Pith. sign in

REVIEW 3 major objections 6 minor 5 cited by

Comparative Analysis of Pooling Mechanisms in LLMs: A Sentiment Analysis Perspective

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that the best pooling mechanism for sentence-level sentiment analysis is model-specific: on IMDB, mean pooling gives BERT its top F1 score of 87.22%, while weighted sum pooling gives GPT-2 its top F1 score of 88.38%.

desk verdict A routine pooling comparison whose architecture-dependent conclusion rests on single-run F1 differences that are likely within noise. read the letter →

arxiv 2411.14654 v3 pith:H3YTQ2BB submitted 2024-11-22 cs.CL cs.AI

Jinming Xing , Dongwen Luo , Chang Xue , Ruilin Xing This is my paper · ORCID
classification cs.CLcs.AI
keywords poolingmechanismssentimentanalysisBERTGPT-2meanmaxweightedsumsentenceembeddings
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 tries to establish that no single pooling mechanism is universally best for sentence-level sentiment analysis: the best way to compress token embeddings into one sentence vector depends on the model architecture. Using the IMDB movie-review benchmark, it compares Mean, Max, and Weighted Sum pooling on BERT and GPT-2, reporting that BERT peaks at 87.22% F1 with Mean pooling while GPT-2 peaks at 88.38% F1 with Weighted Sum pooling. The paper's practical message is that pooling selection should be tuned per architecture and application rather than assumed by default.

What carries the argument

Three pooling operations applied to the token-level embeddings produced by the attention mechanism: Mean pooling averages all token vectors, Max pooling takes the elementwise maximum across tokens, and Weighted Sum pooling uses learnable per-token weights. The paper observes that mean pooling is weighted sum with uniform weights and max pooling is the limit where weight concentrates on one token, so the three methods form a single family of aggregation functions. These functions map the sequence of token embeddings to the sentence-level representation that feeds the sentiment classifier.

What would settle it

Repeat the BERT and GPT-2 experiments with multiple random seeds per pooling method and report the distribution of F1 scores. If the confidence intervals for Mean and Weighted Sum overlap for either model, the central claim that each architecture has a distinct best pooling mechanism is not supported by the data.

Watch

Extended reading notes

Core claim

The central claim is that pooling mechanism effectiveness is model-specific, not universal. On binary sentiment classification of IMDB reviews, BERT's best F1 score is 87.22% with Mean pooling, and GPT-2's best F1 score is 88.38% with Weighted Sum pooling. The authors interpret the results to mean that each pooling strategy carries distinct trade-offs: mean pooling gives a balanced, computationally cheap representation; max pooling selects the most salient features and yields the highest true-positive rate, though with a lower true-negative rate; and weighted sum pooling adds learnable token weights whose value depends on the model and task. Therefore the choice of pooling layer should be aligned with the model architecture and the application's requirements.

Load-bearing premise

The load-bearing premise is that single-run F1 and accuracy numbers from one 60/10/30 split of IMDB reliably rank the pooling methods; with no repeated seeds or significance testing, the reported gaps could be ordinary training noise.

Editorial extensions

If this is right

  • Mean pooling is the recommended default when computational cost matters, since it is simple and gives BERT its best measured performance.
  • Weighted Sum pooling is preferable when the model can exploit learned token emphasis, as it does for GPT-2 in this setup.
  • Max pooling is the choice when maximizing detection of the positive class is the priority, because it produced the highest true-positive rate at the cost of more false positives.
  • Pooling recommendations should not be transferred between architectures; the same pooling method performs differently on BERT than on GPT-2.

Reading between the lines

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

  • Since mean and max are special cases of weighted sum pooling, a fully converged weighted-sum layer should in principle match or exceed both; the fact that it fails to do so on BERT suggests optimization or overparameterization effects that the paper does not investigate.
  • The reported ordering comes from single-run scores, so an immediate testable extension is to repeat each configuration over multiple seeds and check whether the mean-versus-weighted-sum gap for BERT and GPT-2 persists.
  • Because the evidence is confined to one dataset and one task, the natural next experiment is to apply the same three pooling layers on other sentence-level benchmarks such as entailment, paraphrase, and question answering to see whether the architecture-pooling interaction generalizes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper reports an empirical comparison of Mean, Max, and Weighted Sum pooling for sentence-level sentiment classification using BERT-base and GPT2 on the IMDB dataset. The methodology section formalizes the three pooling operations, and the experiments report confusion matrices, precision, recall, and F1 scores. The central claim is that pooling effectiveness is architecture- and task-dependent, with BERT peaking at 87.22% F1 under Mean pooling and GPT2 peaking at 88.38% F1 under Weighted Sum pooling, leading to the practical recommendations in §4.3.3.

Significance. If the reported ranking is reliable, the paper would supply a useful practical datapoint for choosing pooling layers in transformer-based sentiment classifiers, and it would challenge the common practice of defaulting to Mean or CLS pooling. The experiments are simple and transparent enough that the central claim is directly testable by other groups. The significance is, however, limited by the absence of statistical uncertainty quantification; as it stands, the empirical contribution is a set of point estimates rather than a demonstrated architecture-dependent ordering. No code or data artifacts are provided, so reproducibility rests entirely on the written description.

major comments (3)
  1. [§4.3.2, Figure 4] The central claim that pooling effectiveness is architecture-dependent rests on single-run F1 point estimates. The text reports, for example, BERT's peak F1 of 87.22% with Mean pooling and GPT2's peak of 88.38% with Weighted Sum pooling, but no standard deviations, no repeated-seed runs, and no significance tests are reported anywhere in §4.3. In fine-tuning experiments of this size, run-to-run variation of 1–2 F1 points is common, and the gaps between pooling methods visible in Figure 4 appear to be in that range. Because every recommendation in §4.3.3 is derived from this ranking, the missing uncertainty quantification is load-bearing: the observed ordering could be training noise.
  2. [§4.1, Table 1] The text in §4.1 states that the dataset is 'meticulously balanced between positive and negative sentiments' and that the 'deliberate equal distribution' mitigates class imbalance, but Table 1 reports unequal counts in every split: 16,358 positive versus 13,642 negative in training, 2,466 versus 2,534 in validation, and 7,634 versus 7,366 in test. This is a factual inconsistency that affects the interpretation of accuracy and F1 as well as the reproducibility of the experimental setup. The authors should either provide the actual balanced split or revise the dataset description and discuss the small class imbalance.
  3. [§3.2.3, Eq. (4)] Weighted Sum pooling is defined as a sum of token embeddings with learnable weights, but the paper does not specify how the weights are produced, whether they share parameters across sequence positions, or how they are normalized. Without this detail the equation is not a well-defined architecture for variable-length inputs, and the claim that Mean and Max pooling are 'special cases' is loose: Mean is a special case only for uniform weights, and Max can be obtained only if the weights are data-dependent and concentrated on the maximum token, which is not how the learnable weighted sum appears to be defined. This methodological gap is relevant because Weighted Sum is reported to be the best method for GPT2.
minor comments (6)
  1. [§2, Related Work] The related work section lists many works in a block ([4–9]) without connecting them to the argument; consider trimming or integrating these citations into the narrative.
  2. [§4.3.1] The claim of an 'aggregate correct prediction rate of approximately 86.07%' is stated without showing the arithmetic or the underlying confusion matrices in the text; please state how this number is computed.
  3. [§3.1, Eq. (1)] Equation (1) has a typographical issue: the square root sign should apply to d_k, and the denominator notation should be made consistent with the softmax expression.
  4. [Header] The paper is labeled as a conference paper (ISMSI '25), but the header omits the proceedings volume and page numbers; please complete the bibliographic metadata.
  5. [Abstract] The phrase 'comprehensive experiments' overstates the scope: the evaluation uses a single dataset, one split, and two model families; a more cautious wording would better match the experiments.
  6. [§4.2.1] The statement that remaining hyperparameters use 'default configurations' is underspecified; please list the exact optimizer, epochs, batch size, warmup, and sequence length used, because the pooling comparison depends on these choices.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the pooling comparison is an empirical benchmark whose conclusions follow from reported measurements rather than from definitions or self-citations.

full rationale

The paper is an empirical benchmark, not a derivation. Its central claim—that pooling effectiveness depends on architecture—is supported by measured F1 values in §4.3.2 (BERT 87.22% with Mean pooling; GPT2 88.38% with Weighted Sum pooling). These numbers are outputs of fine-tuning experiments, not algebraic consequences of Equations (2)–(4). The observation in §3.2.3 that Mean and Max are special or limiting cases of Weighted Sum is a mathematical identity, but it is not used to infer which pooling method performs best; the rankings come from the experimental results in §4.3. The self-references (e.g., [23], [30]) appear in related-work context and do not supply any load-bearing premise for the paper's conclusions. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no ansatz is justified by self-citation. The lack of variance reporting or repeated seeds is a statistical robustness concern about the empirical claim, not circular reasoning; the conclusions are not equivalent to the inputs by construction. Therefore the circularity score is 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper rests on standard definitions of attention and pooling (Eqs. 1-4) and on the unstated assumption that the experimental setup, including the single IMDB dataset and default hyperparameters, is sufficient to support general conclusions about pooling mechanisms. No free parameters or invented entities are introduced.

assumptions (3)
  • standard math Standard attention and multi-head attention formulations (Eq. 1) are valid.
    Used as foundation in §3.1 without proof.
  • domain assumption IMDB dataset is balanced between positive and negative classes as claimed.
    Stated in §4.1 but contradicted by Table 1 counts; this assumption underpins the claim that class imbalance is mitigated.
  • ad hoc to paper One dataset and one train/validation/test split are representative enough to compare pooling mechanisms.
    The paper generalizes from IMDB only in §5 without supporting evidence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Comparative Analysis of Pooling Mechanisms in LLMs: A Sentiment Analysis Perspective." pith.science (2026). https://pith.science/paper/H3YTQ2BB

@misc{pith2026241114654,
  author       = {Pith},
  title        = {Pith review of: Comparative Analysis of Pooling Mechanisms in LLMs: A Sentiment Analysis Perspective},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H3YTQ2BB}},
  note         = {Machine review of arXiv:2411.14654}
}
read the original abstract

Large Language Models (LLMs) have revolutionized natural language processing (NLP) by delivering state-of-the-art performance across a variety of tasks. Among these, Transformer-based models like BERT and GPT rely on pooling layers to aggregate token-level embeddings into sentence-level representations. Common pooling mechanisms such as Mean, Max, and Weighted Sum play a pivotal role in this aggregation process. Despite their widespread use, the comparative performance of these strategies on different LLM architectures remains underexplored. To address this gap, this paper investigates the effects of these pooling mechanisms on two prominent LLM families -- BERT and GPT, in the context of sentence-level sentiment analysis. Comprehensive experiments reveal that each pooling mechanism exhibits unique strengths and weaknesses depending on the task's specific requirements. Our findings underline the importance of selecting pooling methods tailored to the demands of particular applications, prompting a re-evaluation of common assumptions regarding pooling operations. By offering actionable insights, this study contributes to the optimization of LLM-based models for downstream tasks.

Figures

Figures reproduced from arXiv: 2411.14654 by the authors.

Figure 1
Figure 1. Multi-Head Attention Mechanism 3.2 Pooling Strategies The attention mechanism generates token-level embeddings that must be aggregated to form sentence-level representations. Pooling layers play a crucial role in this aggregation process, condensing token embeddings into unified sentence embeddings. We explore three primary pooling strategies: Mean, Max, and Weighted Sum pooling. Given the token-level embeddings 𝑋 =… view at source ↗
Figure 2
Figure 2. Confusion Matrix of Bert with Different Pooling [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Confusion Matrix of GPT2 with Different Pooling [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Precision, Recall, F1 Score of Bert and GPT2 with [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Mechanistic Decomposition of Sentence Representations

    cs.CL 2025-06 conditional novelty 6.0 of 10

    Sentence embeddings can be decomposed into sparse, interpretable atoms via supervised dictionary learning, and mean pooling preserves mainly atoms aligned with the sentence direction.

  2. Looking around you: external information enhances representations for event sequences

    cs.LG 2025-02 conditional novelty 4.0 of 10

    Adding a learned Kernel attention aggregation of other users' event-sequence embeddings to a user's own embedding improves downstream ROC-AUC scores across several event-sequence datasets.

  3. Multi-view Fuzzy Graph Attention Networks for Enhanced Graph Learning

    cs.LG 2024-12 conditional novelty 4.0 of 10

    A graph classification model that adds linear multi-view projections and weighted-sum pooling to a fuzzy graph attention network reports accuracy gains on PROTEINS, NCI1, and Mutagenicity.

  4. FGATT: A Robust Framework for Wireless Data Imputation Using Fuzzy Graph Attention Networks and Transformer Encoders

    cs.LG 2024-12 reject novelty 4.0 of 10

    FGATT, a fuzzy graph attention network paired with a Transformer encoder, is claimed to improve missing-data imputation on two SWaT wireless sensor datasets.

  5. Enhanced Convolutional Neural Networks for Improved Image Classification

    cs.CV 2025-02 reject novelty 2.0 of 10

    An enhanced CNN with standard techniques claims 84.95% on CIFAR-10, but weak baselines and missing evidence undermine the contribution.

Reference graph

Works this paper leans on

33 extracted references · 8 canonical work pages · cited by 5 Pith papers

  1. [1]

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

    Devlin, Jacob. "Bert: Pre-training of deep bidirectional transformers for language understanding. " arXiv preprint arXiv:1810.04805 (2018)

    arXiv 2018

  2. [2]

    Improving language understanding by generative pre-training

    Radford, Alec. "Improving language understanding by generative pre-training. " (2018)

    work page 2018

  3. [3]

    GPT-3: Its nature, scope, limits, and consequences

    Floridi, Luciano, and Massimo Chiriatti. "GPT-3: Its nature, scope, limits, and consequences. " Minds and Machines 30 (2020): 681-694

    work page 2020

  4. [4]

    Patch-aware Vector Quantized Codebook Learning for Unsupervised Visual Defect Detection

    Cheng, Qisen, Shuhui Qu, and Janghwan Lee. "Patch-aware Vector Quantized Codebook Learning for Unsupervised Visual Defect Detection. " In *2024 IEEE 36th International Conference on Tools with Artificial Intelligence (ICTAI)*, pp. 586-592. IEEE, 2024

    2024

  5. [5]

    Unifying Prediction and Explanation in Time-Series Transformers via Shapley-based Pretraining

    Cheng, Qisen, Jinming Xing, Chang Xue, and Xiaoran Yang. "Unifying Prediction and Explanation in Time-Series Transformers via Shapley-based Pretraining. " *arXiv preprint arXiv:2501.15070* (2025)

    arXiv 2025

  6. [6]

    72-3: Deep Learning Based Visual Defect Detection in Noisy and Imbalanced Data

    Cheng, Qisen, Shuhui Qu, and Janghwan Lee. "72-3: Deep Learning Based Visual Defect Detection in Noisy and Imbalanced Data. " In *SID Symposium Digest of Technical Papers*, vol. 53, no. 1, pp. 971-974. 2022

    2022

  7. [7]

    Estimation of Energy and Time Us- age in 3D Printing With Multimodal Neural Network

    Cheng, Qisen, Chang Zhang, and Xiang Shen. "Estimation of Energy and Time Us- age in 3D Printing With Multimodal Neural Network. " In *2022 4th International Conference on Frontiers Technology of Information and Computer (ICFTIC)*, pp. 900-903. IEEE, 2022

    2022

  8. [8]

    Network Traffic Forecasting via Fuzzy Spatial-Temporal Fusion Graph Neural Networks

    Xing, Jinming, Zhaomin Xiao, Yingyi Wu, Jinran Zhang, Zhuoer Xu, and Zhelu Mai. "Network Traffic Forecasting via Fuzzy Spatial-Temporal Fusion Graph Neural Networks. " In *2024 11th International Conference on Soft Computing & Machine Intelligence (ISCMI)*, pp. 282-286. IEEE, 2024

    2024

Show all 33 references
  1. [9]

    SHAPNN: Shapley Value Regular- ized Tabular Neural Network

    Cheng, Qisen, Shuhui Qu, and Janghwan Lee. "SHAPNN: Shapley Value Regular- ized Tabular Neural Network. " *arXiv preprint arXiv:2309.08799* (2023)

    2023 arXiv

  2. [10]

    A survey on large language model (llm) security and privacy: The good, the bad, and the ugly

    Yao, Yifan, et al. "A survey on large language model (llm) security and privacy: The good, the bad, and the ugly. " High-Confidence Computing (2024): 100211

    2024

  3. [11]

    Botnet detection based on machine learning

    Yang, Xiaoran, Zhen Guo, and Zetian Mai. "Botnet detection based on machine learning. " In 2022 International Conference on Blockchain Technology and Infor- mation Security (ICBCTIS), pp. 213-217. IEEE, 2022

    2022

  4. [12]

    Research and application of knowledge graph technology for intelligent question answering

    Lin, Chengrong, Shaofan Chen, Xiaoran Yang, Caimao Li, Cong Qu, and Qiuhong Chen. "Research and application of knowledge graph technology for intelligent question answering. " In 2021 12th International Symposium on Parallel Architec- tures, Algorithms and Programming (PAAP), ...

    2021

  5. [13]

    Research on automatic composition based on multiple machine learning models

    Yang, Xiaoran. "Research on automatic composition based on multiple machine learning models. " In 2021 3rd International Conference on Artificial Intelligence and Advanced Manufacture, pp. 1206-1209. 2021

    2021

  6. [14]

    Contextual Combinatorial Beam Management via Online Probing for Mul- tiple Access mmWave Wireless Networks

    Li, Zhizhen, Xuanhao Luo, Mingzhe Chen, Chenhan Xu, Shiwen Mao, and Yuchen Liu. "Contextual Combinatorial Beam Management via Online Probing for Mul- tiple Access mmWave Wireless Networks. " *arXiv preprint arXiv:2412.10385* (2024)

    2024 arXiv

  7. [15]

    RM- Gen: Conditional Diffusion Model-Based Radio Map Generation for Wireless Networks

    Luo, Xuanhao, L. Zhizhen, Zhiyuan Peng, X. Dongkuan, and Yuchen Liu. "RM- Gen: Conditional Diffusion Model-Based Radio Map Generation for Wireless Networks. " In *2024 IFIP Networking Conference (IFIP Networking)*, pp. 543-548. IEEE, 2024

    2024

  8. [16]

    Image Retrieval Based on Blockchain

    Zhao, Sihao, Jiahui Xie, Chu Lin, Xiaolan Nie, Jun Ye, Xiaoran Yang, and Pengzhi Xu. "Image Retrieval Based on Blockchain. " In 2022 International Conference on Blockchain Technology and Information Security (ICBCTIS), pp. 210-212. IEEE, 2022

    2022

  9. [17]

    Analysis of Lifting Posture by Two Inertial Measurement Units and a Classification Model Based on a Convolutional Neural Network

    Liu, Juncheng, Xiaoran Yang, Jing-Chen Hong, and Hiroyasu Iwata. "Analysis of Lifting Posture by Two Inertial Measurement Units and a Classification Model Based on a Convolutional Neural Network. " In 2024 10th IEEE RAS/EMBS Inter- national Conference for Biomedical Robotics a...

    2024

  10. [18]

    Yang, Xiaoran, Yang Zhan, Yukiko Iwasaki, Miaohui Shi, Shijie Tang, and Hi- royasu Iwata. "Balancing Real-world Interaction and VR Immersion with AI Comparative Analysis of Pooling Mechanisms in LLMs: A Sentiment Analysis Perspective ISMSI ’25, Tokyo, Japan, Vision Robotic Arm...

    2023

  11. [19]

    Short-Term Traffic Condition Prediction Based on Multi-source Data Fusion

    Deng, Xiaoru, Hui Zhou, Xiaoran Yang, and Chunyang Ye. "Short-Term Traffic Condition Prediction Based on Multi-source Data Fusion. " In International Con- ference on Data Mining and Big Data, pp. 327-335. Singapore: Springer Singapore, 2021

    2021

  12. [20]

    TCellSI: A novel method for T cell state assessment and its applications in immune environment prediction

    Yang, Jing-Min, Nan Zhang, Tao Luo, Mei Yang, Wen-Kang Shen, Zhen-Lin Tan, Yun Xia et al. "TCellSI: A novel method for T cell state assessment and its applications in immune environment prediction. " *Imeta* 3, no. 5 (2024): e231

    2024

  13. [21]

    Improvement and Implementation of a Speech Emotion Recognition Model Based on Dual-Layer LSTM

    Yang, Xiaoran, Shuhan Yu, and Wenxi Xu. "Improvement and Implementation of a Speech Emotion Recognition Model Based on Dual-Layer LSTM. " arXiv preprint arXiv:2411.09189 (2024)

    2024 arXiv

  14. [22]

    Attention is all you need

    Vaswani, A. "Attention is all you need. " Advances in Neural Information Process- ing Systems (2017)

    2017

  15. [23]

    Enhancing Link Prediction with Fuzzy Graph Attention Networks and Dynamic Negative Sampling

    Xing, Jinming. "Enhancing Link Prediction with Fuzzy Graph Attention Networks and Dynamic Negative Sampling. " arXiv preprint arXiv:2411.07482 (2024)

    2024 arXiv

  16. [24]

    Roberta: A robustly optimized bert pretraining approach

    Liu, Yinhan. "Roberta: A robustly optimized bert pretraining approach. " arXiv preprint arXiv:1907.11692 364 (2019)

    2019 arXiv

  17. [25]

    DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter

    Sanh, V. "DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. " arXiv preprint arXiv:1910.01108 (2019)

    2019 arXiv

  18. [26]

    Language models are unsupervised multitask learners

    Radford, Alec, et al. "Language models are unsupervised multitask learners. " OpenAI blog 1.8 (2019): 9

    2019

  19. [27]

    Language models are few-shot learners

    Brown, Tom B. "Language models are few-shot learners. " arXiv preprint arXiv:2005.14165 (2020)

    2020 arXiv

  20. [28]

    Gpt-4 technical report

    Achiam, Josh, et al. "Gpt-4 technical report. " arXiv preprint arXiv:2303.08774 (2023)

    2023 arXiv

  21. [29]

    Chatting about ChatGPT: how may AI and GPT impact academia and libraries?

    Lund, Brady D., and Ting Wang. "Chatting about ChatGPT: how may AI and GPT impact academia and libraries?. " Library hi tech news 40.3 (2023): 26-29

    2023

  22. [30]

    Weighted fuzzy rough sets-based tri- training and its application to medical diagnosis

    Xing, Jinming, Can Gao, and Jie Zhou. "Weighted fuzzy rough sets-based tri- training and its application to medical diagnosis. " Applied Soft Computing 124 (2022): 109025

    2022

  23. [31]

    Parameterized maximum-entropy-based three-way approximate attribute reduction

    Gao, Can, et al. "Parameterized maximum-entropy-based three-way approximate attribute reduction. " International Journal of Approximate Reasoning 151 (2022): 85-100

    2022

  24. [32]

    Detecting formal thought disorder by deep contextualized word representations

    Sarzynska-Wawer, Justyna, et al. "Detecting formal thought disorder by deep contextualized word representations. " Psychiatry Research 304 (2021): 114135

    2021

  25. [33]

    Supervised learning of universal sentence representations from natural language inference data

    Conneau, Alexis, et al. "Supervised learning of universal sentence representations from natural language inference data. " arXiv preprint arXiv:1705.02364 (2017)

    2017 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.