Pith. sign in

REVIEW 5 major objections 5 minor 46 references

Towards Better Multi-task Learning: A Framework for Optimizing Dataset Combinations in Large Language Models

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

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

Pith's one-line read Paper claims that an iterative neural-network-guided search finds better multi-task dataset combinations than single-task baselines using only a handful of fine-tuning runs.

desk verdict A sensible surrogate-based search idea that is undermined by internally inconsistent reported numbers. read the letter →

arxiv 2412.11455 v1 pith:B6KZEJUL submitted 2024-12-16 cs.CL cs.AI

Zaifu Zhan , Rui Zhang This is my paper · ORCID
classification cs.CLcs.AI
keywords multi-tasklearningdatasetcombinationselectionlargelanguagemodelsneuralnetworksurrogateinstructiontuningbiomedicalNLPF1-scoreoptimizationiterativesearch
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

Multi-task learning for large language models should benefit from choosing which datasets to train on together, but the obvious way to choose—fine-tune on every subset and keep the best—costs exponentially many runs. This paper proposes a feedback loop in which a small neural network learns to predict a dataset combination's F1 score from which datasets are included, and each iteration fine-tunes the LLM only on the combination the network currently ranks best. Across 12 biomedical datasets covering named entity recognition, relation extraction, event extraction, and text classification, the loop finds combinations that beat single-task training in a dozen or so iterations, rather than trying the 2,048 possible auxiliary combinations per target. The central finding is that the surrogate-guided search can improve tasks that look unpromising from a human perspective, such as three NER/TC datasets where every hand-chosen combination had decreased performance. A sympathetic reader would take this as evidence that learned dataset-combination search is a viable route to unlocking more of multi-task learning's benefit.

What carries the argument

The central object is the surrogate network: a two-layer neural network with 12 binary inputs (one per dataset, set to 1 if the dataset is included) and a single regression output, the predicted F1 score. It is trained on the data table of tested combinations and their measured F1 scores, retrained after every iteration, and then used to enumerate all $2^{11} = 2048$ auxiliary combinations for the target dataset and pick the predicted best. This surrogate is what converts dataset selection from an exponential number of fine-tuning runs into a feedback loop: test, retrain, repredict, repeat, terminating when the predicted best has already been tested. The same mechanism can in principle wrap any LLM, any dataset pool, and any scoring metric, which is what carries the generality claim.

What would settle it

Using the paper's logged combination-score tables, train the same two-layer surrogate on a random half of the combinations, then measure how well its predicted F1 ranks the held-out half (e.g., Spearman correlation); if the ranking is not significantly better than random, or if a random-search loop given the same fine-tuning budget matches the surrogate-guided loop's best F1, the framework's search-efficiency claim would be refuted.

Watch

Extended reading notes

Core claim

The paper's claim is that a cheap two-layer neural network can act as a surrogate for expensive fine-tuning and reliably steer an iterative search toward high-F1 dataset combinations. The surrogate takes a 12-bit vector indicating which datasets are in a combination and outputs a predicted F1 score; after each real fine-tuning run, it is retrained on the growing table of observed combination-score pairs, then used to score all remaining combinations so the next run tests the top-ranked one. The loop stops when the network's current top pick is a combination already tested. On the RE-task data, 30 of 36 combinations improved BioRED while all tested combinations hurt GIT, yet the framework still found an improved combination for GIT; across the full experiment it found gains for three datasets (BC2GM, BC5CDR, HealthAdvice) where every human-screened combination had failed. The authors conclude that the framework is model-, dataset-, and domain-independent and greatly improves search efficiency over brute-force enumeration.

Load-bearing premise

The framework assumes that a small neural network trained on a limited set of observed combination-F1 pairs can rank untested combinations accurately enough to guide the search; if that ranking is poor, the loop degenerates to near-random exploration and the efficiency and optimality claims no longer follow.

Editorial extensions

If this is right

  • For a fixed target dataset, the framework reaches a near-best combination in roughly a dozen fine-tuning runs instead of trying all $2^{11}=2048$ auxiliary combinations.
  • Datasets with low single-task F1 (roughly below 60 on these biomedical tasks) tend to gain the most from multi-task training, while already-strong tasks gain little or nothing.
  • Useful dataset combinations can be invisible to human judgment: three datasets improved only after the surrogate proposed combinations that a human would not have tried.
  • Because the surrogate only sees binary dataset indicators and a score, the same loop transfers to other LLMs, dataset pools, and metrics without architectural changes.
  • The stopping rule means the search ends when the surrogate's predicted best candidate has already been evaluated, which is an automatic budget cap rather than a fixed number of runs.

Reading between the lines

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

  • A natural extension would be to replace the final fine-tune-and-evaluate step with a cheaper proxy (smaller model, fewer steps, or few-shot inference) to lower cost further; the paper itself does not test this substitution.
  • The bit-vector encoding treats each dataset as an independent binary switch; adding pairwise interaction terms to the surrogate could capture synergy and may improve ranking accuracy for tasks where combinations matter.
  • The stopping rule stops when the surrogate predicts an already-tested combination, which conflates 'no better combination exists' with 'the surrogate's confidence is exhausted'; a more robust stop would require independent validation on held-out combinations.
  • The observation that weak baselines benefit most suggests a usable pre-screening heuristic—spend search budget on low-F1 tasks—that the paper reports as a pattern but does not elevate to a decision rule.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes an iterative framework for selecting auxiliary training datasets in multi-task instruction tuning of LLMs. A neural-network regression surrogate is trained on evaluated dataset-combination F1 scores; after each fine-tuning run the surrogate is retrained, all 2^11 auxiliary combinations are enumerated at low cost, and the predicted best combination is tested until the surrogate repeats an already-tested combination. Experiments are reported on 12 biomedical datasets across NER, RE, EE, and TC with LLaMA-3-8B. The paper claims that the method finds F1-improving combinations with far fewer fine-tuning runs than brute-force enumeration, including for tasks where human intuition sees no MTL benefit.

Significance. If the claims were backed by reliable measurements, the framework would be a useful practical contribution to MTL dataset selection: it is model- and dataset-agnostic in principle, and surrogate-guided search has the potential to reduce computational cost substantially. The paper also deserves credit for evaluating on 12 datasets and for formulating the selection problem as an iterative, feedback-driven search. However, the current evidence is not trustworthy. The central empirical claim rests on F1 curves that contradict the corresponding tables and text, and the surrogate's ranking ability is never validated against held-out predictions or a random-search baseline. The scientific contribution is therefore not established.

major comments (5)
  1. [§5.2, Fig. 4(b), Table 1] The text states that for DDI 'the framework cannot find better combinations and stops soon,' and Table 1 shows only 4 of 36 DDI combinations improving over the 71.10 baseline, with the best at 73.94; nevertheless Fig. 4(b) plots best F1 rising from about 71 to about 78 over roughly 18 iterations. These are incompatible, and because Fig. 4 is the primary evidence for the RE task, the central claim is not supported for DDI.
  2. [§5.3, Fig. 5(b), Fig. 5(e)] The text reports that GENIA2013 improved 'to around 58,' while the y-axis of Fig. 5(b) ends at 55 and the plotted values do not reach 58. Similarly, Fig. 5(e) for BC4CHEMD shows a y-axis extending past 100 and F1 values above 100, which is impossible for F1. These inconsistencies make the figures unusable as quantitative evidence for the framework's effectiveness.
  3. [§3, §4.4, §5.3] The framework's stopping rule and efficiency claim depend on the surrogate's ability to predict untested combinations, but the paper provides no held-out evaluation of the surrogate, no correlation analysis, and no random-search or other model-based baseline. Without such a baseline, the claim that the method 'significantly improved efficiency' relative to brute force is not established; the loop could simply be performing noisy exploration.
  4. [§4.5, §5.1] All results are reported without seeds, error bars, or repeated runs, and the best checkpoint is selected from checkpoints saved every 1000 steps. Given that many reported F1 differences are small (e.g., BioRED improvements of about 0.5–1 point in Table 1), fine-tuning noise could dominate the reported gains, and the paper should at minimum report variance or significance tests.
  5. [§5.3, Figs. 5(d), 5(f), 5(h), Tables 2–4] The text says that for NER and TC 'every combination we attempted resulted in a performance decline' and that the framework later found improvements for BC2GM, BC5CDR, and HealthAdvice; however, the appendix tables mark the baselines as the best among the data-preparation runs, and no tabled framework-iteration F1 values are provided for these three datasets. The reader therefore cannot verify the claimed 'unpromising tasks' improvements from the reported data.
minor comments (5)
  1. [§5.2] The sentence 'we enumerate 211 = 2048kinds of combinations' contains a typo: it should read '2^{11} = 2048,' and 'nerual' should be 'neural.'
  2. [Fig. 3] The prompt templates contain typos: 'You need to to accurately recognize' and 'sentece' should be corrected.
  3. [Title page] The correspondence email address is listed as zhan1386@umn.edu for Rui Zhang, while the affiliations list Zaifu Zhan as the first author; this should be checked for accuracy.
  4. [§4.2] The model name is written 'LLama3-8B'; the standard spelling is 'LLaMA-3-8B,' and the name should be used consistently.
  5. [Tables 1–4] The appendix tables are dense and repeat checkmark patterns in a way that is hard to scan; a compact representation of the selected datasets, such as bit strings, would improve readability.

Circularity Check

2 steps flagged · score 8.0 of 10

The framework's claimed improvements are the surrogate's own fitted predictions, and its stopping rule equates optimality with the surrogate's self-agreement.

  1. fitted input called prediction [Section 4.4 and Section 5.3 (Figures 5(d), 5(f), 5(h)) vs. Tables 2 and 4]
    "The output of the neural network is the score we aim to predict, such as the F1 score in our experiments. ... For the NER and TC tasks, the six datasets involved all achieved their best F1 scores with the baseline settings, and every combination we attempted resulted in a performance decline. ... Surprisingly, as shown in Figures 5(d), 5(f), and 5(h), our framework identified promising combinations that led to significant improvements for three of the datasets."

    The neural network is regressed on the measured F1 scores (Section 4.4), and the 'Best F1 score' curves in Figures 4-5 are then presented as the framework's results. These curves cannot be measured values: Fig. 5(e) plots F1 above 100, and the text/tables say BC2GM, BC5CDR, DDI, and HealthAdvice baselines are best or near-best while the figures show large rises. The 'improvements' are therefore the fitted surrogate's own unrestricted regression outputs, not evaluations of fine-tuned LLMs. Since the NN is trained on the same F1 scores that define success, reporting its predictions as 'better combinations' makes the claimed gain an artifact of the fit rather than an empirical discovery.

  2. self definitional [Section 3 (loop termination) and Section 5.3 (final summary)]
    "The process continues until the neural network predicts an optimal combination that has already been tested, at which point the loop terminates. ... However, in most cases, we were able to find relatively optimal combinations within just a dozen iterations, and the framework predicted that no better combinations existed."

    The termination rule defines the target ('an optimal combination') as whatever the neural network's argmax happens to be and requires only that this argmax has already been tested. The paper then reports as the outcome that it 'found relatively optimal combinations' and that 'the framework predicted that no better combinations existed.' The conclusion restates the surrogate's own prediction; there is no external optimality check, random-search baseline, or held-out validation of the surrogate (Section 7 only notes single-model and no grid search). The efficiency/optimality claim is thus true by construction of the stopping rule: when the NN predicts its own previously chosen best, the loop stops and 'optimal' is asserted.

full rationale

The paper does contain genuinely measured F1 values in the data-preparation tables (e.g., BioRED 53.52 to 64.05, GENIA2011 to 57.08), so not every claim is circular. The framework's distinctive contribution—iterative NN-guided search that 'effectively identifies better combinations'—is, however, supported only by the surrogate's own predicted-F1 curves and by a stopping rule that equates the surrogate's self-agreement with optimality. No seeds, error bars, random-search baseline, or held-out validation of the surrogate are given; Section 7 lists only computational limitations. The score is 8 because the central demonstration reduces to the fitted model's outputs, even though some ancillary measured gains exist. Self-citation is not load-bearing in this paper.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The framework rests on the assumption that a simple MLP over binary dataset indicators can predict MTL F1 scores well enough to guide search. This premise is neither validated nor compared against a random-search baseline. The paper also assumes a single fine-tuning run per combination is representative. No new physical or conceptual entities are introduced.

free parameters (3)
  • Neural network hyperparameters
    The two-layer MLP architecture, learning rate, epochs, hidden size, and optimizer are not specified, so the surrogate model is underdetermined and not reproducible.
  • 48-hour exploration budget = 48 hours
    The framework stops after a fixed wall-clock time, an arbitrary choice that determines how many combinations are explored and directly affects the reported efficiency.
  • 5000 training steps = 5000
    All fine-tuning uses 5000 steps with checkpoints every 1000 steps, chosen for fairness but not justified; the best-checkpoint selection adds another hand-chosen element.
assumptions (4)
  • domain assumption The binary inclusion vector of selected datasets is a sufficient representation for predicting MTL performance
    The neural network takes only the 12 binary indicators and outputs a score; any interaction depending on dataset size, domain, or similarity must be captured implicitly, which the paper does not justify.
  • domain assumption A surrogate trained on a small set of combinations can rank untested combinations accurately enough to guide search
    This is the load-bearing premise of the framework; the paper never validates it with held-out predictions or comparisons to random search.
  • domain assumption Equal sampling from each dataset and fixed 5000 steps make all tested combinations comparable
    The paper fixes training steps and sampling to be fair, but this also assumes the optimal training budget for each combination is the same.
  • domain assumption Reported micro F1 scores are stable across runs
    Each combination is trained once; no variance estimates are given, yet F1 differences of fractions of a point are treated as meaningful improvements.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Better Multi-task Learning: A Framework for Optimizing Dataset Combinations in Large Language Models." pith.science (2026). https://pith.science/paper/B6KZEJUL

@misc{pith2026241211455,
  author       = {Pith},
  title        = {Pith review of: Towards Better Multi-task Learning: A Framework for Optimizing Dataset Combinations in Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B6KZEJUL}},
  note         = {Machine review of arXiv:2412.11455}
}
read the original abstract

To efficiently select optimal dataset combinations for enhancing multi-task learning (MTL) performance in large language models, we proposed a novel framework that leverages a neural network to predict the best dataset combinations. The framework iteratively refines the selection, greatly improving efficiency, while being model-, dataset-, and domain-independent. Through experiments on 12 biomedical datasets across four tasks - named entity recognition, relation extraction, event extraction, and text classification-we demonstrate that our approach effectively identifies better combinations, even for tasks that may seem unpromising from a human perspective. This verifies that our framework provides a promising solution for maximizing MTL potential.

Figures

Figures reproduced from arXiv: 2412.11455 by the authors.

Figure 1
Figure 1. Comparison of (a) single-task learning and (b) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the proposed framework. The combination generator produces dataset combinations to be instruction-tuned on an LLM. In each iteration, the neural network identifies and refines the best combination until no further improvements can be made. nations that are unlikely to yield good results, or directly predict the best combination, it would save significant time and computational power. Driven by this idea,… view at source ↗
Figure 3
Figure 3. Training and testing prompt template. our framework comprehensively. For Event Extraction task, we utilize the fol￾lowing datasets: • PHEE (Sun et al., 2022): A comprehen￾sive dataset providing high-accuracy annota￾tions for 2 events: ’Adverse_event’, ’Poten￾tial_therapeutic_event’. • GENIA2011 (Kim et al., 2011): A bench￾mark EE dataset from the biomedical domain. The event types include ’Regulation’, ’Localization… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Framework results for datasets in the RE task. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Finding best F1 results using proposed framework for EE (GENIA2011, GENIA2013, PHEE), NER [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 17 canonical work pages

  1. [1]

    Joachim Bingel and Anders S gaard. 2017. Identifying beneficial task relations for multi-task learning in deep neural networks. arXiv preprint arXiv:1702.08303

    arXiv 2017

  2. [2]

    Shijie Chen, Yu Zhang, and Qiang Yang. 2024. Multi-task learning in natural language processing: An overview. ACM Computing Surveys, 56(12):1--32

    work page 2024

  3. [3]

    Michael Crawshaw. 2020. Multi-task learning with deep neural networks: A survey. arXiv preprint arXiv:2009.09796

    arXiv 2020

  4. [4]

    Franck Dernoncourt and Ji Young Lee. 2017. Pubmed 200k rct: a dataset for sequential sentence classification in medical abstracts. arXiv preprint arXiv:1710.06071

    arXiv 2017

  5. [5]

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

    arXiv 2018

  6. [6]

    John C Doyle, Bruce A Francis, and Allen R Tannenbaum. 2013. Feedback control theory. Courier Corporation

    work page 2013

  7. [7]

    Chris Fifty, Ehsan Amid, Zhe Zhao, Tianhe Yu, Rohan Anil, and Chelsea Finn. 2021. Efficiently identifying task groupings for multi-task learning. Advances in Neural Information Processing Systems, 34:27503--27516

    2021

  8. [8]

    Han Guo, Ramakanth Pasunuru, and Mohit Bansal. 2019. Autosem: Automatic task selection and mixing in multi-task learning. arXiv preprint arXiv:1904.04153

    arXiv 2019

Show all 46 references
  1. [9]

    Harsha Gurulingappa, Abdul Mateen Rajput, Angus Roberts, Juliane Fluck, Martin Hofmann-Apitius, and Luca Toldo. 2012. https://doi.org/10.1016/j.jbi.2012.04.008 Development of a benchmark corpus to support the automatic extraction of drug-related adverse effects from medical ca...

    2012 doi

  2. [10]

    Jin-Dong Kim, Yue Wang, Toshihisa Takagi, and Akinori Yonezawa. 2011. Overview of genia event task in bionlp shared task 2011. In Proceedings of BioNLP shared task 2011 workshop, pages 7--15

    2011

  3. [11]

    Jin-Dong Kim, Yue Wang, and Yamamoto Yasunori. 2013. The genia event extraction shared task, 2013 edition-overview. In Proceedings of the BioNLP Shared Task 2013 Workshop, pages 8--15

    2013

  4. [12]

    Bahare Kiumarsi, Frank L Lewis, Hamidreza Modares, Ali Karimpour, and Mohammad-Bagher Naghibi-Sistani. 2014. Reinforcement q-learning for optimal tracking control of linear discrete-time systems with unknown dynamics. Automatica, 50(4):1167--1175

    2014

  5. [13]

    Sotiris B Kotsiantis, Ioannis Zaharakis, P Pintelas, et al. 2007. Supervised machine learning: A review of classification techniques. Emerging artificial intelligence applications in computer engineering, 160(1):3--24

    2007

  6. [14]

    Martin Krallinger, Obdulia Rabal, Florian Leitner, Miguel Vazquez, David Salgado, Zhiyong Lu, Robert Leaman, Yanan Lu, Donghong Ji, Daniel M Lowe, et al. 2015. The chemdner corpus of chemicals and drugs and its annotation principles. Journal of cheminformatics, 7:1--17

    2015

  7. [15]

    Jiao Li, Yueping Sun, Robin J Johnson, Daniela Sciaky, Chih-Hsuan Wei, Robert Leaman, Allan Peter Davis, Carolyn J Mattingly, Thomas C Wiegers, and Zhiyong Lu. 2016. Biocreative v cdr task corpus: a resource for chemical disease relation extraction. Database, 2016

    2016

  8. [16]

    Mingchen Li, Ming Chen, Huixue Zhou, and Rui Zhang. 2023. Petailor: Improving large language model by tailored chunk scorer in biomedical triple extraction. arXiv preprint arXiv:2310.18463

    2023 arXiv

  9. [17]

    Pengfei Liu, Weizhe Yuan, Jinlan Fu, Zhengbao Jiang, Hiroaki Hayashi, and Graham Neubig. 2023. Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. ACM Computing Surveys, 55(9):1--35

    2023

  10. [18]

    Ling Luo, Po-Ting Lai, Chih-Hsuan Wei, Cecilia N Arighi, and Zhiyong Lu. 2022. Biored: a rich biomedical relation extraction dataset. Briefings in Bioinformatics, 23(5):bbac282

    2022

  11. [19]

    Jiaqi Ma, Zhe Zhao, Xinyang Yi, Jilin Chen, Lichan Hong, and Ed H Chi. 2018. Modeling task relationships in multi-task learning with multi-gate mixture-of-experts. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, pages 1930--1939

    2018

  12. [20]

    Ishan Misra, Abhinav Shrivastava, Abhinav Gupta, and Martial Hebert. 2016. Cross-stitch networks for multi-task learning. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3994--4003

    2016

  13. [21]

    Rajvardhan Patil, Sorio Boit, Venkat Gudivada, and Jagadeesh Nandigam. 2023. A survey of text representation and embedding techniques in nlp. IEEE Access, 11:36120--36146

    2023

  14. [22]

    Yada Pruksachatkun, Jason Phang, Haokun Liu, Phu Mon Htut, Xiaoyi Zhang, Richard Yuanzhe Pang, Clara Vania, Katharina Kann, and Samuel R Bowman. 2020. Intermediate-task transfer learning with pretrained models for natural language understanding: When and why does it work? arXi...

    2020 arXiv

  15. [23]

    Yunjia Qi, Hao Peng, Xiaozhi Wang, Bin Xu, Lei Hou, and Juanzi Li. 2024. Adelie: Aligning large language models on information extraction. arXiv preprint arXiv:2405.05008

    2024 arXiv

  16. [24]

    Alec Radford, Karthik Narasimhan, Tim Salimans, Ilya Sutskever, et al. 2018. Improving language understanding by generative pre-training

    2018

  17. [25]

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. 2019. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9

    2019

  18. [26]

    Oscar Sainz, Iker Garc \' a-Ferrero, Rodrigo Agerri, Oier Lopez de Lacalle, German Rigau, and Eneko Agirre. 2023. Gollie: Annotation guidelines improve zero-shot information-extraction. arXiv preprint arXiv:2310.03668

    2023 arXiv

  19. [27]

    Isabel Segura-Bedmar, Paloma Mart \' nez, and Mar \' a Herrero-Zazo. 2013. Semeval-2013 task 9: Extraction of drug-drug interactions from biomedical texts (ddiextraction 2013). In Second Joint Conference on Lexical and Computational Semantics (* SEM), Volume 2: Proceedings of ...

    2013

  20. [28]

    Ozan Sener and Vladlen Koltun. 2018. Multi-task learning as multi-objective optimization. Advances in neural information processing systems, 31

    2018

  21. [29]

    Larry Smith, Lorraine K Tanabe, Rie Johnson nee Ando, Cheng-Ju Kuo, I-Fang Chung, Chun-Nan Hsu, Yu-Shi Lin, Roman Klinger, Christoph M Friedrich, Kuzman Ganchev, et al. 2008. Overview of biocreative ii gene mention recognition. Genome biology, 9:1--19

    2008

  22. [30]

    Trevor Standley, Amir Zamir, Dawn Chen, Leonidas Guibas, Jitendra Malik, and Silvio Savarese. 2020. Which tasks should be learned together in multi-task learning? In International conference on machine learning, pages 9120--9132. PMLR

    2020

  23. [31]

    Zhaoyue Sun, Jiazheng Li, Gabriele Pergola, Byron C Wallace, Bino John, Nigel Greene, Joseph Kim, and Yulan He. 2022. Phee: A dataset for pharmacovigilance event extraction from text. arXiv preprint arXiv:2210.12560

    2022 arXiv

  24. [32]

    Kim-Han Thung and Chong-Yaw Wee. 2018. A brief review on multi-task learning. Multimedia Tools and Applications, 77(22):29705--29725

    2018

  25. [33]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems, 30

    2017

  26. [34]

    Chenguang Wang, Xiao Liu, Zui Chen, Haoyun Hong, Jie Tang, and Dawn Song. 2022 a . Deepstruct: Pretraining of language models for structure prediction. arXiv preprint arXiv:2205.10475

    2022 arXiv

  27. [35]

    Haoxiang Wang, Han Zhao, and Bo Li. 2021. Bridging multi-task learning and meta-learning: Towards efficient training and effective adaptation. In International conference on machine learning, pages 10991--11002. PMLR

    2021

  28. [36]

    Xiao Wang, Weikang Zhou, Can Zu, Han Xia, Tianze Chen, Yuansen Zhang, Rui Zheng, Junjie Ye, Qi Zhang, Tao Gui, et al. 2023. Instructuie: Multi-task instruction tuning for unified information extraction. arXiv preprint arXiv:2304.08085

    2023 arXiv

  29. [37]

    Xingyao Wang, Sha Li, and Heng Ji. 2022 b . Code4struct: Code generation for few-shot event structure prediction. arXiv preprint arXiv:2210.12810

    2022 arXiv

  30. [38]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824--24837

    2022

  31. [39]

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Rémi 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

  32. [40]

    Gokul Yenduri, M Ramalingam, G Chemmalar Selvi, Y Supriya, Gautam Srivastava, Praveen Kumar Reddy Maddikunta, G Deepti Raj, Rutvij H Jhaveri, B Prabadevi, Weizheng Wang, et al. 2024. Gpt (generative pre-trained transformer)--a comprehensive review on enabling technologies, pot...

    2024

  33. [41]

    Bei Yu, Yingya Li, and Jun Wang. 2019. https://doi.org/10.18653/v1/D19-1473 Detecting causal language use in science findings . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Langu...

    2019 doi

  34. [42]

    Shengyu Zhang, Linfeng Dong, Xiaoya Li, Sen Zhang, Xiaofei Sun, Shuhe Wang, Jiwei Li, Runyi Hu, Tianwei Zhang, Fei Wu, et al. 2023. Instruction tuning for large language models: A survey. arXiv preprint arXiv:2308.10792

    2023

  35. [43]

    Yu Zhang and Qiang Yang. 2018. An overview of multi-task learning. National Science Review, 5(1):30--43

    2018

  36. [44]

    Yu Zhang and Qiang Yang. 2021. A survey on multi-task learning. IEEE transactions on knowledge and data engineering, 34(12):5586--5609

    2021

  37. [45]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  38. [46]

    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 gl...

Pith tools

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