Pith. sign in

REVIEW 4 major objections 4 minor 44 references

Fine-tuning a small open-source language model on GPT-4-generated synthetic tasks with reasoning traces yields robust zero-shot job understanding.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-02 10:27 UTC pith:JWQPYOIE

load-bearing objection A credible industrial system paper whose central zero-shot claim rests on an unverified data-cleaning assumption; worth reviewing but needs ablations and a contamination check. the 4 major comments →

arxiv 2607.24783 v1 pith:JWQPYOIE submitted 2026-06-22 cs.AI

Unified Semantic Modeling Framework for Large-Scale Job Understanding at LinkedIn

Dan Xu , Baofen Zheng , Jianqiang Shen , Qi Xiao , Benjamin Hoan Le , Wen Pu , Saurabh Gupta , Ran Zhou
show 6 more authors
Neha Saraf Alice Leung Qianqi Shen Liangjie Hong Jingwei Wu Wenjing Zhang
This is my paper
classification cs.AI
keywords job understandingsmall language modelssynthetic data generationreasoning traceszero-shot generalizationLoRA adaptersattribute groupingentity extraction
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper claims that a small language model (SLM), fine-tuned on GPT-4-generated synthetic job-understanding tasks that include labels and reasoning traces, gains strong zero-shot generalization across job attributes—so a single base model can classify and extract attributes from job postings without per-attribute training. It then adds a multi-adapter design with semantics-aware attribute grouping, letting business-critical attributes be fine-tuned with roughly 300 annotated samples per task via lightweight LoRA adapters on the frozen base model. The authors report that this synthetic-first approach outperforms a model fine-tuned on existing human-annotated attribute data (e.g., nurse shift precision/recall 93/85 vs. 28/21) and beats legacy per-attribute models on occupation, seniority, and workplace type, with online A/B tests showing engagement and relevance gains. If it works at scale, one small model with a handful of adapters could replace many bespoke classifiers, lowering serving cost and maintenance.

Core claim

On the paper's own terms, the central discovery is that synthetic task distillation transfers broad job semantics to a small language model: GPT-4 is prompted to generate diverse job-understanding classification tasks, each with a taxonomy pool of labels enriched with definitions and aliases plus a reasoning paragraph explaining the chosen label, and a complementary taxonomy-agnostic entity extraction task. Fine-tuning Flan-T5-XL on over 100k sampled job postings labeled this way yields a base model with strong zero-shot performance on held-out attributes such as nurse specialty and shift, without any explicit nurse-related training. Adding a small amount of linguist-annotated data through L

What carries the argument

The central mechanism is GPT-4-generated synthetic tasks with reasoning traces: for each generated attribute, GPT-4 constructs a taxonomy pool of candidate labels with definitions and aliases, selects the correct label for a job posting, and provides a reasoning paragraph; fine-tuning on these instruction-conditioned examples teaches the SLM the job-understanding skill itself rather than memorized labels. The companion mechanism is low-rank adaptation (LoRA) with attribute grouping: each task-specific adapter is a small set of low-rank matrices on the frozen base model, and attributes are clustered by the average embedding of their taxonomy values so that semantically related attributes (e.g

Load-bearing premise

That GPT-4-generated synthetic taxonomies and reasoning traces, which the authors acknowledge are not aligned with LinkedIn's production taxonomies, are similar enough to real job attributes that a model trained on them transfers zero-shot—and specifically that the synthetic training set truly contains no nurse-related content, since the authors assert but do not verify this.

What would settle it

Inspect the 100k+ synthetic training records for nurse-related taxonomy values (e.g., 'registered nurse', specialty or shift labels) or for job postings describing nursing work. If any appear, the zero-shot nurse results in Table 2 are partially in-distribution and the out-of-distribution generalization claim is weakened. A cleaner test: re-run SYNTH training on synthetic data generated only from non-nursing job postings and check whether nurse specialty/shift precision-recall still reaches 93/85 and 89/64.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • A single SLM base model, fine-tuned only on synthetic tasks, can be served directly for attributes with no labeled data, making zero-shot job understanding practical.
  • Task-specific adaptation costs about 300 annotated samples per attribute via a LoRA adapter, so adding a new job attribute no longer requires training and maintaining a separate deep model.
  • The multi-adapter serving infrastructure loads the base model once and switches adapters per request, so 15 attributes are served through one pipeline with negligible adapter-switch overhead.
  • The same adapter-plus-grouping framework is intended to extend to embedding-based retrieval for future large-cardinality attributes, not just the current occupation case.
  • Online A/B tests show the new SLM improves product metrics such as job sessions, qualified applications, and nurse-segment weekly active users while reducing negative feedback like job recommendation facepalms.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same synthetic-distillation-plus-adapter recipe could transfer to other rapidly evolving taxonomies (product catalogs, legislation, medical codes), where a small model with a few adapters replaces a zoo of per-taxonomy classifiers.
  • The paper does not report a leakage check on the synthetic training data for nurse-related terms; since the 100k+ postings are sampled from real job traffic, confirming the absence of nurse taxonomy values in training would pin down how much of the zero-shot gain is genuinely out-of-distribution.
  • The semantic grouping of attributes by averaged taxonomy embeddings suggests a cheap onboarding path for new professional segments: embed the new attribute's labels and assign it to the nearest existing adapter cluster, avoiding new model training; the paper's K-Means step makes this testable.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper describes a production job-understanding system at LinkedIn. The authors fine-tune Flan-T5-XL on a large corpus of GPT-4-generated synthetic classification tasks (with reasoning traces and taxonomy definitions) to obtain a base SLM with claimed zero-shot generalization over job attributes. On top of this base model, they introduce a LoRA multi-adapter architecture with K-means attribute grouping to enable cheap task-specific adaptation. Offline evaluations on Nurse Specialty/Shift (Table 2) and Occupation/Seniority/Workplace Type (Table 3) show large gains over non-fine-tuned baselines and legacy models; online A/B tests report statistically significant product-metric improvements (Table 4). The paper claims this shows that an industry-scale job-understanding system can be built from one small model with minimal human annotation and low serving cost.

Significance. If the zero-shot generalization claim is clean, this is a strong practical result: it demonstrates that a small open-source language model fine-tuned on synthetic data can replace a collection of expensive per-attribute pipelines, and that multi-adapter serving with attribute grouping is operationally viable. The paper includes credible production evidence — deployment on 15 attributes, 50+ A100 GPUs, online A/B tests with p<.05 — which is rare and valuable. The multi-adapter serving architecture and the nearline pipeline description are useful for practitioners. However, the core scientific claim depends on an unverified data-cleanliness assumption about the synthetic training corpus, and several design choices are not ablated. The paper is therefore of interest but needs substantial revision before the central claims are fully supported.

major comments (4)
  1. [§3.2, §4.1, Table 2] The load-bearing zero-shot comparison for Nurse attributes rests on the assertion that 'COMB and SYNTH were trained without any Nurse related tasks' (§4.1). This is not demonstrated. The synthetic data are generated by prompting GPT-4 on 'over 100k job postings' sampled from LinkedIn (§3.2), which almost certainly include nursing roles. GPT-4 is asked to 'flexibly generate diverse attribute types' and construct taxonomies; no filtering of nurse-related content is described. If any generated task or taxonomy entry is semantically related to nurse specialties or shifts, the SYNTH gains in Table 2 (e.g., Shift 93/85 vs COMB 28/21) are partially in-distribution rather than zero-shot. Please provide evidence that the synthetic training set contains no nurse-related tasks — e.g., an audit of generated attribute/taxonomy types, or confirmation that nurse postings were excluded from the sampled
  2. [§3.2, §4.1, Table 2] The paper attributes SYNTH's effectiveness to three factors: reasoning traces, taxonomy definitions/aliases, and large-scale diverse data. No ablation separates these factors. COMB and SYNTH differ simultaneously in data volume, attribute diversity, and the presence of reasoning traces, so none of the three can be identified as the cause of the observed gains. To support the claim that reasoning traces and taxonomy definitions are essential, the authors should include a variant of SYNTH without reasoning traces (or with labels only), and a variant with a smaller sample to show the effect of scale. Without such ablations, the design rationale in §3.2 remains untested.
  3. [Table 2, Table 3, §4.1] Offline results are reported as single P/R numbers with no error bars, confidence intervals, significance tests, or sample sizes. With only two Nurse attributes and three other attributes, the claim of 'significant gains' is statistically unsupported. Additionally, Table 3's Occupation evaluation is 'collected only from challenging cases where the job title model failed to produce a valid occupation' — this is a selected subset and should be described as such; a random-split evaluation or a discussion of selection bias is needed before comparing with legacy models. Please provide dataset sizes, standard deviations, and appropriate significance tests, and clarify the nature of the evaluation sets.
  4. [§3.3, §4] The multi-adapter architecture with attribute grouping is a central contribution, but no experiment validates it. The paper states that K-means grouping 'reduces the number of adapters while preserving task-specific performance,' yet no offline or online comparison is shown between grouped adapters and per-attribute adapters. There is also no sensitivity analysis for K or for the choice of embedding SLM. Because reduced operational complexity is one of the claimed practical benefits, the paper should quantify the reduction (e.g., number of adapters before/after grouping) and report the quality impact on the affected attributes.
minor comments (4)
  1. [References] References [8] and [24] are duplicates: both are titled 'Enhancing E-Commerce Query Rewriting: A Large Language Model Approach with Domain-Specific Pre-Training and Reinforcement Learning.' Please merge or correct.
  2. [General] Several typos and formatting artifacts need correction: 'Table. reftab:taxonomy' (Table 1 caption), 'differnt' (§1), 'Fig.??' (§3.4), and 'arXiv:2607.24783v1' in the PDF header appears to be the manuscript's own ID rather than a reference.
  3. [§4.1] The statement 'We observe comparable performance across other SLMs when applying the same methodology' is not accompanied by any data. Either provide the comparison or remove the claim.
  4. [§3.2] The taxonomy-free entity extraction task is mentioned but 'omitted due to space constraints.' Since this task is part of the proposed framework, its omission from the paper makes the method description incomplete. If space is a constraint, at least summarize its structure and contribution in an appendix.

Circularity Check

0 steps flagged

No circular dependency; the main caveat is an unverified data-exclusion claim, which is a correctness risk rather than a circularity.

full rationale

The paper's derivation chain is: (1) prompt GPT-4 to generate synthetic classification tasks with reasoning traces from 100k+ real job postings; (2) fine-tune Flan-T5-XL on this synthetic data; (3) evaluate on independent human-annotated held-out data for Nurse Specialty and Shift; (4) optionally add a LoRA adapter trained on ~300 annotated nurse samples. The central zero-shot claim is empirical and is tested against external human labels, not against the same synthetic labels used for training. No equation reduces a predicted quantity to a fitted parameter by construction, and no load-bearing result is justified solely by a self-citation chain. The one genuinely load-bearing statement is in §4.1: 'COMB and SYNTH were trained without any Nurse related tasks.' This exclusion is asserted but not demonstrated. Because the synthetic data are GPT-4 outputs derived from 100k+ sampled LinkedIn job postings, which plausibly include nursing roles, the zero-shot interpretation of Table 2 depends on an unverified data-hygiene assumption. That is a serious external-validity or contamination concern, but it is not a circularity: even under contamination, the model is not being evaluated on its own training labels, and the paper does not define 'zero-shot' in terms of the result it is trying to prove. Similarly, the use of LoRA, attribute grouping, and K-Means clustering follows standard external methodology, not an author-imported uniqueness theorem. Comparing against legacy models [22,25] involves the same company's prior systems, but those comparisons are benchmarks, not circular justifications. Overall, the framework is self-contained with respect to the claimed prediction mechanism; the unverified nurse-task exclusion should be addressed but does not make the derivation circular.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The central claim relies on GPT-4-generated synthetic data transferring to production taxonomies, on the effectiveness of reasoning traces, and on the semantic coherence of K-Means attribute grouping—none of which are independently evidenced beyond the headline results. The main unmeasured design degrees of freedom (K, LoRA rank, prompts) act as free parameters that are not disclosed.

free parameters (4)
  • K (number of attribute clusters for adapter grouping)
    K-Means clustering in §3.3; the number of clusters K and resulting adapter assignments are not reported but influence how many adapters are shared.
  • LoRA rank and alpha
    LoRA hyperparameters for task adapters are not disclosed (§3.3), though they control adaptation capacity and performance.
  • GPT-4 prompt design for synthetic tasks
    Exact prompts, sampling temperature, number of tasks per posting, and any filtering of GPT-4 outputs are not given (§3.2); these choices determine the synthetic training distribution.
  • Number and selection of adapter training samples = ~300 samples
    The paper says 'approximately 300 samples are sufficient' (§4.1) but does not state the exact count, split, or selection procedure for the nurse attributes.
axioms (5)
  • domain assumption GPT-4 can generate accurate, diverse, domain-specific classification tasks with reasoning traces from job postings.
    Section 3.2 relies entirely on this; no quality audit of synthetic labels or reasoning traces is provided.
  • domain assumption Reasoning traces plus taxonomy definitions are the cause of the gains.
    No ablation separates reasoning traces from labels or definitions; §3.2 asserts this without evidence.
  • domain assumption Flan-T5-XL is representative of other SLMs for this methodology.
    The paper states 'comparable performance across other SLMs' (§4.1) but shows no results for any other model.
  • domain assumption The human-annotated evaluation data are accurate, comparable across systems, and free of leakage with the synthetic training data.
    Evaluation in §4.1 assumes correctness and unbiasedness of the test labels and no train/eval contamination, but the paper does not demonstrate the latter.
  • domain assumption Average-then-K-Means over taxonomy-value embeddings yields semantically coherent attribute groups.
    §3.3 uses this grouping; the paper provides no validation that grouped attributes indeed share an adapter without task interference.

pith-pipeline@v1.3.0-alltime-deepseek · 9686 in / 14897 out tokens · 138303 ms · 2026-08-02T10:27:39.505709+00:00 · methodology

0 comments
read the original abstract

Job understanding is critical to LinkedIn's mission of connecting talent with opportunity. This task involves transforming unstructured and noisy job postings into standardized or derived job attributes that power numerous LinkedIn products. However, building a scalable, cost-efficient, and high-performing job understanding system remains challenging. In this paper, we present a unified semantic modeling framework powered by a small language model (SLM) to address the challenges. We begin by fine-tuning an open-source SLM using a suite of carefully curated synthetic tasks augmented with reasoning traces. These tasks jointly target taxonomy-guided classification and taxonomy-agnostic entity extraction. This allows the resulting model to acquire robust zero-shot generalization for job understanding in structured and unstructured contexts. Building upon this foundation, we introduce a multi-adapter architecture with attribute grouping to facilitate efficient task-specific adaptation while streamlining model management across diverse downstream attributes. Offline evaluations and online A/B tests demonstrate significant performance improvement while reducing operational complexity. Our work provides practical insights into building industry-scale text understanding systems.

Figures

Figures reproduced from arXiv: 2607.24783 by Alice Leung, Baofen Zheng, Benjamin Hoan Le, Dan Xu, Jianqiang Shen, Jingwei Wu, Liangjie Hong, Neha Saraf, Qianqi Shen, Qi Xiao, Ran Zhou, Saurabh Gupta, Wenjing Zhang, Wen Pu.

Figure 1
Figure 1. Figure 1: Multi-adapter model training and serving architecture with attribute grouping. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Job understanding two-stage model architecture for large-cardinality attributes. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 13 linked inside Pith

  1. [1]

    Kato, and Masafumi Oyamada

    Kenya Abe, Kunihiro Takeoka, Makoto P. Kato, and Masafumi Oyamada. 2025. LLM-based Query Expansion Fails for Unfamiliar and Ambiguous Queries.arXiv preprint arXiv:2505.12694(2025)

    Pith/arXiv arXiv 2025

  2. [2]

    Palakorn Achananuparp, Ee-Peng Lim, and Yao Lu. 2026. A Multi-Stage Frame- work with Taxonomy-Guided Reasoning for Occupation Classification Using Large Language Models. InProceedings of the 20th International AAAI Conference on Web and Social Media (ICWSM): Understanding the World Through the Web

    2026

  3. [3]

    Boselli, M

    R. Boselli, M. Cesarini, F. Mercorio, and M. Mezzanzanica. 2017. Using machine learning for labour market intelligence. Springer International Publishing, Cham, 330–342

    2017

  4. [4]

    Tom Brown et al. 2020. Language Models are Few-Shot Learners.Advances in Neural Information Processing Systems33 (2020), 1877–1901

    2020

  5. [5]

    Chi, Jeff Dean, Jacob De- vlin, Adam Roberts, Denny Zhou, Quoc V

    Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Eric Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, Albert Web- son, Shixiang Shane Gu, Zhuyun Dai, Mirac Suzgun, Xinyun Chen, Aakanksha Chowdhery, Sharan Narang, Gaurav Mishra, Adams Yu, Vincent Zhao, Yanping Huang, Andrew Dai, Hongkun Yu, Slav Petrov, Ed H. Chi, Jeff Dea...

    Pith/arXiv arXiv doi:10.48550/arxiv.2210.11416 2022

  6. [6]

    Benjamin Clavié and Guillaume Soulié. 2023. Large Language Models as Batteries- Included Zero-Shot ESCO Skills Matchers.arXiv preprint arXiv:2307.03539(2023)

    Pith/arXiv arXiv 2023

  7. [7]

    Colombo, F

    E. Colombo, F. Mercorio, and M. Mezzanzanica. 2019. AI meets labor market: Exploring the link between automation and skills.Information Economics and Policy47 (2019), 27–37

    2019

  8. [8]

    Aijun Dai, Zhenyu Zhu, Haiqing Hu, Guoyu Tang, Lin Liu, and Sulong Xu. 2024. Enhancing E-Commerce Query Rewriting: A Large Language Model Approach with Domain-Specific Pre-Training and Reinforcement Learning. InProceedings of the 33rd ACM International Conference on Information and Knowledge Management. 4439–4445

    2024

  9. [9]

    Decorte, J

    J.-J. Decorte, J. Van Hautte, T. Demeester, and C. Develder. 2021. JobBERT: Understanding Job Titles through Skills.arXiv preprint arXiv:2109.09605(2021)

    Pith/arXiv arXiv 2021

  10. [10]

    Decorte, S

    J.-J. Decorte, S. Verlinden, J. Van Hautte, J. Deleu, C. Develder, and T. Demeester

  11. [11]

    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

    2019

  12. [12]

    Mahak Goindani, Qiaoling Liu, Josh Chao, and Valentin Jijkoun. 2017. Employer Industry Classification Using Job Postings. InProceedings of the IEEE International Conference on Data Mining Workshops (ICDMW). IEEE, 183–188

    2017

  13. [13]

    Junxian He, Chunting Zhou, Xuezhe Ma, Taylor Berg-Kirkpatrick, and Graham Neubig. 2022. Towards a Unified View of Parameter-Efficient Transfer Learning. InInternational Conference on Learning Representations

    2022

  14. [14]

    Shwai He, Liang Ding, Daize Dong, Jeremy Zhang, and Dacheng Tao. 2022. SparseAdapter: An Easy Approach for Improving the Parameter-Efficiency of Adapters. InFindings of the Association for Computational Linguistics: EMNLP

    2022

  15. [15]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. LoRA: Low-Rank Adaptation of Large Language Models.arXiv preprint arXiv:2106.09685(2021)

    Pith/arXiv arXiv 2021

  16. [16]

    Ikudo, J

    A. Ikudo, J. I. Lane, J. Staudt, and B. A. Weinberg. 2019. Occupational classifica- tions: A machine learning approach.Journal of Economic and Social Measurement 44, 2-3 (2019), 57–87

    2019

  17. [17]

    Javed, Q

    F. Javed, Q. Luo, M. McNair, F. Jacob, M. Zhao, and T. S. Kang. 2015. Carotene: A Job Title Classification System for the Online Recruitment Domain. InIEEE BigDataService. IEEE, 286–293

    2015

  18. [18]

    Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. 2022. Large Language Models Are Zero-Shot Reasoners. InProceedings of the 36th International Conference on Neural Information Processing Systems (NeurIPS). 22199–22213

    2022

  19. [19]

    Dikaiakos, Dimosthenis Stefanidis, and George Pallis

    Arina Kostina, Marios D. Dikaiakos, Dimosthenis Stefanidis, and George Pallis

  20. [20]

    Kouretsis, A

    A. Kouretsis, A. Bampouris, P. Morfiris, and K. Papageorgiou. 2020. LabourR: clas- sify multilingual labour market free-text to standardized hierarchical occupations. https://cran.r-project.org/web/packages/labourR/index.html

    2020

  21. [21]

    Nan Li, Bo Kang, and Tijl De Bie. 2023. Unsupervised Occupation Extrac- tion and Standardization Leveraging Large Language Models.arXiv preprint arXiv:2309.09708(2023)

    Pith/arXiv arXiv 2023

  22. [22]

    Shan Li, Baoxu Shi, and Jaewon Yang. 2024. Encoding a Job Posting as an Embed- ding Using a Graph Neural Network. US Patent No. 11,861,295 (US11861295B2)

    2024

  23. [23]

    Shan Li, Baoxu Shi, Jaewon Yang, Ji Yan, Shuai Wang, Fei Chen, and Qi He. 2020. Deep Job Understanding at LinkedIn. InProceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. ACM, 2145–2148

    2020

  24. [24]

    Ping Liu and et al. 2026. Enhancing E-Commerce Query Rewriting: A Large Lan- guage Model Approach with Domain-Specific Pre-Training and Reinforcement Learning. InTo appear in the 35th ACM International Conference on Information and Knowledge Management

    2026

  25. [25]

    Ping Liu, Haichao Wei, Xiaochen Hou, Jianqiang Shen, Shihai He, Qianqi Shen, Zhujun Chen, Fedor Borisyuk, Daniel Hewlett, Liang Wu, et al. 2025. Linksage: Optimizing job matching using graph neural networks. InProceedings of the 31st ACM SIGKDD V. 1. 2448–2457

    2025

  26. [26]

    Chen Luo, William Headden, Neela Avudaiappan, Haoming Jiang, Tianyu Cao, Qingyu Yin, Yifan Gao, Zheng Li, Rahul Goutam, Haiyang Zhang, et al . 2022. Query attribute recommendation at amazon search. InProceedings of the 16th ACM conference on recommender systems. 506–508

    2022

  27. [27]

    Chen Luo, Xianfeng Tang, Hanqing Lu, Yaochen Xie, Hui Liu, Zhenwei Dai, Limeng Cui, Ashutosh Joshi, Sreyashi Nag, Yang Li, et al. 2024. Exploring Query Understanding for Amazon Product Search. In2024 IEEE International Conference on Big Data (BigData). IEEE, 2343–2348

    2024

  28. [28]

    Sourab Mangrulkar, Sylvain Gugger, Lysandre Debut, Younes Belkada, Sayak Paul, and Benjamin Bossan. 2022. PEFT: State-of-the-art Parameter-Efficient Fine-Tuning methods. https://github.com/huggingface/peft

    2022

  29. [29]

    OpenAI. 2022. Introducing ChatGPT. https://openai.com/blog/chatgpt. Accessed: YYYY-MM-DD

    2022

  30. [30]

    OpenAI. 2023. GPT-4 Technical Report.arXiv preprint arXiv:2303.08774(2023). Job Understanding At LinkedIn SIGIR ’26, July 20–24, 2026, Melbourne, VIC, Australia

    Pith/arXiv arXiv 2023

  31. [31]

    Chuan Qin, Hengshu Zhu, Chen Zhu, Tong Xu, Fuzhen Zhuang, Chao Ma, Jing- shuai Zhang, and Hui Xiong. 2019. DuerQuiz: A Personalized Question Rec- ommender System for Intelligent Job Interview. InProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD). 2165–2173

    2019

  32. [32]

    D. E. Russ, K. Y. Ho, J. S. Colt, K. R. Armenti, D. Baris, W. H. Chow, F. Davis, A. Johnson, M. P. Purdue, M. R. Karagas, et al . 2016. Computer-based coding of free-text job descriptions to efficiently identify occupations in epidemiological studies.Occupational and Environmental Medicine73, 6 (2016), 417–424

    2016

  33. [33]

    Krishna Srinivasan, Karthik Raman, Anupam Samanta, Lingrui Liao, Luca Bertelli, and Michael Bendersky. 2022. QUILL: Query Intent with Large Language Models using Retrieval Augmentation and Multi-stage Distillation. InProceedings of the 2022 Conference on Empirical Methods in Natural Language Processing: Industry Track. 492–501

    2022

  34. [34]

    Xiaofei Sun et al. 2023. Text Classification via Large Language Models (CARP). InFindings of the Association for Computational Linguistics: EMNLP 2023

    2023

  35. [35]

    Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B Hashimoto. 2023. Stanford alpaca: An instruction-following Llama model

    2023

  36. [36]

    Qwen Team. 2025. Qwen3 Technical Report. arXiv:2505.09388 [cs.CL] https: //arxiv.org/abs/2505.09388

    Pith/arXiv arXiv 2025

  37. [37]

    Nandan Thakur, Nils Reimers, Johannes Daxenberger, and Iryna Gurevych. 2021. Augmented SBERT: Data Augmentation Method for Improving Bi-Encoders for Pairwise Sentence Scoring Tasks. InProceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 296–310

    2021

  38. [38]

    Sowmya Vajjala and Shwetali Shimangaud. 2025. Text Classification in the LLM Era – Where Do We Stand?arXiv preprint arXiv:2502.11830(2025). https: //arxiv.org/abs/2502.11830

    Pith/arXiv arXiv 2025

  39. [39]

    Yaqing Wang, Sahaj Agarwal, Subhabrata Mukherjee, Xiaodong Liu, Jing Gao, Ahmed Hassan Awadallah, and Jianfeng Gao. 2022. AdaMix: Mixture- of-Adaptations for Parameter-efficient Model Tuning.arXiv preprint arXiv:2205.12410(2022). Accepted at EMNLP 2022

    Pith/arXiv arXiv 2022

  40. [40]

    Dai, and Quoc V Le

    Jason Wei, Maarten Bosma, Vincent Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M. Dai, and Quoc V Le. 2022. Finetuned Language Models are Zero-Shot Learners. InInternational Conference on Learning Representations. https://openreview.net/forum?id=gEZrGCozdqR

    2022

  41. [41]

    Zhang, K

    M. Zhang, K. N. Jensen, and B. Plank. 2022. Kompetencer: Fine-grained skill classification in Danish job postings via distant supervision and transfer learning. arXiv:2205.01381 [cs.CL] https://arxiv.org/abs/2205.01381 arXiv preprint

    Pith/arXiv arXiv 2022

  42. [42]

    Ruqing Zhang, Jiafeng Guo, Yixing Fan, Yanyan Lan, and Xueqi Cheng. 2020. Query understanding via intent description generation. InProceedings of the 29th ACM International Conference on Information & Knowledge Management. 1823–1832

    2020

  43. [2023]

    Extreme Multi-Label Skill Extraction Training Using Large Language Models.arXiv preprint arXiv:2307.10778(2023)

    Pith/arXiv arXiv 2023

  44. [2025]

    Large Language Models For Text Classification: Case Study And Compre- hensive Review.arXiv preprint arXiv:2501.08457(2025)

    Pith/arXiv arXiv 2025