Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

Towards Objective and Unbiased Decision Assessments with LLM-Enhanced Hierarchical Attention Networks

T0 review · 5 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 A hierarchical attention network augmented by an LLM analysis agent predicts university admission outcomes with 9.6% higher F1 and accuracy than human shortlisting and interview recommendation, suggesting that structured AI workflows can…

desk verdict The 9.6% human-beating claim is invalid because the model is trained on final outcomes and compared to intermediate human stages, and the reference list contains signs of fabrication. read the letter →

arxiv 2411.08504 v2 pith:LNXIDJQT submitted 2024-11-13 cs.CL cs.AI

Junhua Liu , Kwan Hui Lim , Roy Ka-Wei Lee This is my paper · ORCID
classification cs.CLcs.AI
keywords AI-AugmentedDecisionsDecisionBiasMitigationAgenticWorkflowHierarchicalLearningLargeLanguageModelsUniversityAdmissionsAttentionNetworkCognitive
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 argues that cognitive bias and inconsistency in high-stakes human decision making can be detected statistically and reduced by an AI-augmented workflow. Using real-world university admission data, it shows that correlations among human decision points (shortlisting, interview recommendation, final offer) are far from perfect, implying subjectivity. The authors propose BGM-HAN, a hierarchical attention network with byte-pair encoding, gated residual connections, and multi-head attention, plus a Shortlist-Analyse-Recommend (SAR) agentic workflow that mimics the human process. They report that the full workflow outperforms human evaluators by over 9.6% in F1-score and accuracy when predicting final admission decisions. If correct, this suggests that structured AI intervention can improve consistency and fairness in high-stakes evaluations.

What carries the argument

The central object is BGM-HAN, an enhanced Hierarchical Attention Network that processes each student profile field (GCE A-Level, GCE O-Level, leadership records, personal insight essays) through three levels — token, sentence, field — using multi-head self-attention with gated residual connections and layer normalization at each level. Byte-pair encoding supplies the subword vocabulary, and the resulting field embeddings are concatenated and passed to an MLP for classification. Around this backbone, the Shortlist-Analyse-Recommend (SAR) workflow chains three agents: a shortlisting agent (A_S), an LLM-based analysis agent (A_A) using Gemini-1.5-Pro, and a recommendation agent (A_R) that combines profile and analysis embeddings. The key mechanism is hierarchical representation learning: the model can attend to the most informative parts of a semi-structured profile while the gated residual connections stabilize training.

What would settle it

Recompute human shortlisting and interview-recommendation metrics against the same final admission labels, using the same stratified test split and class weights, and report confusion matrices; if the resulting human F1/accuracy approach or exceed 0.8945/0.8966, the claimed 9.6% improvement disappears. Additionally, test the trained model on profiles where the final outcome was decided by a different committee; if performance drops to human levels, the model may have memorized one committee's bias rather than removed it.

Watch

Extended reading notes

Core claim

The central claim is that the proposed BGM-HAN-W_SAR pipeline predicts final university admission outcomes more accurately than the human decision stages it is designed to assist, achieving F1 0.8945 and accuracy 0.8966 against human shortlisting (F1 0.8156) and interview recommendation (F1 0.8321). The paper further claims that each architectural enhancement contributes measurably: BPE tokenization adds about 1.8%, multi-head attention 5.2%, and gated residual connections 2.6% over the base HAN, and that the agentic workflow with an LLM-generated analysis improves results further regardless of which LLM is used. On the paper's own terms, the workflow mitigates inconsistency and cognitive bias across decision makers while providing interpretable rationales.

Load-bearing premise

The paper assumes that the final admission outcome is an unbiased ground truth and that human shortlisting and interview decisions are directly comparable benchmarks for predicting that outcome; if the final outcome inherits the same human biases, or if the intermediate stages are not meant to match the final decision exactly, the 'surpasses human judgment' result collapses.

Editorial extensions

If this is right

  • If the reported results hold, admissions offices could deploy the workflow as a screening and recommendation aid that matches or exceeds the consistency of human shortlisting and interview stages.
  • The ablation numbers imply that the improvement is not from any single trick but from the combination of subword tokenization, multi-head attention, and gated residuals on hierarchical data.
  • The workflow's dependence on an LLM-generated analysis is weak: swapping Gemini-1.5-Pro for other LLMs changes accuracy by only about 0.049-0.051, so the gain is robust to the choice of analysis generator.
  • The same approach could be transferred to other high-stakes decisions with semi-structured text inputs, such as job applicant screening, vendor selection, or loan approval, as the paper states in its impact statement.

Reading between the lines

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

  • The paper's comparison treats the final admission outcome as ground truth; a natural extension would be to test whether the model's decisions are less biased than the human stages by measuring prediction errors against protected attributes, which the paper does not do.
  • Because the human metrics are computed on different decision stages, the 9.6% improvement is not apples-to-apples; an external reader should compare the model against a human committee asked to predict the same final outcome on the same candidate set.
  • The correlation analysis shows SL-DO at 0.67 and AR-DO at 0.62; a stronger consistency test would be inter-rater reliability (e.g., Cohen's kappa) between multiple human evaluators on the same profiles, which would quantify bias more directly.
  • The workflow's analysis agent is prompted to look for 'red flags' and institutional values; this introduces a normative framework that could itself encode biases, so an extension would be to audit the LLM's generated analyses for biased language.
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 / 6 minor

Summary. This paper addresses cognitive bias in university admission decisions. The authors analyze correlations among decision points (shortlisting, interview recommendation, degree offer) and interpret moderate correlations as evidence of bias. They propose BGM-HAN, a hierarchical attention network with byte-pair encoding, gated residual connections, and multi-head attention, and a Shortlist-Analyze-Recommend (SAR) agentic workflow that uses an LLM to generate profile analyses and a second classifier to produce final recommendations. On a real admissions dataset, the full workflow is reported to achieve F1 0.8945 and accuracy 0.8966, which the authors claim is over 9.6% higher than human evaluation metrics. They also report improvements over several baselines and provide an ablation study.

Significance. The paper has the ingredients of a useful applied study: it uses a real-world admissions dataset, releases code, evaluates a broad set of baselines, and combines a hierarchical model with LLM-generated features. The architecture is plausible and the relative ranking of baselines is internally consistent. However, the central contributions as stated—surpassing human judgment and mitigating cognitive bias—are not supported by the evidence. The human comparison is undefined and not apples-to-apples, and the bias-mitigation claim is never directly tested. As such, the current version cannot be recommended for publication; with a redefined evaluation and explicit bias metrics, the underlying method might be publishable elsewhere.

major comments (5)
  1. [Section 5.5, Table 3] The manuscript never states how the human evaluation metrics (precision, recall, F1, accuracy) for Shortlisting and Interview-Rec are computed, nor which ground-truth labels are used. Without this definition, the headline claim of over 9.6% improvement over human judgment is uninterpretable and cannot be audited.
  2. [Section 5.5, Table 3] The comparison in Table 3 is between a model trained to predict the final admission outcome and human decisions made at intermediate stages (shortlisting and interview recommendation). These are different prediction targets: an intermediate screening decision need not match the final committee decision, and the human evaluators at those stages may not have access to the full information used by the model. The reported 9.6% improvement therefore does not establish that the model surpasses human judgment on the same task.
  3. [Section 3.2, Figure 3] The correlation coefficients of 0.67 (SL-DO) and 0.62 (AR-DO) are interpreted as evidence of cognitive bias, but this inference is not justified. The paper does not rule out legitimate stage-specific differences in available information or decision objectives, and it provides no direct measure of bias (e.g., subgroup differences or consistency across raters).
  4. [Overall evaluation, Sections 5 and 8] The paper claims that the SAR workflow mitigates inconsistency and cognitive bias, but no bias or fairness metric is ever computed. Because the model is trained on the same human committee outcomes that are claimed to be biased, the reported accuracy may amount to replicating the committee's decision surface. An explicit evaluation of bias mitigation (e.g., calibration across applicant groups, consistency of decisions on matched profiles) is needed to support the title's claim.
  5. [Section 4.5 and 5.2] The decision thresholds tau and delta are free parameters, and the manuscript does not describe how they were selected. If they were tuned on the test set, the reported results would be optimistic. In addition, no confidence intervals or statistical significance tests are provided for any of the comparisons, including the headline '9.6%' improvement.
minor comments (6)
  1. [Section 3.1] The dataset size is not reported. Please provide the number of profiles and the class balance, since the 90-5-5 split and weighted loss depend on these figures.
  2. [Tables 3 and 4] Tables 3 and 4 are inconsistent for the HAN baseline (F1 0.7711 in Table 3 vs 0.5833 in Table 4); please clarify the experimental setting or correct the numbers.
  3. [Section 5.6.1] The component-wise improvements attributed to BPE (1.8%), multi-head attention (5.2%), and gated residual connections (2.6%) are not reported in any table; please add the corresponding ablation results.
  4. [Throughout] There are numerous typos and grammar errors, including 'among in' (Abstract), 'Connetions' (Section 1), 'optimial' (Section 5.2), 'provding' (Section 5.5.3), and 'difference decision points' (Section 3.2).
  5. [References] Reference [20] duplicates reference [19] with different page numbers, and reference [14] attributes a 2023 article on sustainable decision making to Kahneman and Tversky; please verify these citations.
  6. [Section 4.4.1 and 5.5.1] The class-weight formula w_y = N/(2N_y) is stated without reporting the actual class distribution; also, Section 5.5.1 calls the F1 score 'macro-averaged' for a binary task, which should be clarified.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline 'over 9.6% improvement over human evaluation' is scored on the same final-admission label the model is trained to predict, while the human baselines are intermediate decision points that the paper itself shows are only weakly coupled to that label.

  1. fitted input called prediction [Section 3.1, Section 3.2, Section 4.4.1, Table 3 (Section 5.5)]
    "Each profile is labeled with a binary admission decision indicating whether an offer was made (1) or not (0). ... despite the strong correlations between SL and DO (0.67) and between AR and DO (0.62), they are far from perfect correlations of 1.0. ... Human Evaluation: Shortlisting 0.8464 0.8418 0.8156 0.8155; Interview-Rec 0.8011 0.8193 0.8321 0.8087."

    Section 4.4.1 trains the agents with weighted cross-entropy on the binary admission decision, and Section 4.5 outputs the final admission decision from the same label. Table 3 reports 'Human Evaluation' using only Shortlisting and Interview-Rec, which are different, intermediate decision points with correlations of only 0.67 and 0.62 to the final degree offer. Since no other reference label is defined for the human rows, the comparison necessarily scores human intermediate decisions against the very final-offer label the model was explicitly optimized to predict.

full rationale

The BGM-HAN architecture is a standard composition of published components (HAN, BPE, multi-head attention, gated residual connections), and no load-bearing self-citation or imported uniqueness theorem appears in the derivation. The main circularity is in the headline human-comparison: the model is trained to reproduce the final admission label, while the human baselines are shortlisting and interview-recommendation stages that the paper's own correlation analysis shows are far from identical to that final label. Scoring those intermediate human judgments against the final-label target makes the model's superior F1 and accuracy partly a consequence of the evaluation setup. Additionally, the 9.6% figure is used inconsistently: Section 5.6.1 attributes it to BGM-HAN over HAN, whereas the abstract and conclusion attribute it to comparison with human evaluation. These are benchmark-construction and reporting problems rather than a formal derivation that reduces to its inputs, so the circularity is partial rather than total.

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

The central claim depends on several unreported or post-hoc choices: decision thresholds, grid-searched hyperparameters, BPE vocabulary size, and the final-outcome-as-ground-truth assumption. No new physical entities are introduced. The LLM 'agents' are software components, not invented scientific entities.

free parameters (5)
  • Decision thresholds tau and delta = not reported
    The final decision rule (Section 4.5) requires P_s(p_i) > tau and P_r(p_i) > delta; the reported accuracy and F1 depend on these values, which are omitted. They appear tuned on the validation set.
  • Optimal hyperparameters from grid search = not reported
    Section 5.2 lists the search space (hidden dimension, heads, dropout, learning rate, batch size) and says the optimal set is bolded, but the bold values do not appear in the text, so the chosen configuration is not stated.
  • BPE vocabulary size V = 5000
    Chosen by hand in Section 4.1.2; no sensitivity analysis.
  • Maximum sentences and words per sentence (s, w) = (10, 50)
    Fix the tensor dimensions in Algorithm 2; no justification is given.
  • LLM generation parameters = temperature 0.3, top-p 0.8, top-k 40
    Appendix A.2; these parameters affect the analysis text and therefore the recommendation outcome.
assumptions (4)
  • domain assumption The final admission outcome is a valid unbiased ground truth for both the human decision stages and the model.
    The entire evaluation in Table 3 compares predictions against this label, yet the label is produced by the same human process the paper claims is biased. Section 3.1 and Section 4 define the mapping to {0,1} without questioning the label's own bias.
  • ad hoc to paper Correlation coefficients of 0.67 (SL-DO) and 0.62 (AR-DO) indicate cognitive bias rather than legitimate differences in stage-wise judgment.
    Section 3.2 concludes 'variability in evaluations, which could be influenced by subjectivity and cognitive biases' solely from correlations below 1.0, with no significance tests or alternative explanations considered.
  • domain assumption LLM-generated analyses are faithful, complete summaries that add signal beyond the raw profile.
    The Recommendation Agent consumes the Gemini-generated analysis as an input feature (Sections 4.2 and 5.3.1); if the analysis paraphrases only parts of the profile or injects the LLM's own biases, the training signal changes.
  • domain assumption The 90-5-5 stratified split gives a reliable test estimate.
    No dataset size or confidence intervals are provided, so the few test samples may make the reported differences statistically insignificant.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Objective and Unbiased Decision Assessments with LLM-Enhanced Hierarchical Attention Networks." pith.science (2026). https://pith.science/paper/LNXIDJQT

@misc{pith2026241108504,
  author       = {Pith},
  title        = {Pith review of: Towards Objective and Unbiased Decision Assessments with LLM-Enhanced Hierarchical Attention Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LNXIDJQT}},
  note         = {Machine review of arXiv:2411.08504}
}
read the original abstract

How objective and unbiased are we while making decisions? This work investigates cognitive bias identification in high-stake decision making process by human experts, questioning its effectiveness in real-world settings, such as candidates assessments for university admission. We begin with a statistical analysis assessing correlations among different decision points among in the current process, which discovers discrepancies that imply cognitive bias and inconsistency in decisions. This motivates our exploration of bias-aware AI-augmented workflow that surpass human judgment. We propose BGM-HAN, an enhanced Hierarchical Attention Network with Byte-Pair Encoding, Gated Residual Connections and Multi-Head Attention. Using it as a backbone model, we further propose a Shortlist-Analyse-Recommend (SAR) agentic workflow, which simulate real-world decision-making. In our experiments, both the proposed model and the agentic workflow significantly improves on both human judgment and alternative models, validated with real-world data.

Figures

Figures reproduced from arXiv: 2411.08504 by the authors.

Figure 1
Figure 1. University Admission Decision Process: overview [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Architecture overview of the proposed BGM-HAN model. The multi-level model learns features from token to sentence [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Intuitively, we expect the admission decision (i.e. DO) to adhere to SL and AR, where students who are shortlisted and/or recom￾mended for admission should have a high ratio of receiving degree 1Due to confidentiality, the content of the options are not disclosed. offers. However, looking at [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (1 more)
Figure 3
Figure 3. Figure 3: Correlation matrix of different decision points in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Balancing Accuracy and Efficiency in Multi-Turn Intent Classification for LLM-Powered Dialog Systems in Production

    cs.CL 2024-11 conditional novelty 4.0 of 10

    Compressing intent labels for LLM fine-tuning and using self-consistency-filtered LLM pseudo-labeling improve multi-turn intent classification accuracy and enable small, low-latency production models.

Reference graph

Works this paper leans on

30 extracted references · 12 canonical work pages · cited by 1 Pith paper

  1. [1]

    Rohan Alur, Loren Laine, Darrick Li, Manish Raghavan, Devavrat Shah, and Den- nis Shung. 2024. Auditing for human expertise. Advances in Neural Information Processing Systems 36 (2024)

    work page 2024

  2. [2]

    Julia Angwin, Jeff Larson, Surya Mattu, and Lauren Kirchner. 2016. Machine Bias: There’s software used across the country to predict future criminals. And it’s biased against blacks. ProPublica (2016). https://www.propublica.org/article/ machine-bias-risk-assessments-in-criminal-sentencing

    2016

  3. [3]

    Anthropic. 2024. The Claude 3 Model Family: Opus, Sonnet, Haiku. https: //www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/ Model_Card_Claude_3.pdf

    work page 2024

  4. [4]

    Soumya Basu, Ankit Singh Rawat, and Manzil Zaheer. 2023. A Statistical Perspec- tive on Retrieval-Based Models. InProceedings of the 40th International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 202), Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jonathan Scarlett (Eds.). PMLR, 1852–1886

    work page 2023

  5. [5]

    Kyunghyun Cho, Bart van Merrienboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using RNN encoder–decoder for statistical machine translation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP). 1724–1734

    work page 2014

  6. [6]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies . 4171–4186

    work page 2019

  7. [7]

    Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jégou. 2024. The Faiss library. (2024). arXiv:2401.08281 [cs.LG]

    arXiv 2024

  8. [8]

    Julia Dressel and Hany Farid. 2018. The accuracy, fairness, and limits of predicting recidivism. Science Advances 4, 1 (2018), eaao5580. https://doi.org/10.1126/sciadv. aao5580

    doi:10.1126/sciadv 2018

Show all 30 references
  1. [9]

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783 (2024)

    2024 arXiv

  2. [10]

    Jessica Echterhoff and et al. 2024. Cognitive Bias in High-Stakes Decision-Making with LLMs. arXiv preprint arXiv:2403.00811 (2024). https://arxiv.org/abs/2403. 00811

    2024 arXiv

  3. [11]

    Bhavya Ghai and Klaus Mueller. 2022. D-BIAS: A Causality-Based Human-in- the-Loop System for Tackling Algorithmic Bias. arXiv preprint arXiv:2208.05126 (2022). https://arxiv.org/abs/2208.05126

    2022 arXiv

  4. [12]

    Felix Haag, Carlo Stingl, Katrin Zerfass, Konstantin Hopf, and Thorsten Staake

  5. [13]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory.Neural Computation 9, 8 (1997), 1735–1780

    1997

  6. [14]

    Kahneman and A

    D. Kahneman and A. Tversky. 2023. Cognitive Bias and How to Improve Sustainable Decision Making. Frontiers in Psychology 14 (2023), 10071311. https://pmc.ncbi.nlm.nih.gov/articles/PMC10071311/

    2023

  7. [15]

    Guillaume Lample and Alexis Conneau. 2019. Cross-lingual language model pretraining. In Advances in Neural Information Processing Systems , Vol. 32

    2019

  8. [16]

    Junhua Liu, Yung Chuen Ng, Zitong Gui, Trisha Singhal, Lucienne T. M. Blessing, Kristin L. Wood, and Kwan Hui Lim. 2022. Title2Vec: a contextual job title embedding for occupational named entity recognition and other applications. Journal of Big Data 9, 1 (2022), 1–16

    2022

  9. [17]

    Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. RoBERTa: A robustly optimized BERT pretraining approach. arXiv preprint arXiv:1907.11692 (2019)

    2019 arXiv

  10. [18]

    OpenAI. 2024. GPT-4 Technical Report. arXiv:2303.08774 [cs.CL]

    2024 arXiv

  11. [19]

    Gloria Phillips-Wren, Daniel J Power, and Manuel Mora. 2019. Cognitive bias, decision styles, and risk attitudes in decision making and DSS. , 63–66 pages

    2019

  12. [20]

    Gloria Phillips-Wren, Daniel J Power, and Manuel Mora. 2019. Cognitive bias, decision styles, and risk attitudes in decision making and DSS. Decision Support Systems 63 (2019), 63–66

    2019

  13. [21]

    João G Ribeiro, Frederico S Felisberto, and Isabel C Neto. 2020. Pruning and Sparsemax Methods for Hierarchical Attention Networks. arXiv preprint arXiv:2004.04343 (2020). https://arxiv.org/abs/2004.04343

    2020 arXiv

  14. [22]

    Pedro HP Savarese, Leonardo O Mazza, and Daniel R Figueiredo. 2016. Learning identity mappings with residual gates. arXiv preprint arXiv:1611.01260 (2016)

    2016 arXiv

  15. [23]

    John Smith, Jane Doe, and Andrew Lee. 2024. Bias and Fairness in High-Stakes AI: Challenges of Data Sensitivity and Access. Ethics and Information Technology 26 (2024), 85–102. https://doi.org/10.1007/s10676-024-09746-w

    2024 doi

  16. [24]

    A Vaswani. 2017. Attention is all you need. Advances in Neural Information Processing Systems (2017)

    2017

  17. [25]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in Neural Information Processing Systems , Vol. 30

    2017

  18. [26]

    Mingzhe Yang and et al. 2024. Fair Machine Guidance to Enhance Fair Decision Making in Biased People. arXiv preprint arXiv:2404.05228 (2024). https://arxiv. org/abs/2404.05228

    2024 arXiv

  19. [27]

    Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alex Smola, and Eduard Hovy. 2016. Hierarchical attention networks for document classification. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language...

    2016

  20. [28]

    Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alex Smola, and Eduard Hovy. 2016. Hierarchical attention networks for document classification. In Proceedings of the 2016 conference of the North American chapter of the association for computational linguistics: human language...

    2016

  21. [29]

    Vilém Zouhar, Clara Meister, Juan Gastaldi, Li Du, Tim Vieira, Mrinmaya Sachan, and Ryan Cotterell. 2023. A Formal Perspective on Byte-Pair Encoding. In Findings of the Association for Computational Linguistics: ACL 2023 , Anna Rogers, Jordan Boyd-Graber, and Naoaki Okazaki (E...

    2023 doi

  22. [2024]

    arXiv preprint arXiv:2405.04972 (2024)

    Overcoming Anchoring Bias: The Potential of AI and XAI-based Decision Support. arXiv preprint arXiv:2405.04972 (2024). https://arxiv.org/abs/2405.04972

    2024 arXiv

Pith tools

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