Y.; Wei, F.; Zhao, J.; Zhang, C.; and Zhang, H. 2023. Rain: Your language models can align themselves without finetuning. arXiv preprint arXiv:2309.07124
2023 arXiv
-
[21]
Liu, A.; Han, X.; Wang, Y.; Tsvetkov, Y.; Choi, Y.; and Smith, N. A. 2024. Tuning language models by proxy. arXiv preprint arXiv:2401.08565
2024 arXiv
-
[22]
Lu, X.; Brahman, F.; West, P.; Jung, J.; Chandu, K.; Ravichander, A.; Ammanabrolu, P.; Jiang, L.; Ramnath, S.; Dziri, N.; et al. 2023. Inference-time policy adapters (ipa): Tailoring extreme-scale lms without fine-tuning. In Proceedings of the 2023 Conference on Empirical Meth...
2023
-
[23]
Meta, A. 2024. Introducing meta llama 3: The most capable openly available llm to date. Meta AI
2024
-
[24]
Ouyang, L.; Wu, J.; Jiang, X.; Almeida, D.; Wainwright, C.; Mishkin, P.; Zhang, C.; Agarwal, S.; Slama, K.; Ray, A.; et al. 2022. Training language models to follow instructions with human feedback. Advances in neural information processing systems, 35: 27730--27744
2022
-
[25]
Paul, C.; Ajeya, C.; and Xu, M. 2024. Eliciting latent knowledge: How to tell if your eyes deceive you. Google Docs
2024
-
[26]
D.; Ermon, S.; and Finn, C
Rafailov, R.; Sharma, A.; Mitchell, E.; Manning, C. D.; Ermon, S.; and Finn, C. 2024. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36
2024
-
[27]
Taori, R.; Gulrajani, I.; Zhang, T.; Dubois, Y.; Li, X.; Guestrin, C.; Liang, P.; and Hashimoto, T. B. 2023. Alpaca: A strong, replicable instruction-following model. Stanford Center for Research on Foundation Models, 3(6): 7
2023
-
[28]
S.; Love, J.; et al
Team, G.; Mesnard, T.; Hardin, C.; Dadashi, R.; Bhupatiraju, S.; Pathak, S.; Sifre, L.; Rivi \`e re, M.; Kale, M. S.; Love, J.; et al. 2024. Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295
2024 arXiv
-
[29]
Team, Q. 2024. Introducing Qwen1.5
2024
-
[30]
Touvron, H.; Lavril, T.; Izacard, G.; Martinet, X.; Lachaux, M.-A.; Lacroix, T.; Rozi \`e re, B.; Goyal, N.; Hambro, E.; Azhar, F.; et al. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971
2023 arXiv
-
[31]
Turner, A.; Thiergart, L.; Udell, D.; Leech, G.; Mini, U.; and MacDiarmid, M. 2023. Activation addition: Steering language models without optimization. arXiv preprint arXiv:2308.10248
2023 arXiv
-
[32]
Welleck, S.; Lu, X.; West, P.; Brahman, F.; Shen, T.; Khashabi, D.; and Choi, Y. 2022. Generating sequences by learning to self-correct. arXiv preprint arXiv:2211.00053
2022 arXiv
-
[33]
Y.; and Poovendran, R
Xu, Z.; Jiang, F.; Niu, L.; Jia, J.; Lin, B. Y.; and Poovendran, R. 2024. SafeDecoding: Defending against Jailbreak Attacks via Safety-Aware Decoding. arXiv preprint arXiv:2402.08983
2024 arXiv
-
[34]
Yang, K.; and Klein, D. 2021. FUDGE: Controlled text generation with future discriminators. arXiv preprint arXiv:2104.05218
2021 arXiv
-
[35]
Yang, K.; Liu, Z.; Xie, Q.; Zhang, T.; Song, N.; Huang, J.; Kuang, Z.; and Ananiadou, S. 2024. MetaAligner: Conditional Weak-to-Strong Correction for Generalizable Multi-Objective Alignment of Language Models. arXiv preprint arXiv:2403.17141
2024 arXiv
-
[36]
Zou, A.; Phan, L.; Chen, S.; Campbell, J.; Guo, P.; Ren, R.; Pan, A.; Yin, X.; Mazeika, M.; Dombrowski, A.-K.; et al. 2023. Representation engineering: A top-down approach to ai transparency. arXiv preprint arXiv:2310.01405
2023 arXiv
-
[37]
, " * write output.state after.block = add.period write newline
ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...
-
[38]
write newline
" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...
Reviewed August 10, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.