Pith. sign in

REVIEW 5 major objections 5 minor 33 references

Precise Length Control in Large Language Models

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Decoder-only LLMs can be given precise, token-level response-length control by adding a reversed positional encoding that counts down to the target end; low-rank fine-tuning yields mean token errors under 3 tokens.

desk verdict Solid, honest adaptation of LDPE to decoder-only LLMs with clear length-control gains, but the evaluation is thin and the countdown mechanism is not isolated from alternative positional signals. read the letter →

arxiv 2412.11937 v1 pith:ANGKSFLY submitted 2024-12-16 cs.CL

Bradley Butcher , Michael O'Keefe , James Titchener This is my paper · ORCID
classification cs.CL
keywords lengthcontrolpositionalencodingdecoder-onlylanguagemodelstokenbudgetlow-rankadaptationsummarizationquestionansweringMaxNewTokens++
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that decoder-only large language models can be given precise, token-level control over response length without hurting quality. The mechanism is a second positional encoding, added to the token embeddings, that counts down from the desired response length; after a one-epoch low-rank fine-tune, the model learns to end its response when the countdown reaches zero. On question answering and summarization, the authors report mean token errors below 3 tokens, with summary quality scores comparable to a prompt-fine-tuned baseline. The practical value is that production systems needing answers of a specified length could use fine-tuning rather than prompt engineering or hard truncation.

What carries the argument

The length-difference positional encoding (LDPE): a sinusoidal positional encoding, $PE(i,2k)=\sin(i/10000^{2k/d})$ and $PE(i,2k+1)=\cos(i/10000^{2k/d})$, added to the token embeddings with the index reversed so that $i_{\mathrm{LDPE}}=(L+1)-i$ gives each token a countdown of remaining tokens to the end. The added signal is rescaled by the ratio of the Frobenius norms of the token embeddings and the positional encodings, so it does not drown out the learned embeddings, and a low-rank fine-tune teaches the model to terminate at the countdown's end. The offset variant (ORPE) applies the countdown only to the response portion; Max New Tokens++ adds a random half-normal shift to the countdown during training so the model can stop before an upper bound.

What would settle it

Evaluate an LDPE fine-tuned model on a large held-out prompt set at target lengths of 5, 300, and 1000 tokens, and compare the mean absolute token error against lengths inside the 10-200 training range; if the error grows steeply outside the range, the model is memorizing training-length associations rather than using the countdown as a general remaining-token budget.

Watch

Extended reading notes

Core claim

Adding a second, reversed sinusoidal positional encoding to input token embeddings, with an optional offset so the countdown begins only at the response, teaches a decoder-only LLM during low-rank fine-tuning to associate the remaining encoding index with the number of tokens left to generate. At inference, setting the encoding to a desired total length yields responses that terminate near that length; the paper reports mean absolute token errors of 2.8 and 2.4 tokens on summarization for the two tested 7-8B parameter instruction-tuned models, and near-perfect length alignment on question-answering targets from 10 to 200 tokens, while BERT and ROUGE scores match a prompt-fine-tuned baseline.

Load-bearing premise

The reverse positional encoding is learned as a general 'remaining token budget' that transfers from the 110,000-sample fine-tuning set to unseen prompts and to target lengths across the 10-200 token range, even though the evaluation used 20 hand-selected question-answering prompts and no held-out model-selection split.

Editorial extensions

If this is right

  • Question-answering responses can be generated at any user-chosen length between 10 and 200 tokens with near-perfect target alignment.
  • Summarization reaches a mean absolute length error of 2.4-2.8 tokens, versus roughly 24.8 tokens for a prompt-based length-control baseline, at comparable BERT and ROUGE quality.
  • The method works on existing decoder-only models without changing the pretrained rotary positional embeddings, so it can be added as a low-rank fine-tune rather than a full retraining.
  • Max New Tokens++ extends exact-length control to upper-bound control, allowing the model to end before the limit when its response is naturally complete.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the remaining-budget interpretation is correct, the same countdown should transfer to word- or character-level targets by rescaling the index; the paper itself lists this as future work.
  • The conditioning signal is not inherently tied to the end-of-sequence token, so the same mechanism might steer other discrete choices, such as how many facts to state or where to place a citation.
  • A stronger test would evaluate out-of-range lengths, such as 5, 300, or 1000 tokens, on a large held-out prompt set; the paper's current evidence does not fully isolate the countdown from memorized length associations.
  • Replacing the sinusoidal countdown with a learned scalar 'remaining budget' embedding could preserve control while avoiding interference with existing positional encodings, and would be a direct way to test whether the countdown or the fine-tune itself carries the effect.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes a method for precise token-level length control in decoder-only LLMs. It adapts a reverse sinusoidal positional encoding, called LDPE (and an offset variant ORPE), adds it to the input token embeddings, and fine-tunes the model with LoRA. At inference, the encoding counts down from a user-specified target length, and the model is expected to terminate at the set length. The authors report mean absolute length errors below three tokens on a summarization task, qualitative evidence on 20 question-answering prompts, a Max New Tokens++ extension for upper-bound control, and benchmark results indicating preserved general quality. The paper is clearly written and the method is simple, but the empirical evidence is thin in several respects, and a missing control leaves the causal role of the countdown direction unestablished.

Significance. If the central claim is correct, the method provides a lightweight, LoRA-based way to give decoder-only LLMs explicit token-budget awareness, which is practically valuable for production systems requiring structured or length-constrained outputs. The approach builds on existing encoder-decoder length-control ideas (Takase and Okazaki, 2019) and adapts them to modern decoder-only models, and the Max New Tokens++ extension is a plausible way to relax exact-length targets. The paper uses standard open models (Mistral-7B, Llama3-8B) and standard datasets (OpenOrca, MMLU, CNN/DailyMail), which is a strength for reproducibility, and the limitations section is candid about several remaining gaps. However, the evidence base is too narrow to fully support the claims: the control condition problem, the tuning-on-evaluation issue, and the small hand-selected QA set all bear directly on the paper's central assertions.

major comments (5)
  1. [Section 3.1 and Section 4] The claim that the reversed countdown semantics of LDPE/ORPE are what enables length control is not isolated by the experiments. The only length-control baseline is prompt-based fine-tuning, which differs from LDPE in multiple ways: it adds no per-token encoding, the target length is textual, and the model receives the target only once. The observed low errors could therefore be due to the added per-token signal generally (e.g., any consistent positional feature), to a length prior learned from the 110,000-sample training set, or to the countdown specifically. A control condition using the same added-embedding procedure but with forward sinusoidal encodings, or with fixed random per-token encodings, under identical fine-tuning is necessary to attribute the effect to the countdown direction. Without such a control, the causal role of LDPE is not independently established.
  2. [Section 4, Hyperparameters] The manuscript states that hyperparameters were "obtained via manual experimentation and evaluation until satisfactory results were obtained." No separate held-out model-selection split is described; the only evaluation set consists of 200 samples drawn in the same manner as the training set, and the QA length-control results use 20 hand-selected prompts. This means the reported numerical results are likely optimistically biased, because the same evaluation data may have been used to guide tuning. The paper should report results on a genuinely held-out test split, describe the tuning procedure, and provide error bars (e.g., bootstrap confidence intervals or multiple training seeds) for the mean token errors.
  3. [Section 5.1.1] The question-answering length-control evaluation is based on only 20 hand-selected samples, and the results are presented solely as scatter plots. No quantitative summary (mean or median absolute error, standard deviation, or per-target-length breakdown) is given, and the selection criterion for the 20 prompts is not stated. The visual claim of "near-perfect alignment" is not sufficient to support the abstract's "mean token errors of less than 3 tokens," which is actually only supported by the summarization experiment. The authors should provide numeric length-error statistics for both models and for ORPE, along with confidence intervals and a description of how the 20 prompts were chosen.
  4. [Section 5.1.2] The response-quality claim "without compromising response quality" is primarily evaluated by comparing generated summaries to GPT-3.5-turbo-generated summaries using BERT and ROUGE scores. This is an automatic proxy, not a human judgment, and the target summaries themselves are generated by another LLM, so the comparison measures similarity to a particular model's style and content rather than absolute quality. The standard benchmarks in Section 5.1.3 partially mitigate this, but one benchmark (Mistral on HellaSwag) shows a reduction, and the evaluation harness was modified for QA tasks. A human evaluation of summary faithfulness, or at least evaluation against a dataset with human-written reference summaries, would substantially strengthen the quality-preservation claim.
  5. [Section 5.1.3] The description of the LM Evaluation Harness modification is insufficiently precise. The text says the harness was "modified to optionally apply the LDPE countdown to the queries" for QA tasks, and that this gave the model "additional information about the response length, but no information about whether a specific question response pair was correct or incorrect." It is unclear whether the base (non-fine-tuned) models also received these added LDPE encodings during evaluation. If they did, adding untrained encodings could distort their log-likelihoods; if they did not, the comparison is not controlled. The exact evaluation protocol for each model and task should be specified in detail.
minor comments (5)
  1. [Abstract and Section 5.1.1] The abstract's claim of "mean token errors of less than 3 tokens" is presented without specifying that it comes from the summarization experiment; the QA experiment does not report a mean error. The scope of this claim should be stated explicitly.
  2. [Section 5.2] The Max New Tokens++ results are shown only as a plot with no numeric metrics. The statement that the 95% confidence interval "still follows closely" should be defined (e.g., bootstrap over samples) and quantified with a table of, for example, median absolute error and coverage of the confidence interval.
  3. [Appendix A, Table 1] The table caption says "ORPE fine tuned Llama3 8B" while Section 5.1.1 reports the LDPE results in the main text; please verify that the model and encoding names are consistent throughout.
  4. [Section 4, Data] The OpenOrca reference contains a malformed URL ("https://https://huggingface.co/..."). Please correct the citation.
  5. [Section 5.1.2] There are minor grammatical issues, for example "for the Mistral-prompted was an order of magnitude larger" and "the target length as was added to the prompt." A copy-editing pass is recommended.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the length-control result is an empirical fine-tuning measurement, not a claim derived from its own inputs.

full rationale

LDPE is an input-side conditioning mechanism: the reverse sinusoidal encoding of Eqs. (1)-(2) with i_LDPE=(L+1)-i explicitly injects the target total length into the token embeddings, and the fine-tuning loss in Eq. (4) is standard next-token prediction on responses of known length. The paper's central claim is an empirical measurement—fine-tuned models stop near the target count—not a derivation from an assumption that already contains the result. No equation reduces to the claimed outcome by construction; the model could in principle ignore the added encoding, and the prompt-tuned baseline shows that simply knowing the target as text gives much larger errors (mean 24.8 tokens vs. 2.8 for Mistral-LDPE). The method adapts LDPE from Takase and Okazaki [27], an external source, and no load-bearing self-citation or imported uniqueness theorem is used. The admission that hyperparameters were tuned until satisfactory results (Section 4) is a validation and selection-bias concern, not a circular step, because the reported errors are measured outcomes rather than parameter-free forecasts. Consequently, no circular step meets the quoted-evidence bar.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method is empirical and imports the reverse positional encoding idea from prior work. The central claim rests on assumptions that the added encoding is compatible with RoPE-based pretrained models, that Frobenius-norm scaling is a sufficient way to merge the signals, and that QA fine-tuning transfers to summarization. The free parameters are mostly standard training hyperparameters; the Max New Tokens++ schedule adds two hand-picked scales.

free parameters (5)
  • Max New Tokens++ sigma_0 = 0.1
    Initial scale of half-normal shift in Section 5.2; chosen by hand, controls how early the model may terminate.
  • Max New Tokens++ sigma_max = 2048
    Maximum scale of half-normal shift; chosen by hand for length variability.
  • LoRA rank = 16
    LoRA rank used for all fine-tuning; selected by manual experimentation.
  • LoRA alpha = 32
    LoRA scaling; selected by manual experimentation.
  • Learning rate = 0.0003
    AdamW learning rate; hyperparameters obtained via manual experimentation until satisfactory results.
assumptions (4)
  • domain assumption Sinusoidal positional encodings in reversed order provide a usable countdown signal to a transformer pretrained with rotary embeddings.
    Section 3.1 assumes adding absolute LDPE to token embeddings will not interfere destructively with RoPE; supported only by empirical results.
  • domain assumption The Frobenius-norm scaling makes the reverse positional embeddings comparable in magnitude to token embeddings.
    Section 3.2.1, Eq. 5; no theoretical justification is given for this specific scaling choice.
  • domain assumption Fine-tuning on OpenOrca/MMLU QA data teaches a general length-control behavior that transfers to summarization.
    Section 4; the transfer is assumed, and the summarization evaluation is used to test it.
  • domain assumption BERT/ROUGE scores against GPT-3.5-generated summaries are a valid proxy for response quality.
    Section 5.1.2; no human evaluation is included, so the quality claim depends on this proxy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Precise Length Control in Large Language Models." pith.science (2026). https://pith.science/paper/ANGKSFLY

@misc{pith2026241211937,
  author       = {Pith},
  title        = {Pith review of: Precise Length Control in Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ANGKSFLY}},
  note         = {Machine review of arXiv:2412.11937}
}
read the original abstract

Large Language Models (LLMs) are increasingly used in production systems, powering applications such as chatbots, summarization, and question answering. Despite their success, controlling the length of their response remains a significant challenge, particularly for tasks requiring structured outputs or specific levels of detail. In this work, we propose a method to adapt pre-trained decoder-only LLMs for precise control of response length. Our approach incorporates a secondary length-difference positional encoding (LDPE) into the input embeddings, which counts down to a user-set response termination length. Fine-tuning with LDPE allows the model to learn to terminate responses coherently at the desired length, achieving mean token errors of less than 3 tokens. We also introduce Max New Tokens++, an extension that enables flexible upper-bound length control, rather than an exact target. Experimental results on tasks such as question answering and document summarization demonstrate that our method enables precise length control without compromising response quality.

Figures

Figures reproduced from arXiv: 2412.11937 by the authors.

Figure 1
Figure 1. A simplified diagram of the reverse positional encoding methods [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Comparison of target and response lengths for different length control [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Results for length the length controlled summarisation task. Mod [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Performance of length controlled and baseline models against a range [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Plot of token limit vs response length for using a [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Comparison of target and response lengths for different length control [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 14 canonical work pages

  1. [1]

    Curriculum learning

    Yoshua Bengio, J´ erˆ ome Louradour, Ronan Collobert, and Jason Weston. Curriculum learning. In Proceedings of the 26th annual international con- ference on machine learning , pages 41–48, 2009

    work page 2009

  2. [2]

    PIQA: reasoning about physical commonsense in natural language

    Yonatan Bisk, Rowan Zellers, Ronan Le Bras, Jianfeng Gao, and Yejin Choi. PIQA: reasoning about physical commonsense in natural language. CoRR, abs/1911.11641, 2019

    arXiv 1911

  3. [3]

    Language models are few-shot learners

    Tom B Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Ka- plan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sas- try, Amanda Askell, et al. Language models are few-shot learners. arXiv preprint arXiv:2005.14165, 2020. 14

    arXiv 2005

  4. [4]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805 , 2018

    arXiv 2018

  5. [5]

    Recent automatic text summarization techniques: a survey

    Mahak Gambhir and Vishal Gupta. Recent automatic text summarization techniques: a survey. Artificial Intelligence Review , 47(1):1–66, 2017

    work page 2017

  6. [6]

    A framework for few-shot language model evaluation, 12 2023

    Leo Gao, Jonathan Tow, Baber Abbasi, Stella Biderman, Sid Black, Anthony DiPofi, Charles Foster, Laurence Golding, Jeffrey Hsu, Alain Le Noac’h, Haonan Li, Kyle McDonell, Niklas Muennighoff, Chris Ociepa, Jason Phang, Laria Reynolds, Hailey Schoelkopf, Aviya Skowron, Lintang Sutawika, Eric Tang, Anish Thite, Ben Wang, Kevin Wang, and Andy Zou. A framework...

    2023

  7. [7]

    Contextual position encoding: Learning to count what’s important

    Olga Golovneva, Tianlu Wang, Jason Weston, and Sainbayar Sukhbaatar. Contextual position encoding: Learning to count what’s important. arXiv preprint arXiv:2405.18719, 2024

    arXiv 2024

  8. [8]

    News summarization and evaluation in the era of gpt-3

    Tanya Goyal, Junyi Jessy Li, and Greg Durrett. News summarization and evaluation in the era of gpt-3. arXiv preprint arXiv:2209.12356 , 2022

    arXiv 2022

Show all 33 references
  1. [9]

    Measuring massive multitask language understanding, 2021

    Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. Measuring massive multitask language understanding, 2021

    2021

  2. [10]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models. CoRR, abs/2106.09685, 2021

    2021 arXiv

  3. [11]

    Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, L´ elio Renard Lavaud, Marie- Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thom...

    2023

  4. [12]

    Prompt-based length controlled generation with reinforcement learning

    Renlong Jie, Xiaojun Meng, Lifeng Shang, Xin Jiang, and Qun Liu. Prompt-based length controlled generation with reinforcement learning. ArXiv, abs/2308.12030, 2023

    2023 arXiv

  5. [13]

    Controlling output length in neural encoder-decoders

    Yuta Kikuchi, Graham Neubig, Ryohei Sasano, Hiroya Takamura, and Manabu Okumura. Controlling output length in neural encoder-decoders. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 1328–1338, 2016

    2016

  6. [14]

    Chao Lei, Nir Lipovetzky, and Krista A. Ehinger. Generalized planning for the abstraction and reasoning corpus, 2024

    2024

  7. [15]

    Openorca: An open dataset of gpt augmented flan reasoning traces

    Wing Lian, Bleys Goodson, Eugene Pentland, Austin Cook, Chanvichet Vong, and ”Teknium”. Openorca: An open dataset of gpt augmented flan reasoning traces. https://https://huggingface.co/Open-Orca/ OpenOrca, 2023. 15

    2023

  8. [16]

    Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing

    Pengfei Liu, Weizhe Yuan, Jinlan Fu, Zhengbao Jiang, Hiroaki Hayashi, and Graham Neubig. Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. ACM Computing Surveys, 55:1 – 35, 2021

    2021

  9. [17]

    Length control in abstractive summa- rization by pretraining information selection

    Yizhu Liu, Qi Jia, and Kenny Zhu. Length control in abstractive summa- rization by pretraining information selection. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages 6885–6895, 2022

    2022

  10. [18]

    Decoupled weight decay regularization, 2019

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization, 2019

    2019

  11. [19]

    Transformers can do arithmetic with the right embeddings

    Sean McLeish, Arpit Bansal, Alex Stein, Neel Jain, John Kirchenbauer, Brian R Bartoldson, Bhavya Kailkhura, Abhinav Bhatele, Jonas Geiping, Avi Schwarzschild, et al. Transformers can do arithmetic with the right embeddings. arXiv preprint arXiv:2405.17399 , 2024

    2024 arXiv

  12. [20]

    Introducing Meta Llama 3: The most capable openly available LLM to date — ai.meta.com

    Meta. Introducing Meta Llama 3: The most capable openly available LLM to date — ai.meta.com. https://ai.meta.com/blog/meta-llama-3/,

  13. [21]

    Abstractive text summarization using sequence-to- sequence rnns and beyond, 2016

    Ramesh Nallapati, Bowen Zhou, Cicero Nogueira dos santos, Caglar Gul- cehre, and Bing Xiang. Abstractive text summarization using sequence-to- sequence rnns and beyond, 2016

    2016

  14. [22]

    GPT3.5 Turbo-0125, 2024

    OpenAI. GPT3.5 Turbo-0125, 2024

    2024

  15. [23]

    Recipes for building an open-domain chatbot

    Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric Michael Smith, et al. Recipes for building an open-domain chatbot. arXiv preprint arXiv:2004.13637, 2021

    2004 arXiv

  16. [24]

    WINOGRANDE: an adversarial winograd schema challenge at scale

    Keisuke Sakaguchi, Ronan Le Bras, Chandra Bhagavatula, and Yejin Choi. WINOGRANDE: an adversarial winograd schema challenge at scale. CoRR, abs/1907.10641, 2019

    1907 arXiv

  17. [25]

    Self-attention with relative position representations

    Peter Shaw, Jakob Uszkoreit, and Ashish Vaswani. Self-attention with relative position representations. arXiv preprint arXiv:1803.02155 , 2018

    2018 arXiv

  18. [26]

    Roformer: Enhanced transformer with rotary position embedding

    Jianlin Su, Murtadha Ahmed, Yu Lu, Shengfeng Pan, Wen Bo, and Yunfeng Liu. Roformer: Enhanced transformer with rotary position embedding. Neurocomputing, 568:127063, 2024

    2024

  19. [27]

    Positional encoding to control output se- quence length

    Sho Takase and Naoaki Okazaki. Positional encoding to control output se- quence length. In Proceedings of the 2019 Conference of the North Amer- ican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) , pages...

    2019

  20. [28]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. In Advances in neural information processing systems , pages 5998– 6008, 2017

    2017

  21. [29]

    Huggingface’s transformers: State-of-the-art natural lan- guage processing

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, R´ emi Louf, Morgan Funtowicz, et al. Huggingface’s transformers: State-of-the-art natural lan- guage processing. arXiv preprint arXiv:1910.03771 , 2019

    1910 arXiv

  22. [30]

    Lenatten: An effective length controlling unit for text summarization

    Zhongyi Yu, Zhenghao Wu, Hao Zheng, Zhe XuanYuan, Jefferson Fong, and Weifeng Su. Lenatten: An effective length controlling unit for text summarization. CoRR, abs/2106.00316, 2021

    2021 arXiv

  23. [31]

    Hellaswag: Can a machine really finish your sentence? CoRR, abs/1905.07830, 2019

    Rowan Zellers, Ari Holtzman, Yonatan Bisk, Ali Farhadi, and Yejin Choi. Hellaswag: Can a machine really finish your sentence? CoRR, abs/1905.07830, 2019

    1905 arXiv

  24. [32]

    Latent prompt tuning for text summarization

    Yubo Zhang, Xingxing Zhang, Xun Wang, Si-Qing Chen, and Furu Wei. Latent prompt tuning for text summarization. ArXiv, abs/2211.01837, 2022. 17 Appendix A Example length controlled outputs Examples of length controlled outputs from the LDPE finetuned Llama model are shown in Ta...

    2022 arXiv

  25. [2024]

    [Accessed 13-05-2024]

    2024

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.