Pith. sign in

REVIEW 4 major objections 5 minor 36 references

Interim Report on Human-Guided Adaptive Hyperparameter Optimization with Multi-Fidelity Sprints

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a human-guided, multi-fidelity sprint workflow with post-training threshold meta-learning identifies a high-performing joint entity-relation extraction configuration, reporting about 49.64 relation F1 on Re-DocRED…

desk verdict An honest interim HPO engineering report whose headline result is undermined by an internally inconsistent table and a missing same-dataset baseline; it deserves a serious referee, but major revision before the claims can stand. read the letter →

arxiv 2505.09792 v1 pith:WNI3Y5UZ submitted 2025-05-14 cs.CL cs.LG

Michael Kamfonas This is my paper · ORCID
classification cs.CLcs.LG
keywords hyperparameteroptimizationmulti-fidelityBayesianjointentityandrelationextractionthresholdcalibrationmulti-tasklearninghuman-in-the-loopRe-DocRED
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 proposes and tests a practical workflow for tuning expensive multitask NLP models: run cheap low-fidelity Bayesian optimization sprints on rotating data subsets, let a human shrink the hyperparameter ranges based on the top results, then progress through warmer and finally full-fidelity sprints with early pruning, and finish by calibrating decision thresholds with a hill-climbing meta-learner. The claim is that this phased process finds strong configurations within a constrained compute budget, and that threshold calibration is a real performance lever. It demonstrates the workflow on sixteen variants of the JEREX-L joint entity and relation extraction model trained on Re-DocRED, reporting the best result of about 49.64 relation F1 for the long-document transformer variant with dynamic task-loss weighting. A sympathetic reader would care because the framework is reusable: it offers a blueprint for comparing model variants and tuning hyperparameters when full-fidelity training is too expensive to brute-force.

What carries the argument

The load-bearing mechanism is the sprint thread: a sequence of Bayesian optimization sessions at increasing fidelity, each with its own hyperparameter space and initialization. Phase 1 runs about 120 single-epoch trials on rotating one-sixth training and one-third validation subsets with the learning-rate scheduler disabled, then prunes each hyperparameter dimension to the range spanned by the top 10 trials plus a margin. Phase 2 runs 60–90 trials at the same data fidelity with the scheduler enabled and an early stop at the end of the warmup period, optimizing the scheduler's maximum learning rate and warmup length. Phase 3 cold-primes a full-fidelity tree-structured Parzen estimator study with Hyperband pruning using the top 3 Phase 2 configurations and runs nine 25-epoch trials plus a calibration phase. The calibration component is a hill-climbing meta-learner that perturbs mention, coreference, and relation thresholds, optionally per relation class, to maximize validation relation F1. Cold priming—sharing top hyperparameter sets without carrying over scores—is what lets knowledge cross a fidelity boundary without contamination.

What would settle it

Run full 25-epoch calibrated training for hyperparameter combinations that the Phase 1 top-10 pruning would have excluded; if any such combination scores above 49.64 relation F1 on Re-DocRED, the consistency assumption behind low-fidelity pruning is violated.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a three-sprint pipeline—single-epoch low-fidelity exploration, warmup-period tuning with the scheduler active, and a short full-fidelity study with Hyperband pruning—combined with post-training threshold meta-learning can identify a high-performing joint entity and relation extraction configuration. The best configuration found is the long-document transformer variant with Dynamic Task Loss Weighting and partitioned per-module optimizer rates, reaching 49.64 relation F1 on Re-DocRED; the same model with global parameter rates is essentially tied at 49.63. The paper also reports that dynamic task-loss weighting helps when parameters are partitioned and seems to hurt with global rates, and that score rankings are unstable across fidelity levels, with the top low-fidelity configuration dropping to rank 8 at full fidelity. The paper presents these as evidence that the human-guided multi-fidelity process, not just the model changes, contributes to the outcome.

Load-bearing premise

The load-bearing premise is that optimal hyperparameter distributions stay consistent across fidelity levels, so pruning on cheap low-fidelity results will not discard the best full-fidelity configuration.

Editorial extensions

If this is right

  • Multi-fidelity data-subset sprints can shrink the hyperparameter search space before expensive full runs, lowering the compute cost of comparing model variants.
  • Post-training threshold calibration yields noticeable score gains and can be reused at inference time without retraining.
  • Ranking model configurations by low-fidelity scores is unreliable; the winning configuration at low fidelity can be mid-pack at full fidelity.
  • Dynamic task-loss weighting is most beneficial when the optimizer uses partitioned per-module rates, an interaction worth checking in other multitask models.
  • The sprint/thread structure gives practitioners a concrete protocol for mixing automated search with human judgment and checkpoint management.

Reading between the lines

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

  • If the cross-fidelity consistency of hyperparameter distributions holds, the same three-sprint pattern could be applied to other expensive fine-tuning workloads, such as long-document summarization or multimodal extraction, where data-subset fidelities are natural.
  • A direct test would compare top-10 pruning against random pruning under an equal full-fidelity budget; if random pruning matches or beats it, the human-guided pruning step is not the source of the gain.
  • The 49.64 figure is an interim case-study result with nine full-fidelity trials on a single GPU, so readers should treat the paper as an existence proof for the workflow rather than a head-to-head benchmark statement.
  • The per-class relation thresholds suggest a cheap deployment-time adaptation: re-calibrating thresholds on a new validation distribution may recover performance after domain shift without touching the trained weights.
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

4 major / 5 minor

Summary. The manuscript describes an interim case study of a phased, human-guided hyperparameter optimization (HPO) workflow applied to JEREX-L, a collection of variants of the Eberts and Ulges (2021) joint entity and relation extraction model. The workflow consists of three sprints: a low-fidelity single-epoch Gaussian-process sprint used to prune the hyperparameter space, a warmup-phase sprint run under the cyclic learning-rate schedule, and a full-fidelity Optuna/TPE sprint with Hyperband pruning that includes threshold calibration. A second contribution is a threshold meta-learner that tunes mention, coreference, and relation thresholds on validation data using SCut and hill-climbing. The paper reports F1-micro scores for 16 model configurations on Re-DocRED and claims that the Longformer-based model with dynamic task loss weighting (Long-DTL) reaches scores close to 50% and shows the best improvement.

Significance. If the reported results were properly supported, the work would be a useful practitioner-oriented demonstration of multi-fidelity HPO with human-in-the-loop space pruning, and the threshold-calibration component is a concrete, falsifiable mechanism that could benefit joint extraction models. The manuscript is strong on procedural detail: Section 3 and Appendix B describe the sprint protocol, the Hyperband budget parameters, the hill-climbing meta-learner equations, and Algorithm 2 with test-set evaluation. However, the significance is currently undercut by the empirical reporting issues detailed below: the headline numbers are presented ambiguously as validation or test scores, no same-dataset baseline is provided, and the three-run protocol described in Appendix B.3 is not reflected in the reported results. These issues must be resolved before the claimed enhancement of model performance can be accepted.

major comments (4)
  1. [Table 2 and Section 4] The central claim of Section 4, that Long-DTL shows the best improvement and reaches scores close to 50%, depends entirely on the '25E+Cal' column of Table 2. The table title calls these 'F1-micro test scores,' while the caption states that the same column contains 'validation F1 scores' from the full-fidelity training-calibration-validation cycle. The abstract and Section 4 treat these values as test performance on Re-DocRED, and Appendix B.3 describes a separate final test on Dtest after threshold calibration. A validation score used for model selection and threshold calibration does not support the test-set claim. The authors must clarify which quantity is reported, provide the actual test column, or revise the claims in the abstract and Section 4 accordingly.
  2. [Section 4 and Table 2] No same-dataset baseline is reported. The paper does not evaluate the original Eberts-Ulges JEREX model on Re-DocRED, does not report a random-search or default-configuration comparator, and the statement in Section 4 that 'all variants in JEREX-L benefited when compared with the Doc-Red benchmarks shown here' is not supported because no such benchmark numbers appear. Without a baseline, the abstract's claim that the methodology 'enhances machine learning model performance' and the phrase 'best improvement' cannot be evaluated. A minimal fix would be to report the original JEREX configuration trained and tested under the same pipeline on Re-DocRED, together with at least one standard HPO comparator.
  3. [Appendix B.3 and Table 2] Appendix B.3 states that the fine-tuning and evaluation pipeline is repeated three times and that the mean of the three test scores is reported. Table 2 contains single F1 values with no variance or per-run values. This makes it impossible to assess whether the top difference (49.64 vs. 49.63) is meaningful, and it also obscures the stability of the multi-fidelity pruning decisions. The authors should report mean and standard deviation (or per-run values) for at least the top configurations and ideally for all rows.
  4. [Section 3, Hypothesis 2 and Table 2] The multi-fidelity pruning strategy relies on Hypothesis 2, which states that optimal hyperparameter distributions are more consistent across fidelity levels. This assumption is load-bearing because Phase 1 prunes the hyperparameter space to the top 10 low-fidelity iterations, and a mismatch between low- and full-fidelity optima could exclude the best configuration before expensive sprints run. The paper provides no direct evidence for the hypothesis, and the score rankings in Table 2 change substantially across fidelities (for example, Bert-ASL-POS GLOBAL goes from rank 1 at 1E to rank 8 at 25E+Cal). The authors should either verify the assumption directly by comparing hyperparameter distributions of top trials across fidelities, or temper the claims about the reliability of the pruning step.
minor comments (5)
  1. [Section 3, Experiment A description] The text reads 'considering the two transformer architectures, i.e., Bert and Longfromer'; 'Longfromer' should be 'Longformer.'
  2. [Appendix B.1.3] Equations (16) and (17) are identical, which appears to be a typo; one of them should presumably define the per-class relation threshold vector explicitly, since equation (18) then selects per-class indices.
  3. [Algorithm 1] The line 'B = (smax + 1)R Toral Budget' contains a typo: 'Toral' should be 'Total Budget.'
  4. [Appendix B.1.1] The heading 'Mention Autho-threshold' should be 'Mention Auto-threshold.'
  5. [Abstract and Section 1] The paper states that a prototype visualization and space-pruning tool is published, but no repository, URL, or version is given in the text or references; this should be added to make the reproducibility claim concrete.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: threshold calibration and multi-fidelity HPO are standard empirical procedures, not predictions derived from their own inputs.

full rationale

The paper makes no equation-level derivation; it is an empirical HPO case study. Its threshold meta-learner (Appendix B.1.2) optimizes thresholds against validation data through hill-climbing on relation F1 and then applies those thresholds to the held-out test set in Algorithm 2, which is standard practice and not a case of a fitted parameter being renamed as a prediction. The multi-fidelity sprint pipeline (Phase 1 pruning, Phase 2 warmup, Phase 3 cold-primed full-fidelity Optuna study) is a heuristic search procedure; the paper explicitly labels its two consistency assumptions as hypotheses rather than deriving them, and it even reports rank reordering across fidelities in Table 2, so it does not assert that low-fidelity scores equal high-fidelity scores by construction. There are no load-bearing self-citations: the only JEREX citation is Eberts and Ulges (2021), an external baseline, and no prior work of the author is invoked to forbid alternatives or supply uniqueness. The one notable manuscript-level red flag is a reporting inconsistency: Table 2's title says "F1-micro test scores of 16 model configurations," while its caption says the 25E+Cal column contains "validation F1 scores of the full-fidelity 25-epoch training-calibration-validation cycle." The caption is the more specific statement, so the headline 49.64 may lack demonstrated test-set support, and the absence of a same-dataset baseline undermines the word "improvement". Those are correctness and evidence gaps, not circularity: the reported value is an empirical validation evaluation, not a quantity that equals its input by construction. Therefore the paper is self-contained against circularity, and the score is 0.

Assumptions & free parameters 10 free parameters · 6 assumptions · 0 invented entities

The framework's central claims rest primarily on domain assumptions about fidelity transfer, validation generalization, and the benevolence of human pruning, none of which are empirically isolated. There are no formal mathematical axioms beyond the standard surrogate-model assumptions of Bayesian optimization. No new physical entities are posited.

free parameters (10)
  • max_lr (scheduler maximum learning rate) = not reported
    Optimized in Phases 2 and 3 via gp_minimize and Optuna; final values not listed.
  • weight_decay = not reported
    Optimized in all phases; final values not listed.
  • Task loss weights (four) = not reported
    Weights for mention, coreference, entity, and relation losses; optimized as hyperparameters.
  • pct_start (warmup fraction) = bounded between 6 and 8 epochs
    Warmup period is an optimized hyperparameter in Phase 2; exact values not reported.
  • Mention threshold = not reported
    Optimized analytically by SCut and via hill-climbing on validation data.
  • Coreference threshold = not reported
    Hill-climbed on validation data.
  • Relation threshold vector = not reported
    Per-class thresholds optimized by hill-climbing on validation data.
  • Perturbation delta for hill-climbing = 0.05 during training, 0.025 post-training
    Hand-chosen step sizes for threshold perturbations.
  • Pruning margins = 1.5 for log-uniform, 0.01 for uniform
    Hand-chosen expansion factors around top-10 ranges to build pruned HP spaces.
  • Trial counts and budget = 120 / 90 / 9 trials, B=128 epochs
    Hand-set HPO budgets; final Optuna phase allowed only nine trials per configuration.
assumptions (6)
  • domain assumption Optimal hyperparameter distributions are consistent across fidelity levels
    The pruning of HP spaces from low-fidelity sprints presumes that low-fidelity optima remain near full-fidelity optima; the paper labels this Hypothesis 2, not a proven fact.
  • domain assumption Validation-based threshold calibration and model selection do not overfit the validation set
    Thresholds are hill-climbed on validation and best models are selected on validation before final test; no guard against validation overfitting is demonstrated.
  • domain assumption Re-DocRED labels are treated as ground truth
    All scores depend on the correctness of the Re-DocRED annotations; no error analysis or annotation-quality check is included.
  • domain assumption Data subsetting acts as a cross-validation mechanism
    Section 3 and Hypothesis 2 claim rotating data subsets provides robustness; only anecdotal support from Table 2 rank changes is given.
  • ad hoc to paper The pruning margin factors (1.5, 0.01) and top-10 selection retain the true optimum
    These values are chosen by hand without sensitivity analysis; if the true optimum lies outside the pruned ranges, later sprints cannot recover it.
  • ad hoc to paper Human guidance improves rather than biases the search
    No controlled experiment compares human-guided vs fully automated HPO; human pruning decisions are not logged or quantified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Interim Report on Human-Guided Adaptive Hyperparameter Optimization with Multi-Fidelity Sprints." pith.science (2026). https://pith.science/paper/WNI3Y5UZ

@misc{pith2026250509792,
  author       = {Pith},
  title        = {Pith review of: Interim Report on Human-Guided Adaptive Hyperparameter Optimization with Multi-Fidelity Sprints},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WNI3Y5UZ}},
  note         = {Machine review of arXiv:2505.09792}
}
read the original abstract

This case study applies a phased hyperparameter optimization process to compare multitask natural language model variants that utilize multiphase learning rate scheduling and optimizer parameter grouping. We employ short, Bayesian optimization sessions that leverage multi-fidelity, hyperparameter space pruning, progressive halving, and a degree of human guidance. We utilize the Optuna TPE sampler and Hyperband pruner, as well as the Scikit-Learn Gaussian process minimization. Initially, we use efficient low-fidelity sprints to prune the hyperparameter space. Subsequent sprints progressively increase their model fidelity and employ hyperband pruning for efficiency. A second aspect of our approach is using a meta-learner to tune threshold values to resolve classification probabilities during inference. We demonstrate our method on a collection of variants of the 2021 Joint Entity and Relation Extraction model proposed by Eberts and Ulges.

Figures

Figures reproduced from arXiv: 2505.09792 by the authors.

Figure 1
Figure 1. General framework for HPO experiments. Gray boxes denote external resources that remain fixed. Each “Sprint" uses a model checkpoint and HP space definition to execute a given number of trials (iterations). Each trial entails fine-tuning the model at a chosen fi￾delity level with a proposed set of HPs suggested by a Bayesian, history-aware HPO process. The trial HPs and score results are analyzed by a human who can … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 19 canonical work pages

  1. [1]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

  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]

    Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. 2019. O ptuna: A next-generation hyperparameter optimization framework. In The 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 2623--2631

    2019

  4. [4]

    Peters, and Arman Cohan

    Iz Beltagy, Matthew E. Peters, and Arman Cohan. 2020. Longformer: The long-document transformer. arXiv:2004.05150

    arXiv 2020

  5. [5]

    Emanuel Ben-Baruch, Tal Ridnik, Nadav Zamir, Asaf Noy, Itamar Friedman, Matan Protter, and Lihi Zelnik-Manor. 2020. http://arxiv.org/abs/2009.14119 Asymmetric loss for multi-label classification

    arXiv 2020

  6. [6]

    James Bergstra, R \'e mi Bardenet, Yoshua Bengio, and Bal \'a zs K \'e gl. 2011. https://api.semanticscholar.org/CorpusID:11688126 Algorithms for hyper-parameter optimization . In Neural Information Processing Systems

    work page 2011

  7. [7]

    James Bergstra and Yoshua Bengio. 2012. https://api.semanticscholar.org/CorpusID:15700257 Random search for hyper-parameter optimization . J. Mach. Learn. Res., 13:281--305

    work page 2012

  8. [8]

    Bernd Bischl, Martin Binder, Michel Lang, Tobias Pielok, Jakob Richter, Stefan Coors, Janek Thomas, Theresa Ullmann, Marc Becker, Anne-Laure Boulesteix, Difan Deng, and Marius Lindauer. 2021. http://arxiv.org/abs/2107.05847 Hyperparameter optimization: Foundations, algorithms, best practices and open challenges

    arXiv 2021

Show all 36 references
  1. [9]

    Scikit-Optimize Contributors. 2024. Scikit-optimize: Sequential model-based optimization with a `scipy.optimize` interface. https://github.com/scikit-optimize/scikit-optimize. Accessed: 2024-02-15

    2024

  2. [10]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. https://api.semanticscholar.org/CorpusID:52967399 Bert: Pre-training of deep bidirectional transformers for language understanding . In North American Chapter of the Association for Computational Linguistics

    2019

  3. [11]

    Markus Eberts and Adrian Ulges. 2021. https://doi.org/10.18653/v1/2021.eacl-main.319 An end-to-end model for entity-level relation extraction using multi-instance learning . In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Ling...

    2021 doi

  4. [12]

    Haytame Fallah, Patrice Bellot, Emmanuel Bruno, and Elisabeth Murisasco. 2022. Adapting transformers for multi-label text classification. In Joint Conference of the Information Retrieval Communities in Europe

    2022

  5. [13]

    Rong-En Fan and Chih-Jen Lin. 2007. A study on threshold selection for multi-label classification

    2007

  6. [14]

    Peter I. Frazier. 2018. http://arxiv.org/abs/1807.02811 A tutorial on bayesian optimization

    2018 arXiv

  7. [15]

    Flach, and C

    Jos \'e Hern \'a ndez-Orallo, Peter A. Flach, and C. Ferri. 2012. A unified view of performance metrics: translating threshold choice into expected classification loss. J. Mach. Learn. Res., 13:2813--2869

    2012

  8. [16]

    Matthew Honnibal and Mark Johnson. 2015. https://doi.org/10.18653/v1/D15-1162 An improved non-monotonic transition system for dependency parsing . In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 1373--1378, Lisbon, Portugal. Ass...

    2015 doi

  9. [17]

    Matthew Honnibal and Ines Montani. 2017. spacy 2: Natural language understanding with bloom embeddings, convolutional neural networks and incremental parsing. To appear

    2017

  10. [18]

    Kevin Jamieson and Ameet Talwalkar. 2016. https://proceedings.mlr.press/v51/jamieson16.html Non-stochastic best arm identification and hyperparameter optimization . In Proceedings of the 19th International Conference on Artificial Intelligence and Statistics, volume 51 of Proc...

    2016

  11. [19]

    Kirthevasan Kandasamy, Gautam Dasarathy, Jeff Schneider, and Barnab \'a s P \'o czos. 2017. https://proceedings.mlr.press/v70/kandasamy17a.html Multi-fidelity B ayesian optimisation with continuous approximations . In Proceedings of the 34th International Conference on Machine...

    2017

  12. [20]

    Alex Kendall, Yarin Gal, and Roberto Cipolla. 2018. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7482--7491

    2018

  13. [21]

    Lisha Li, Kevin Jamieson, Giulia DeSalvo, Afshin Rostamizadeh, and Ameet Talwalkar. 2018. http://arxiv.org/abs/1603.06560 Hyperband: A novel bandit-based approach to hyperparameter optimization

    2018 arXiv

  14. [22]

    Girshick, Kaiming He, and Piotr Doll \'a r

    Tsung-Yi Lin, Priya Goyal, Ross B. Girshick, Kaiming He, and Piotr Doll \'a r. 2020. Focal loss for dense object detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42:318--327

    2020

  15. [23]

    Ilya Loshchilov and Frank Hutter. 2016. https://api.semanticscholar.org/CorpusID:14337532 Sgdr: Stochastic gradient descent with warm restarts . arXiv: Learning

    2016

  16. [24]

    Bertil Matérn. 1960. Spatial Variation

    1960

  17. [25]

    Kourosh Meshgi, Maryam Sadat Mirzaei, and Satoshi Sekine. 2022. https://aclanthology.org/2022.wassa-1.8 Uncertainty regularized multi-task learning . In Proceedings of the 12th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis , pages 78--...

    2022

  18. [26]

    Jonas Mockus. 1989. https://api.semanticscholar.org/CorpusID:120322743 Bayesian approach to global optimization: Theory and applications

    1989

  19. [27]

    Matthias Poloczek, Jialei Wang, and P. Frazier. 2016. https://api.semanticscholar.org/CorpusID:30747438 Multi-information source optimization . In Neural Information Processing Systems

    2016

  20. [28]

    C. E. Rasmussen and C. K. I. Williams. 2006. Gaussian Processes for Machine Learning. MIT Press

    2006

  21. [29]

    Jasper Snoek, Hugo Larochelle, and Ryan P. Adams. 2012. http://arxiv.org/abs/1206.2944 Practical bayesian optimization of machine learning algorithms

    2012 arXiv

  22. [30]

    Qingyu Tan, Lu Xu, Lidong Bing, Hwee Tou Ng, and Sharifah Mahani Aljunied. 2022. https://arxiv.org/abs/2205.12696 Revisiting docred – addressing the false negative problem in relation extraction . In Proceedings of EMNLP

    2022 arXiv

  23. [31]

    Shuhei Watanabe. 2023. https://api.semanticscholar.org/CorpusID:258291728 Tree-structured parzen estimator: Understanding its algorithm components and their roles for better empirical performance . ArXiv, abs/2304.11127

    2023 arXiv

  24. [32]

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mari...

    2020 doi

  25. [33]

    Yiming Yang. 2001. A study of thresholding strategies for text categorization. In Annual International ACM SIGIR Conference on Research and Development in Information Retrieval

    2001

  26. [34]

    Yuan Yao, Deming Ye, Peng Li, Xu Han, Yankai Lin, Zhenghao Liu, Zhiyuan Liu, Lixin Huang, Jie Zhou, and Maosong Sun. 2019. Docred: A large-scale document-level relation extraction dataset. ArXiv, abs/1906.06127

    2019 arXiv

  27. [35]

    Sungduk Yu, Mike Pritchard, Po‐Lun Ma, Balwinder Singh, and Sammuel Ramos da Silva. 2023. https://api.semanticscholar.org/CorpusID:256662729 Two-step hyperparameter optimization method: Accelerating hyperparameter search by using a fraction of a training dataset . ArXiv, abs/2...

    2023 arXiv

  28. [36]

    Yuhao Zhang, Victor Zhong, Danqi Chen, Gabor Angeli, and Christopher D. Manning. 2017. https://nlp.stanford.edu/pubs/zhang2017tacred.pdf Position-aware attention and supervised data improve slot filling . In Proceedings of the 2017 Conference on Empirical Methods in Natural La...

    2017

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.