Look Ahead Text Understanding and LLM Stitching

CA; China); City University of Hong Kong; Hong Kong; Junlin Julian Jiang (Piedmont High School; Piedmont; USA); Xin Li (College of Business

REVIEW 4 major objections 5 minor 28 references

Look Ahead Text Understanding and LLM Stitching

T0 review · 4 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 Stitching BERT and GPT predicts the next section of developing text.

desk verdict A modest but genuinely new task framing and a sensible stitching idea; the empirical case for the headline claim is too thin to trust without error bars and a real per-model noise comparison. read the letter →

arxiv 2412.17836 v1 pith:UFPEEZYC submitted 2024-12-16 cs.CL

Junlin Julian Jiang (Piedmont High School , Piedmont , CA , USA) , Xin Li (College of Business , City University of Hong Kong , Hong Kong , China) This is my paper · ORCID

classification cs.CL

keywords lookaheadtextunderstandingsectionidentificationLASIBERTGPTmodelstitchingrepresentationmappingnoiserobustness

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

This paper introduces look-ahead section identification (LASI): predicting the section label of a sentence that has not been written yet, using only the preceding sentences. The authors argue that this task needs both BERT's bidirectional context and GPT's ability to predict what comes next, and they propose two 'stitching' methods that tie GPT's representation of the previous sentence to BERT's representation of the next sentence. On a medical-abstract dataset, the stitched models reach about 74.7-74.8% accuracy, roughly one point above BERT and GPT alone, and their advantage grows when the input text is corrupted by noise, as in generative-AI settings. The paper frames LASI as an example of a broader look-ahead text understanding family, including look-ahead sentiment and topic classification.

What carries the argument

The central identity is the stitching equation B(s_k) = f(G(s_{k-1})), which asserts that a mapping f can convert GPT's output on the previous sentence into BERT's output on the next sentence. The paper implements f in two ways: Loss Stitching (GBLS) fits a two-layer Tanh mapper by adding an MSE loss between the transformed GPT representation and the BERT representation of the upcoming sentence, and Attention Stitching (GBAS) replaces the explicit mapper with an attention operation taking G(s_{k-1}) as query, G(s_{k-2}) as key, and B(s_{k-1}) as value. Both are meant to 'drift' GPT's unidirectional read of the past into BERT's bidirectional space so that a standard classifier head can label the not-yet-written sentence.

What would settle it

Train the two stitching models on one half of PubMed 200k RCT and evaluate on the other half, and also on a different corpus (e.g., news or social media) with section-like labels. If the accuracy gain over BERT disappears out-of-corpus or shrinks to zero when the mapper is tested on sentence pairs it never saw, the stitching equation is not a general representation alignment but an artifact of the training distribution.

Watch

Extended reading notes

Core claim

LASI is a harder task than classic section identification because the sentence to be classified is missing; BERT alone loses 10-15 accuracy points compared with the classic setting. The paper's central claim is that a learnable stitching between two pre-trained models recovers part of that loss. The authors posit that GPT's hidden state on sentence s_{k-1} and BERT's hidden state on sentence s_k share task-relevant information, so a mapping B(s_k) = f(G(s_{k-1})) should exist. They implement f by a loss-stitched nonlinear mapper (GBLS) and by an attention-based mapper (GBAS). Experiments on PubMed 200k RCT show GBLS at 74.7% and GBAS at 74.8% accuracy, both above BERT, GPT, a GPT-BERT concatenation, and BART; when words are dropped or added as noise, the stitched models' lead over BERT widens.

Load-bearing premise

Everything rests on the assumption that a mapping trained on adjacent sentence pairs in one corpus will keep working on new, unseen documents — that GPT's reading of a previous sentence really can be converted into BERT's reading of the next sentence, rather than the mapper merely memorizing the training pairs.

Editorial extensions

If this is right

If the stitching equation holds generally, LASI is reducible to classic SI once the mapper is learned, so any SI classifier can be applied to developing text.
The roughly one-point accuracy gain and improved noise robustness suggest stitching is a lightweight alternative to training a full encoder-decoder like BART for look-ahead tasks.
Because the stitching framework only assumes two pre-trained models with aligned adjacent-sentence representations, it can be applied to other look-ahead text understanding tasks such as look-ahead sentiment or topic classification.
The masking step used during training indicates that the high-capacity mapper can overfit to training sentence pairs; the paper argues this is mitigated by random input masking.

Reading between the lines

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

The paper does not test whether the learned mapping transfers across domains (e.g., from biomedical abstracts to news or social media); a direct test would be to train the mapper on one corpus and evaluate on another.
The reported gains are small (~0.7-0.8 accuracy points), so the practical value hinges on whether the 10-15 point gap between LASI and classic SI can be further closed by better mappers or by combining stitching with label-sequence models such as CRF, which the paper explicitly set aside.
The stitching identity suggests a possible curriculum: use GPT to propose candidate next sentences, then stitch only the representations rather than generating text, reducing generation cost while retaining predictive information.
If noise robustness is confirmed on other datasets, stitching could serve as a defense in generative-AI writing assistants, where the text being classified is partial and error-prone.

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.

Desk Editor's Note

read the letter

Colleague,

Here's the short version: the paper introduces a genuinely new task, look-ahead section identification (LASI), and a sensible way to stitch GPT and BERT representations using adjacent sentence pairs. The task-validity part holds up. The experimental support for the headline claim does not.

What's actually new: framing SI as predicting the label of a not-yet-written sentence from prior context, and the two stitching variants, GBLS (loss-based) and GBAS (attention-based), which map GPT's representation of the previous sentence toward BERT's representation of the next. That is a clean, lightweight idea, and the paper honestly distinguishes it from LoRA, distillation, and BART.

What the paper does well: the SI-versus-LASI difficulty gap is convincing — LASI is 10–15 points harder, which makes sense. The writing is clear, the dataset is public, and the limitations section is candid. The authors also disclose the mapping error (RMSE ~0.26, about 10% of the BERT representation norm).

The soft spots are in the evidence for the main claim. Table 3 shows single-run accuracies: GBAS at 74.8 versus BERT at 74.0, a 0.8-point gap on one test split. No error bars, no multiple seeds, no significance tests. That gap could easily be run-to-run noise. More importantly, the noise-robustness claim in Experiment III is unsupported: Table 4 lists one accuracy per tweak with no per-model breakdown, so the statement about stitching and BART gaining over BERT as noise increases is not verifiable. The footnote about a GBAS variant and the 'about 1% improvement' also appear in no table. And the GitHub links are broken.

None of this is fatal to the task framing. The mapping assumption in Eq. (1) is a theoretical leap, but the paper's own numbers suggest the mechanism is not the immediate problem; the immediate problem is that the empirical differences may be noise.

This paper is for readers interested in sequential sentence classification or in lightweight ways to combine encoders and decoders. I would not cite its accuracy numbers as evidence. It does deserve a serious referee — the task is new and the idea is worth testing — but before publication the authors should provide multiple seeds, confidence intervals, and a per-model noisy-text comparison.

Recommendation: conditional acceptance at best, contingent on those experiments. If that is not feasible, it is a workshop-level contribution.

Referee Report

4 major / 5 minor

Summary. The paper introduces look-ahead section identification (LASI), a variant of section identification in which the label of the next sentence must be predicted from the preceding sentence(s) rather than from the focal sentence itself. Using PubMed-RCT, the authors first show that LASI is harder than classic SI, then compare several baselines (BERT, GPT, BERT(GPT), GPT+BERT, BART) with two proposed stitching models, GBLS and GBAS, which attempt to map GPT representations of the previous sentence into BERT representations of the next sentence via a loss-based mapper or an attention module. The paper reports that GBLS and GBAS achieve the best accuracies on a clean test split and claims that stitching is especially robust to noisy input text. The manuscript also discusses broader applications of look-ahead understanding and limitations of the study.

Significance. If the central empirical claim were established, LASI would be a useful new task for studying predictive text understanding in developing documents, and the stitching idea would be a lightweight way to combine complementary pre-trained models. The paper is clearly written and provides a public dataset, code, and a reproducible experimental protocol, which are strengths. However, the current evidence is insufficient to support the headline: the reported accuracy gaps over baselines are 0.5–0.8 percentage points from single runs, no statistical testing is performed, and the noise-robustness claim rests on a table that does not show per-model results. The mapping assumption underlying both proposed models is also not validated for generalization. These are load-bearing weaknesses that require additional experiments or analysis.

major comments (4)

[Experiment II, Table 3] The central claim that GBLS and GBAS outperform established models rests on accuracy differences of 0.5–0.8 percentage points (74.7 and 74.8 vs. 74.0 for BERT/GPT and 74.3 for BART) from what appear to be single runs. No error bars, multiple seeds, or significance tests are reported, so these differences may well be within run-to-run variation. Please report results across at least 5–10 random seeds with standard deviations and a paired significance test (e.g., McNemar's test) for the main comparisons.
[Experiment III, Table 4 and Figure 6] The noise-robustness claim is not verifiable from the reported evidence. Table 4 gives accuracies across tweaked settings but does not break down results by model, and the text states that 'the advantage of stitching and BART models over BERT increases when the sentences get noisier' without showing any per-model comparison. Please provide a table (or figure) with accuracies for each model at each noise level, along with significance tests for the interaction between model and noise.
[Methods, Eq. (1) and Experiment II] The core assumption B(s_k) = f(G(s_{k-1})) is fit to training data, but the paper provides no evidence that the learned mapping generalizes to new documents. The reported RMSE of about 0.26 (roughly 10% of the BERT representation norm) is substantial, and the paper does not analyze how this mapping error propagates to classification accuracy or whether the mapping is stable across documents. Please add an evaluation of mapping generalization, for example by training the mapper on one set of documents and testing the stitching model on held-out documents, and report the resulting classification accuracies.
[Experiment I, Table 2] The paper concludes that 'combining the last two sentences in LASI does not make much difference' and therefore uses only s_{k-1} in later experiments. However, Table 2 shows a 3.2-point accuracy gain from using s_{k-2}s_{k-1} (77.2%) over s_{k-1} alone (74.0%). This is a substantial gap, and discarding s_{k-2} may weaken the proposed models relative to what they could achieve. Please justify this design choice, or include experiments with the two-sentence input for the proposed models.

minor comments (5)

[Discussion] There is a typo in the first bullet: 'beacuse' should be 'because'.
[Experiment II, text after Table 3] The text refers to 'BERT+GPT' while the table and method description use 'GPT+BERT'; please use a consistent name.
[Experiment III, Table 4] For the noisy settings, the reported F1 scores are higher than the accuracies (e.g., 74.1% F1 vs. 72.3% accuracy for '-2 words'), which is unusual for a multi-class accuracy/F1 pair; please verify these numbers and report whether they are weight-averaged in the same way as in Table 2.
[Experiment II, footnote] The footnote describing a GBAS variant that incorporates B(s_{k-2}) and claims 'about 1% improvement' is not backed by any table or figure; either report the result explicitly or remove the claim.
[References] The reference list includes FORCE11 (2020) and Gebru et al. (2021), but these are not cited in the body of the paper; please either cite them where relevant or remove them.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation; the only self-citation is background and not load-bearing.

full rationale

The central claim is empirical: GBLS and GBAS outperform baselines on LASI. Eq. (1) is a stated modeling assumption, not a result derived from itself; the mapping f is fit to adjacent sentence pairs in training and evaluated on held-out documents, so the method is standard supervised learning rather than a prediction forced by construction. The loss-stitching objective uses B(s_k) as an auxiliary target during training, but at inference only G(s_{k-1}) (and for GBAS, G(s_{k-2}) and B(s_{k-1})) is used; no evaluation uses the target sentence representation. The paper cites the authors' earlier SI work (Zhou and Li 2020) for background and as a feature-engineering baseline, but that citation is not invoked to justify Eq. (1), the stitching architecture, or the reported gains. The unsupported noise-robustness claim in Experiment III, where Table 4 gives no per-model breakdown, is an evidence and correctness concern, not circularity. No equation reduces to its own inputs, and no fitted parameter is renamed as a prediction. The score of 2 reflects only a minor, non-load-bearing self-citation.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

LASI is a task definition, not a new physical or formal entity, and 'stitching' is a training objective rather than a new mechanism; no invented entities require independent evidence. The main ledger items are the half-dozen hyperparameters tuned on validation data and the unproven Eq. (1) mapping assumption.

free parameters (6)

MSE loss weight for GBLS = 0.05
Set after small-scale experiments; directly controls the strength of the stitching objective and therefore the reported accuracy.
BERT input masking ratio = 10%
Chosen over 0% and 20% in training for GBLS and GBAS; affects how much the model relies on memorized sentence pairs.
Mapper architecture = Tanh between two linear layers
Decided through small-scale comparison with a single linear layer; this design choice is part of the proposed GBLS method.
Number of attention heads in GBAS = 8
Set arbitrarily, not tuned; could affect stitching performance.
Learning rate and weight decay = 2e-5 and 0.01
Set without much tuning; standard choices that still affect final accuracy.
Epoch selection = Best of 10 epochs by validation accuracy
Validation-based early selection can inflate test results if validation serves as a second tuning set.

assumptions (5)

ad hoc to paper A fixed mapping f exists such that B(s_k) = f(G(s_{k-1})), and it can be learned from adjacent sentence pairs.
Eq. (1) in Methods is the starting point for both stitching designs; the paper provides intuition, not proof, that such a mapping exists.
domain assumption The section label of a sentence can be predicted from the preceding sentence(s) with a structure that is not just a positional prior.
LASI defines the classifier {s_{k-2}, s_{k-1}} -> l_k in Problem Setting; if labels were mostly determined by sentence position in the abstract, the task would not test semantic look-ahead.
domain assumption GPT-2 hidden states and BERT hidden states, despite different tokenizers and objectives, can be meaningfully compared and aligned after fine-tuning.
Eqs. (2), (3), and (5) apply MSE or attention between the two representation spaces; the paper does not report any geometric or ablation evidence that the spaces are compatible.
ad hoc to paper Randomly masking 10% of BERT input tokens during training prevents the mapper from memorizing training sentence pairs.
The Methods section introduces masking specifically to counter memorization, but no memorization test or masking ablation beyond the 0/10/20 percent comparison is given.
standard math Standard neural network machinery (Transformer attention, backpropagation, cross-entropy, XGBoost) works as expected.
Used without proof throughout the experiments; this is a reasonable background assumption.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Look Ahead Text Understanding and LLM Stitching." pith.science (2026). https://pith.science/paper/UFPEEZYC

@misc{pith2026241217836,
  author       = {Pith},
  title        = {Pith review of: Look Ahead Text Understanding and LLM Stitching},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UFPEEZYC}},
  note         = {Machine review of arXiv:2412.17836}
}

read the original abstract

This paper proposes a look ahead text understanding problem with look ahead section identification (LASI) as an example. This problem may appear in generative AI as well as human interactions, where we want to understand the direction of a developing text or conversation. We tackle the problem using transformer-based LLMs. We show that LASI is more challenging than classic section identification (SI). We argue that both bidirectional contextual information (e.g., BERT) and unidirectional predictive ability (e.g., GPT) will benefit the task. We propose two approaches to stitch together BERT and GPT. Experiments show that our approach outperforms the established models, especially when there is noise in the text (which is often the case for developing text in generative AI). Our paper sheds light on other look ahead text understanding tasks that are important to social media, such as look ahead sentiment classification, and points out the opportunities to leverage pre-trained LLMs through stitching.

Figures

Figures reproduced from arXiv: 2412.17836 by the authors.

**Figure 1.** Look Ahead Section Identification involved in reading. It has also been shown that a document’s alignment with their expected structures will increase the audience’s trust in the document (Rowley and Johnson 2013). In practice, researchers often conduct SI on finished documents. We argue that the SI task could go beyond finished documents to developing texts. For example, when AI responds to humans on social medi… view at source ↗

**Figure 2.** illustrates the process of our study, which has three sets of experiments. In the first experiment, we explore the difficulty of LASI compared to classic SI. Aligning with our problem setting elaborated above, we compare setups using different sentences to predict the label of sk. In the second experiment, we examine the performance of multiple baseline models and our proposed model in tackling LASI, which we elabo… view at source ↗

**Figure 3.** BERT and GPT Representation [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Loss Stitching While linear mapping is simple, it illustrates the rationale of stitching, which is to convert representations from one space to another. Recently, Din et al. (2023) proposed using linear mapping to replace the complicated transformer and convert betwee…

**Figure 5.** Figure 5: Attention Stitching (sk−1, sk). To prevent this, we include a masking step that randomly masks the input of BERT in training. Masking is not needed in the evaluation/prediction stage. Attention Stitching. The Loss Stitching framework is simple but differs from the popu…

**Figure 6.** Figure 6: Performance on Tweaked Settings [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 24 canonical work pages

[1]

, " * 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 mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...
[2]

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 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Brack, A.; Entrup, E.; Stamatakis, M.; Buschermöhle, P.; Hoppe, A.; and Ewerth, R. 2024. Sequential sentence classification in research papers using cross-domain multi-task learning. International Journal on Digital Libraries

work page 2024
[4]

Cohan, A.; Beltagy, I.; King, D.; Dalvi, B.; and Weld, D. 2019. Pretrained Language Models for Sequential Sentence Classification . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing , 3693--3699. Hong Kong, China: Association for Computational...

work page 2019
[5]

d.; Gojenola, K.; and Atutxa, A

de la Iglesia, I.; Vivó, M.; Chocrón, P.; Maeztu, G. d.; Gojenola, K.; and Atutxa, A. 2023. An Open Source Corpus and Automatic Tool for Section Identification in Spanish Health Records. Journal of Biomedical Informatics, 145: 104461

work page 2023
[6]

Dernoncourt, F.; and Lee, J. Y. 2017. P ub M ed 200k RCT : a Dataset for Sequential Sentence Classification in Medical Abstracts. In Kondrak, G.; and Watanabe, T., eds., Proceedings of the Eighth International Joint Conference on Natural Language Processing, 308--313. Taipei, Taiwan: Asian Federation of Natural Language Processing

work page 2017
[7]

Y.; Karidi, T.; Choshen, L.; and Geva, M

Din, A. Y.; Karidi, T.; Choshen, L.; and Geva, M. 2023. Jump to Conclusions : Short - Cutting Transformers With Linear Transformations . ArXiv:2303.09435 [cs]

arXiv 2023
[8]

FORCE11. 2020. The FAIR Data principles

work page 2020

Show all 28 references

[9]

W.; Wallach, H.; Iii, H

Gebru, T.; Morgenstern, J.; Vecchione, B.; Vaughan, J. W.; Wallach, H.; Iii, H. D.; and Crawford, K. 2021. Datasheets for datasets. , 64(12): 86–92. Communications of the ACM, 64(12): 86–92

2021
[10]

Goldfarb-Tarrant, S.; Chakrabarty, T.; Weischedel, R.; and Peng, N. 2020. Content Planning for Neural Story Generation with Aristotelian Rescoring . In Webber, B.; Cohn, T.; He, Y.; and Liu, Y., eds., Proceedings of the 2020 Conference on Empirical Methods in Natural Language ...

2020
[11]

Gonçalves, S.; Cortez, P.; and Moro, S. 2020. A deep learning classifier for sentence classification in biomedical and computer science abstracts. Neural Computing and Applications, 32(11): 6793--6807

2020
[12]

Gray, M.; Xu, J.; Tong, W.; and Wu, L. 2023. Classifying Free Texts Into Predefined Sections Using AI in Regulatory Documents : A Case Study with Drug Labeling Documents . Chemical Research in Toxicology, 36(8): 1290--1299. Publisher: American Chemical Society

2023
[13]

Hassanzadeh, H.; Groza, T.; and Hunter, J. 2014. Identifying scientific artefacts in biomedical literature: The Evidence Based Medicine use case. Journal of Biomedical Informatics, 49: 159--170

2014
[14]

J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W

Hu, E. J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2022. Lo RA : Low-Rank Adaptation of Large Language Models. In International Conference on Learning Representations

2022
[15]

Hu, Y.; Chen, Y.; and Xu, H. 2023. Towards More Generalizable and Accurate Sentence Classification in Medical Abstracts with Less Data . Journal of Healthcare Informatics Research, 7(4): 542--556

2023
[16]

Jiang, C.-Y.; and Fan, Y.-C. 2023. Biomedical Abstract Sentence Classification by BERT - Based Reading Comprehension . SN Computer Science, 4(4): 395

2023
[17]

Jin, D.; and Szolovits, P. 2018. Hierarchical Neural Networks for Sequential Sentence Classification in Medical Scientific Abstracts . In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing , 3100--3109. Brussels, Belgium: Association for Com...

2018
[18]

Koivisto, M.; and Grassini, S. 2023. Best Humans Still Outperform Artificial Intelligence in A Creative Divergent Thinking Task. Scientific Reports, 13(1): 13601. Number: 1 Publisher: Nature Publishing Group

2023
[19]

Liang, X.; Tang, Z.; Li, J.; and Zhang, M. 2023. Open-ended Long Text Generation via Masked Language Modeling . In Rogers, A.; Boyd-Graber, J.; and Okazaki, N., eds., Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics , 223--241. Toronto, C...

2023
[20]

Liu, J.; Capurro, D.; Nguyen, A.; and Verspoor, K. 2023. Attention-based Multimodal Fusion with Contrast for Robust Clinical Prediction in the Face of Missing Modalities. Journal of Biomedical Informatics, 145: 104466

2023
[21]

Ma, B.; Zhang, C.; Wang, Y.; and Deng, S. 2022. Enhancing Identification of Structure Function of Academic Articles Using Contextual Information. Scientometrics, 127(2): 885--925

2022
[22]

Minaee, S.; Kalchbrenner, N.; Cambria, E.; Nikzad, N.; Chenaghlu, M.; and Gao, J. 2021. Deep Learning --based Text Classification : A Comprehensive Review . ACM Computing Surveys, 54(3): 62:1--62:40

2021
[23]

Rowley, J.; and Johnson, F. 2013. Understanding Trust Formation in Digital Information Sources: The Case of Wikipedia . Journal of Information Science, 39(4): 494--508. Publisher: SAGE Publications Ltd

2013
[24]

N.; Kaiser, L.; and Polosukhin, I

Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, L.; and Polosukhin, I. 2017. Attention is All You Need. In Proceedings of the 31st International Conference on Neural Information Processing Systems, 6000–6010. Red Hook, NY, USA: Curran Asso...

2017
[25]

Yao, L.; Peng, N.; Weischedel, R.; Knight, K.; Zhao, D.; and Yan, R. 2019. Plan-and-Write: Towards Better Automatic Storytelling. In Proceedings of the 33rd AAAI Conference on Artificial Intelligence and 31st Innovative Applications of Artificial Intelligence Conference and 9t...

2019
[26]

Yu, G.; Zhang, Z.; Liu, H.; and Ding, L. 2019. Masked Sentence Model Based on BERT for Move Recognition in Medical Scientific Abstracts . Journal of Data and Information Science, 4(4): 42--55

2019
[27]

Zhou, M.; Li, Z.; and Xie, P. 2021. Self-supervised Regularization for Text Classification . Transactions of the Association for Computational Linguistics, 9: 641--656

2021
[28]

Zhou, S.; and Li, X. 2020. Feature Engineering vs. Deep Learning for Paper Section Identification: Toward Applications in Chinese Medical Literature. Information Processing & Management, 57(3): 102206

2020

Pith tools

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

[1] [1]

, " * 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 mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

[2] [2]

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 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

Brack, A.; Entrup, E.; Stamatakis, M.; Buschermöhle, P.; Hoppe, A.; and Ewerth, R. 2024. Sequential sentence classification in research papers using cross-domain multi-task learning. International Journal on Digital Libraries

work page 2024

[4] [4]

Cohan, A.; Beltagy, I.; King, D.; Dalvi, B.; and Weld, D. 2019. Pretrained Language Models for Sequential Sentence Classification . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing , 3693--3699. Hong Kong, China: Association for Computational...

work page 2019

[5] [5]

d.; Gojenola, K.; and Atutxa, A

de la Iglesia, I.; Vivó, M.; Chocrón, P.; Maeztu, G. d.; Gojenola, K.; and Atutxa, A. 2023. An Open Source Corpus and Automatic Tool for Section Identification in Spanish Health Records. Journal of Biomedical Informatics, 145: 104461

work page 2023

[6] [6]

Dernoncourt, F.; and Lee, J. Y. 2017. P ub M ed 200k RCT : a Dataset for Sequential Sentence Classification in Medical Abstracts. In Kondrak, G.; and Watanabe, T., eds., Proceedings of the Eighth International Joint Conference on Natural Language Processing, 308--313. Taipei, Taiwan: Asian Federation of Natural Language Processing

work page 2017

[7] [7]

Y.; Karidi, T.; Choshen, L.; and Geva, M

Din, A. Y.; Karidi, T.; Choshen, L.; and Geva, M. 2023. Jump to Conclusions : Short - Cutting Transformers With Linear Transformations . ArXiv:2303.09435 [cs]

arXiv 2023

[8] [8]

FORCE11. 2020. The FAIR Data principles

work page 2020