The DOI matches its DataCite record; 23 of the 23 references it cites were found in a registry or an index; 3 of the 4 sources its claims cite were checked for support.
DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter
Open accessAbstract
As Transfer Learning from large-scale pre-trained models becomes more prevalent in Natural Language Processing (NLP), operating these large models in on-the-edge and/or under constrained computational training or inference budgets remains challenging. In this work, we propose a method to pre-train a smaller general-purpose language representation model, called DistilBERT, which can then be fine-tuned with good performances on a wide range of tasks like its larger counterparts. While most prior work investigated the use of distillation for building task-specific models, we leverage knowledge distillation during the pre-training phase and show that it is possible to reduce the size of a BERT model by 40%, while retaining 97% of its language understanding capabilities and being 60% faster. To leverage the inductive biases learned by larger models during pre-training, we introduce a triple loss combining language modeling, distillation and cosine-distance losses. Our smaller, faster and lighter model is cheaper to pre-train and we demonstrate its capabilities for on-device computations in a proof-of-concept experiment and a comparative on-device study.
Overview
AI snapshot
Written from the paper’s full text.
TL;DR
DistilBERT is a distilled version of BERT pre-trained with knowledge distillation, reducing parameters by 40% while retaining 97% of BERT’s GLUE performance and running 60% faster at inference. It is presented as a general-purpose, lighter model suitable for constrained and on-device NLP applications.
Key findings
- DistilBERT has 66M parameters vs. BERT-base’s 110M and takes 410s vs. 668s for a CPU inference pass on STS-B, i.e. 40% smaller and 60% faster.
- On GLUE dev sets, DistilBERT scores 77.0 macro-score vs. BERT-base’s 79.5, retaining 97% of performance; it is always on par or improving over the ELMo baseline.
- On IMDb test accuracy, DistilBERT reaches 92.82 vs. BERT-base’s 93.46; on SQuAD v1.1, DistilBERT reaches 77.7/85.8 EM/F1 vs. BERT-base’s 81.2/88.5.
- Ablations show removing the cosine loss costs -2.96 GLUE macro-score, removing distillation loss costs -1.46, removing MLM loss costs -0.31, and random initialization costs -3.69.
Methods
- Knowledge distillation during pre-training: a student with BERT’s architecture but half the layers, no token-type embeddings and no pooler, initialized from every other teacher layer.
- Training objective is a triple loss combining distillation loss \(L_{ce}\), masked language modeling loss \(L_{mlm}\), and cosine embedding loss \(L_{cos}\), using softmax-temperature.
- Trained on English Wikipedia + Toronto Book Corpus with large batches up to 4K examples, dynamic masking, no next-sentence prediction, on 8 16GB V100 GPUs for approximately 90 hours; evaluated on GLUE, IMDb, SQuAD v1.1, CPU inference, and an iPhone 7 Plus on-device proof-of-concept.
Limitations
- The paper does not state an explicit limitations section.
- Evident from the text: evaluation is limited to English benchmarks and a small set of downstream tasks, and the on-device study is described as a proof-of-concept; the model could be further reduced with quantization.
- Evident from the text: the SQuAD second-distillation result is reported inconsistently—the prose says 79.8 F1 and 70.4 EM, while Table 2 lists DistilBERT (D) as 79.1/86.9.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
15 Own result · 6 Own interpretation · 2 Own method · 2 No basis given · 1 Supported · 1 Partly supported · 1 Sources not found · accuracy not yet measured
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“we leverage knowledge distillation during the pre-training phase and show that it is possible to reduce the size of a BERT model by 40%, while retaining 97% of its language understanding capabilities and being 60% faster.”
- Also stated in: § Size and inference speed · p. 4 · plain
“DistilBERT has 40% fewer parameters than BERT and is 60% faster than BERT.”
- Also stated in: § 6 Conclusion and future work · p. 5 · plain
“We introduced DistilBERT, a general-purpose pre-trained version of BERT, 40% smaller, 60% faster, that retains 97% of the language understanding capabilities.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § Size and inference speed · p. 4 · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“we propose a method to pre-train a smaller generalpurpose language representation model, called DistilBERT, which can then be finetuned with good performances on a wide range of tasks like its larger counterparts.”
- Also stated in: § 1 Introduction · p. 2 · plain
“Our general-purpose pre-trained models can be fine-tuned with good performances on several downstream tasks, keeping the flexibility of larger models.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 1 Introduction · p. 2 · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Further ablation studies indicate that all the components of the triple loss are important for best performances.”
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Excluding the tokenization step, DistilBERT is 71% faster than BERT, and the whole model weighs 207 MB (which could be further reduced with quantization).”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“we leverage knowledge distillation during the pre-training phase and show that it is possible to reduce the size of a BERT model by 40%, while retaining 97% of its language understanding capabilities and being 60% faster.”
- Also stated in: § Size and inference speed · p. 4 · plain
“DistilBERT has 40% fewer parameters than BERT and is 60% faster than BERT.”
- Also stated in: § 6 Conclusion and future work · p. 5 · plain
“We introduced DistilBERT, a general-purpose pre-trained version of BERT, 40% smaller, 60% faster, that retains 97% of the language understanding capabilities.”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: Inference speed measured on a single GLUE task (STS-B) on CPU is representative of DistilBERT's speed advantage generally.
- Also stated in: § Size and inference speed · p. 4 · plain
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“we propose a method to pre-train a smaller generalpurpose language representation model, called DistilBERT”
- Also stated in: § 6 Conclusion and future work · p. 5 · plain
“We introduced DistilBERT, a general-purpose pre-trained version of BERT”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 6 Conclusion and future work · p. 5 · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We found it beneficial to add a cosine embedding loss ( L cos ) which will tend to align the directions of the student and teacher hidden states vectors.”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“On SQuAD, DistilBERT is within 3.9 points of the full BERT.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“we leverage knowledge distillation during the pre-training phase and show that it is possible to reduce the size of a BERT model by 40%, while retaining 97% of its language understanding capabilities and being 60% faster.”
- Also stated in: § 4 Experiments · p. 3 · plain
“DistilBERT also compares surprisingly well to BERT, retaining 97% of the performance with 40% fewer parameters.”
- Also stated in: § 6 Conclusion and future work · p. 5 · plain
“We introduced DistilBERT, a general-purpose pre-trained version of BERT, 40% smaller, 60% faster, that retains 97% of the language understanding capabilities.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 4 Experiments · p. 3 · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“removing the Masked Language Modeling loss has little impact while the two distillation losses account for a large portion of the performance.”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Probably not a substantial claim. 7
- It may state two separate claims as one. 2
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“environmental cost of exponentially scaling these models' computational requirements as mentioned in Schwartz et al. [2019], Strubell et al. [2019].”
What the sources say
- Supportedcites Schwartz et al., 2019 source
The Abstract (p. 1) says the computations required for deep learning research have been doubling every few months and have a large carbon footprint. Section 2, “Red AI” (p. 4), likewise warns that the environmental cost of increasingly large models will not decrease soon. (LLM, conf=0.93)
- Partly supportedcites Strubell et al., 2019 sourceExperimental — accuracy not yet measured
The Introduction (page unavailable) says that computationally hungry NLP models require considerable energy and have associated environmental costs. It also says model development can multiply those costs thousands of times, but it does not establish that the models’ computational requirements are scaling exponentially. (LLM, conf=0.84)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“We applied best practices for training BERT model recently proposed in Liu et al. [2019].”
What the sources say
- Supportedcites Liu et al., 2019 source
The Introduction (p. 1) says Liu et al. propose an improved recipe for training BERT, and Section 4 (p. 4) evaluates which choices matter for BERT pretraining. The citing authors’ statement that they applied those practices to DistilBERT describes their own work, not a finding attributed to Liu et al. (LLM, conf=0.92)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Among the 9 tasks, DistilBERT is always on par or improving over the ELMo baseline (up to 19 points of accuracy on STS-B).”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In this case, we were able to reach interesting performances given the size of the model: 79.8 F1 and 70.4 EM, i.e. within 3 points of the full model.”
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We further demonstrated that DistilBERT is a compelling option for edge applications.”
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“BERT and DistilBERT results are the medians of 5 runs with different seeds.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“current research 1 on pre-trained models indicates that training even larger models still leads to better performances on downstream tasks.”
What the sources say
- Not matchedcites MegatronLM
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
- Not matched
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“We report scores on the development sets for each task by fine-tuning DistilBERT without the use of ensembling or multi-tasking scheme for fine-tuning (which are mostly orthogonal to the present work).”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“While most prior work investigated the use of distillation for building task-specific models, we leverage knowledge distillation during the pre-training phase”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper may give no reason why it holds.
The paper relies on this without support.
Operating large pre-trained models on-device in real-time is hampered by their growing computational and memory requirements.
“the growing computational and memory requirements of these models may hamper wide adoption.”
The paper relies on this without support.
Near-zero probabilities in a trained model's output distribution reflect its generalization capabilities.
“some of these”
The paper relies on this without support.
Inference speed measured on a single GLUE task (STS-B) on CPU is representative of DistilBERT's speed advantage generally.
“Inference time of a full pass of GLUE task STS-B (sentiment analysis) on CPU with a batch size of 1.”
“DistilBERT has 40% fewer parameters than BERT and is 60% faster than BERT.”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“This idea has been successfully applied to LLMs, where distilled students approximate the generation patterns of larger teachers while retaining much of their performance (Sanh et al., 2019).”
From Vision-Guided Iterative Refinement for Frontend Code Generation (2026) · cites this paper as Sanh et al., 2019
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“even benchmarking is not a reliable estimator of parameter count [12], [15], [28], [29]”
From Hollow-LLM Attack: Computationally Trivial Weights in Zero-Knowledge Verification of LLM Inference (2026) · cites this paper as [29]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“It achieves similar performance to the original BERT model while using fewer resources. The approach leveraged knowledge distillation during the pretraining phase and reduced the size of general BERT by 40%, yet it retained a language understanding capability of 97%.”
From A Review on Deep-Learning-Based Cyberbullying Detection (2023) · cites this paper as [53]
The cited source says what the claim attributes to it.
In the citing paper
“Unlike in many similar circumstances, we do not employ knowledge distillation [56], because we discovered that for our sizes of the models, this technique does not provide any quality improvement.”
From Full Line Code Completion: Bringing AI to Desktop (2024) · cites this paper as [56]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ 4.1 Downstream task benchmark · p. 3
“In this case, we were able to reach interesting performances given the size of the model: 79.8 F1 and 70.4 EM, i.e. within 3 points of the full model.”
§ 3 DistilBERT: a distilled version of BERT · p. 3
“| DistilBERT (D) | - | 79.1/86.9 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
“DistilBERT is always on par or improving over the ELMo baseline (up to 19 points of accuracy on STS-B).”
Reported 19 · recomputed 16.5
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325
RoBERTa: A Robustly Optimized BERT Pretraining Approach
Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., & Stoyanov, V. (2019) · arXiv · doi:10.48550/arxiv.1907.11692 · Cited by 16,515
Distilling the Knowledge in a Neural Network
Hinton, G., Vinyals, O., & Dean, J. (2015) · arXiv · doi:10.48550/arxiv.1503.02531 · Cited by 14,204
Deep Contextualized Word Representations
Peters, M., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., & Zettlemoyer, L. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) · doi:10.18653/v1/n18-1202 · Cited by 11,983
Transformers: State-of-the-Art Natural Language Processing
Wolf, T., Debut, L., Sanh, V., Chaumond, J., Delangue, C., Moi, A., Cistac, P., Rault, T., Louf, R., Funtowicz, M., et al. (2020) · Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations · doi:10.18653/v1/2020.emnlp-demos.6 · Cited by 8,741
Self-citationSQuAD: 100,000+ Questions for Machine Comprehension of Text
Rajpurkar, P., Zhang, J., Lopyrev, K., & Liang, P. (2016) · Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d16-1264 · Cited by 6,208
GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding
Wang, A., Singh, A., Michael, J., Hill, F., Levy, O., & Bowman, S. (2018) · Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP · doi:10.18653/v1/w18-5446 · Cited by 4,222
Learning Word Vectors for Sentiment Analysis
Maas, A. L., Daly, R. E., Pham, P. T., Huang, D., Ng, A. Y., & Potts, C. (2011) · Cited by 3,710
Energy and Policy Considerations for Deep Learning in NLP
Strubell, E., Ganesh, A., & McCallum, A. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1355 · Cited by 3,684
Research works printed but not linked: 1
- Cristian Bucila, Rich Caruana, and Alexandru Niculescu-Mizil. Model compression. In KDD , 2006.
Cited by
Sorted by citations
Items 4,501–4,600 of 4,692
Addressing Past Tense Temporal Attacks in Large Language Models through Multiple Methods
Zhai, N.-J., Tuo, X., & Jin, Z. (2024) · International Conference on E-Society, E-Education and E-Technology (ICSET) · doi:10.1145/3704217.3705008
Jekyll institute or Mrs Hyde? gender identification with machine learning
Gombert, A., Sánchez-López, B., & Cerquides, J. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.110087
Kneuer, M., Roch, J., Heid, U., et al. (2025) · European Politics and Society · doi:10.1080/23745118.2024.2446827
ET-SSM: Linear-Time Encrypted Traffic Classification Method Based On Structured State Space Model
Li, Y., Zhao, X., Zha, Z., & Ling, Z.-H. (2024) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc63619.2025.10849155
Ladeira, L. Z., Santos, F. A., & Villas, L. A. (2025) · Journal of Internet Services and Applications · doi:10.5753/jisa.2025.4890
A Light-Weight Image to Description Model Using Knowledge Distillation
Rehan, E., Memon, S., Raza, M., Naeem, M. A., & Bhatti, F. N. (2024) · International Conference on Artificial Intelligence, Blockchain, and Internet of Things (AIBThings) · doi:10.1109/aibthings63359.2024.10863400
Lee, J., Yu, S., Jang, J., Park, K., & Kim, Y. (2025) · IEEE Access · doi:10.1109/access.2025.3540090
A Comprehensive Analysis on Sentiment Using Deep Learning and Transformers
Mohsin, M., Wankhede, J. P., Fatima, A., & Solanki, V. K. (2024) · IEEE International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA) · doi:10.1109/icccmla63077.2024.10871474
sLLM-based Korean Text-to-Motion Retrieval System Using CNN
Choi, J., & Park, K.-H. (2025) · The Journal of Korea Robotics Society · doi:10.7746/jkros.2025.20.1.077
bin Yusuf, Y. L., & Saee, S. B. (2025) · Acta Informatica Pragensia · doi:10.18267/j.aip.263
Valeva, S., Vallejo, A., Klimberg, R. K., et al. (2025) · Journal of Statistics and Data Science Education · doi:10.1080/26939169.2025.2475764
Model for named entity extraction from short fire event-related texts
Šeleng, M., Dlugolinský, Š., Staňo, M., et al. (2025) · Procedia Computer Science · doi:10.1016/j.procs.2025.02.152
Classifier-Guided Captioning Across Modalities
Shaulov, A., Shaharabany, T., Shaar, E., et al. (2025) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49660.2025.10888916
Understanding the Impact of News Articles on the Movement of Market Index: A Case on Nifty 50
Dasgupta, S., Satpati, P., Choudhary, I., & Sen, J. (2024) · IEEE Silchar Subsection Conference (SILCON) · doi:10.1109/silcon63976.2024.10910416
Subjective Answer Evaluation Using Keyword Similarity and Regression Techniques
Kapparad, P. (2024) · IEEE Silchar Subsection Conference (SILCON) · doi:10.1109/silcon63976.2024.10910888
PoliGuilt: Two level guilt detection from social media texts
Meque, A. G. M., Balouchzahi, F., Gelbukh, A. F., & Sidorov, G. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.127187
Peng, J., & Tian, F. (2024) · IEEE Smart World Congress (SWC) · doi:10.1109/swc62898.2024.00230
Social Structure-Based Methodology for User Identity Matching Across Social Networks
Zhu, G. (2025) · 计算机科学辑要 · doi:10.63313/cs.8002
Enhancing Sexism Detection with Ensemble Language Models
Calderón, G. (2024) · IEEE Central America and Panama Convention (CONCAPAN) · doi:10.1109/concapan63470.2024.10933849
Improving Stock Index Prediction Based on Financial Sentiment Analysis and Time Series Models
Yang, L., Liu, Y., Li, Z., Qi, B., & Feng, J. (2024) · doi:10.1109/iccfi63302.2024.00008
Ravindran, P. (2025) · International Journal of Advanced Research in Science Communication and Technology · doi:10.48175/ijarsct-24468 · Cited by 1
Toward Fine-Tuning Large Language Models in Ontology of Microbial Phenotypes Construction
Baidya, A., Do, T., & Gnimpiéba, E. Z. (2024) · IEEE International Conference on Bioinformatics and Biomedicine (BIBM) · doi:10.1109/bibm62325.2024.10947604
A multi-modal transformer-based model for generative visual dialog system
Elshamy, G. M., Alfonse, M., Hegazy, I., & Aref, M. M. (2025) · Applied Computer Science · doi:10.35784/acs_6856
Shymanskyi, V., & Sabov, V. (2025) · Lecture notes on data engineering and communications technologies · doi:10.1007/978-3-031-84228-3_22
The BERT Model Family and Encoder-Only Transformers
Ciapparelli, M., & de Varda, A. G. (2025) · Elsevier eBooks · doi:10.1016/b978-0-323-95504-1.00507-x · Cited by 1
Fine-Grained Spatial and Verbal Losses for 3D Visual Grounding
Dey, S., Unal, O., Sakaridis, C., & Gool, L. V. (2025) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv61041.2025.00475
A Dataset of Semantically Related Multiword Terms of the Electrical Engineering Domain
Druselmann, M., & Harbusch, K. (2024) · International Conference on Natural Language Processing and Information Retrieval (NLPIR) · doi:10.1145/3711542.3711550
Multi-Knowledge Distillation for Constructing Math Word Problem Encoder
Cheng, X., Ming, H., Shen, Z., & Yu, X. (2024) · International Conference on Intelligent Education and Intelligent Research (IEIR) · doi:10.1109/ieir62538.2024.10960253
Compacting Language Model for Natural Language Understanding on English Datasets
Jeremy, N. H., & Suhartono, D. (2024) · International Conference on Informatics and Computing (ICIC) · doi:10.1109/icic64337.2024.10957202
Robotic Process Automation Activity Recommendation using NLP Techniques
Herath, T., & Asanka, D. (2025) · International Conference on Advanced Research in Computing (ICARC) · doi:10.1109/icarc64760.2025.10963182
Ensemble Approach for Religious Hate Speech Detection in Text-Based Images on Facebook
Pavithra, D. R., & Smith, R. (2025) · International Conference on Advanced Research in Computing (ICARC) · doi:10.1109/icarc64760.2025.10962881
Decoding reddit memes virality
Sah, T., & Jordan, K. (2025) · International Journal of Data Science and Analytics · doi:10.1007/s41060-025-00772-5
Pozo-Villafuerte, D. D., & Villacís-Miranda, A. (2025) · Computational Economics · doi:10.1007/s10614-025-10935-y · Cited by 2
Bailey, A. L., Johnson, A., Shankar, N. B., Veeramani, H., Washington, J. A., & Alwan, A. A. (2025) · Journal of Educational Measurement · doi:10.1111/jedm.12435 · Cited by 1
R, P. J. I., Erramchetty, S. K., Kabilan, K., & Sandosh, S. (2025) · Journal of Poetry Therapy · doi:10.1080/08893675.2025.2493818
Disfluency processing for cascaded speech translation involving English and Indian languages
Mujadia, V., Mishra, P., & Sharma, D. M. (2025) · Language Resources and Evaluation · doi:10.1007/s10579-025-09818-3 · Cited by 1
Sasaki, Y., Zhang, J., & Shiraishi, Y. (2024) · IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT) · doi:10.1109/wi-iat62293.2024.00049
Multi-Modal Integration for Predicting Biochemical Recurrence in Prostate Cancer
Dudhabhate, S., Chattopadhyay, N., Parekh, A., et al. (2025) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi60581.2025.10980823
Search method for ship regulations based on semantic similarity using natural language processing
Han, I.-S., Roh, M.-I., Kong, M.-C., Park, S., & Lee, J.-Y. (2025) · Ships and Offshore Structures · doi:10.1080/17445302.2025.2502462 · Cited by 1
Graph Regularized Encoder Training for Extreme Classification
Mittal, A., Mohan, S., Saini, D., Asokan, S., Prabhu, S. C., Kumar, L., Malhotra, P., Jiao, J., Singh, A. P., Agarwal, S., et al. (2025) · International World Wide Web Conference (WWW) · doi:10.1145/3701716.3715230
Leveraging Large Language Models for Systems-Based Learning from Hospital Incident ReportsPreprint
Liu, D., Hegde, S., & Ehrett, C. (2025) · doi:10.36227/techrxiv.174803009.90361955/v1
A Conceptual Framework for Efficient and Sustainable Pruning Techniques in Deep Learning Models
Smarts, N., Rajalakshmi, S., & Kuthadi, V. M. (2025) · Journal of Computer Science · doi:10.3844/jcssp.2025.1113.1128
Sentiment analysis with XAI algorithm for movies recommendation
Wafaa, A., Karima, E. H., Houda, R. N. E., & Ilham, S. (2025) · International Conference on Innovative Research in Applied Science, Engineering and Technology (IRASET) · doi:10.1109/iraset64571.2025.11008255
Sentence Classification Using Transfer Learning with BERTPreprint
Verma, A., & Nallarasan, V. (2025) · Preprints.org · doi:10.20944/preprints202505.2360.v1
A GAN-Enhanced Approach to Bengali Intent Classification with Integrated Entity Recognition
Labib, M. A. W. A., Mahin, M. S. H., Mahin, M. R. R., Hossain, M. D., & Sultana, T. (2025) · International Conference on Electrical, Computer and Communication Engineering · doi:10.1109/ecce64574.2025.11013928
Small pre-trained model for background understanding in multi-round question answering
Huang, X., Song, H., & Lu, M. (2025) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2024.1308206
Rating Expectations Based Analysis for Predicting Hotel Review Sentiment
Kurniawan, A. F., Samo, R., Sungkono, K. R., Haryono, A. T., & Sunaryo, D. (2025) · International Conference on Computer Sciences, Engineering, and Technology Innovation (ICoCSETI) · doi:10.1109/icocseti63724.2025.11019085
Lubwama, B., Ontiveros, J., Carlyle, R. C., Kumar, S., Berkane, T., Puri, A., Tregub, A., Nitirahardjo, C., Morgan, E., Lawler, B. C., et al. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5276034
Transforming Bengali Emails: Efficient Multi-Class Classification Using Transformers
Mahin, R. B., & Rahman, M. O. (2024) · International Conference on Computer and Information Technology (ICCIT) · doi:10.1109/iccit64611.2024.11022329
Exploring Humor Detection in Bengali: A Case Study Leveraging Transformer Models
Roza, I. I., Abid, F., Nishat, T. T., et al. (2024) · International Conference on Computer and Information Technology (ICCIT) · doi:10.1109/iccit64611.2024.11021927
Zhang, X., Zhang, R. ‐., Xing, X., Zhou, S., & Chen, J. (2024) · Asian Conference on Artificial Intelligence Technology (ACAIT) · doi:10.1109/acait63902.2024.11022264
Can AI Tell More than the Available Abstracts?
Zhou, H., & Huan, X. (2024) · International Conference on Sustainable Technology and Engineering (i-COSTE) · doi:10.1109/i-coste63786.2024.11024734
A Cost-Optimization Approach of Placing Student Models in Edge Networks
Zeng, W., Zheng, D., Xing, H., Chen, X., Ren, J., Zhang, Y., Peng, C., & Cao, X. (2025) · World Conference on Computing and Communication Technologies (WCCCT) · doi:10.1109/wccct65447.2025.11027982
Social Transmission along Multiple Pathways Promotes Information Fidelity and Reduces DivisivenessPreprint
Neu, Z., Garcia‐Bernardo, J., Hardy, M., Santoro, A., Morgan, A. C., Alorić, A., Griffiths, T. L., & Krafft, P. M. (2025) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/dg8zj_v1
MTR-MSE: Motion-Text Retrieval Method Based on Motion Semantics Expansion
Xu, D., Zheng, T., Zhang, Y., Yang, X., & Fu, W. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.130632 · Cited by 1
GenAI at the Edge: Comprehensive Survey on Empowering Edge DevicesPreprint
Navardi, M., Aalishah, R., Fu, Y., Lin, Y., Li, H., Chen, Y., & Mohsenin, T. (2025) · Qeios · doi:10.1609/aaaiss.v5i1.35586 · Cited by 29
Fully Quantized Matrix Arithmetic-Only BERT Model and Its FPGA-Based Accelerator
Fuketa, H., Katashita, T., Hori, Y., & Hioki, M. (2025) · IEEE Access · doi:10.1109/access.2025.3581957
MindRon: An AI-Driven NLP Framework for Story Analysis, Summarization, and Emotion Recognition
Kundu, P., & Shanmuganathan, C. (2025) · doi:10.1109/icdsbs63635.2025.11031723
Decoding revision mechanisms in Wikipedia: Collaboration, moderation, and collectivities
Zhang, X. (2025) · New Media & Society · doi:10.1177/14614448251336418
Masker: Speculative Decoding for Long-Text QA with CopyHead
Zhou, Y., Pan, H., Liu, R., & Zhang, L. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-6954-7_3
A centroid-based fine-tuning method for out-of-scope classification
Cai, X., Tsai, P., Zhang, Y., Tian, J., Zhang, K., Yu, K., Xiao, H., & Chen, J. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.130828
Classification of Economic Activities in Indonesia Using IndoBERT Language Model
Syazali, M. R., & Yulianti, E. (2025) · Jurnal Ilmu Komputer dan Informasi · doi:10.21609/jiki.v18i2.1446 · Cited by 5
DiffBCE: Difference contrastive learning for binary code embeddings
Zhang, Y., & Cheng, G. (2025) · Information and Software Technology · doi:10.1016/j.infsof.2025.107822
Social context in political stance detection: Impact and extrapolation
Cox, R. A. V., Williams, E., & Carley, K. M. (2025) · PLoS ONE · doi:10.1371/journal.pone.0324697
Black-box model functionality stealing for Vietnamese sentiment analysis
Pham, C., Do, V.-B., Hoang, T. N., et al. (2025) · Journal of Military Science and Technology · doi:10.54939/1859-1043.j.mst.104.2025.144-154
Kengne, D., Nkambou, R., Tato, A. A. N., & Lacourarie, C. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-8889-0_36 · Cited by 1
Automatic Detection and Identification of Causal Relationships in Polish Legal Texts
Kurant, Ł. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-97557-8_15
NLP Security and Ethics, in the Wild
Lent, H., Galinkin, E., Chen, Y., Pedersen, J. M., Derczynski, L., & Bjerva, J. (2025) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00762 · Cited by 1
Distillation-Enhanced Medical Diagnosis And Clinic Referral With Fine-Tuned Gemma-3-12b
İşeri, İ., Yeniçeri, B., & Tarar, C. (2025) · Cognitive Models and Artificial Intelligence Conference (AICCONF) · doi:10.1109/aicconf64766.2025.11064054
AI-Powered Virtual Mental Health Assistant for Early-Stage NLP-Based Mental Health Screening
Loy, D. Y. K., Yau, P. C., & Wong, D. (2025) · Cognitive Models and Artificial Intelligence Conference (AICCONF) · doi:10.1109/aicconf64766.2025.11063866 · Cited by 1
Smart Assistance for Linux Beginners: A MultiModel Chatbot Approach
Shetty, B., Solapure, S., Adhyapak, S., et al. (2025) · International Conference on Electronics, Computing, Communication and Control Technology (ICECCC) · doi:10.1109/iceccc65144.2025.11063986
An Automated MCQ Generator using NLP
Neupane, A., Chaudhari, S., & Shah, S. (2025) · doi:10.38124/ijisrt/25jun1878
Dikshit, S., Dixit, R., Tiwari, R., & Jain, P. (2025) · SN Computer Science · doi:10.1007/s42979-025-04146-3 · Cited by 1
Kongthon, A., & Sae-Lim, P. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5352102
A Novel Approach to Web Article Summarization
Suresh Babu, P., Kavya, V., S, V. G., Harshitha, B., & Ashritha, D. (2025) · International Conference on Inventive Computing and Informatics (ICICI) · doi:10.1109/icici65870.2025.11069807 · Cited by 1
Sentiment Classification with Large Language Models: A Study on Generalization and Performance
Verma, B., Meel, P., & Vishwakarma, D. K. (2025) · International Conference on Inventive Computing and Informatics (ICICI) · doi:10.1109/icici65870.2025.11069494
LLM-Driven Data Augmentation for Visual Question Answering
Boussaid, H., Kwon, N., Kurtz, C., et al. (2025) · Joint Urban Remote Sensing Event (JURSE) · doi:10.1109/jurse60372.2025.11076083
A One-class variational autoencoder for smart contract vulnerability detection
Guan, S., & Law, N.-F. (2025) · International Journal of Information Security · doi:10.1007/s10207-025-01102-3
Abdeljaber, L., Alatrush, N., Alsarra, S., Billah, M., Alshammari, A. S., & Khan, L. R. (2025) · IEEE International Conference on Healthcare Informatics (ICHI) · doi:10.1109/ichi64645.2025.00033
Rastakhiz, F. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5346208
Pınar, M., & Girgin, A. B. A. (2025) · DÜMF Mühendislik Dergisi · doi:10.24012/dumf.1643370
Liu, Y., Wong, A., Fong, M. W. M., Metts, C. L., Shi, Y., & Lee, S. I. (2026) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2026.3688513
Citation Recognition in Large-Scale Legal Platforms Using Transformer Models
Eggmann, M., Fitz, J., Glasl, D., & Purandare, M. (2025) · IEEE Swiss Conference on Data Science (SDS) · doi:10.1109/sds66131.2025.00017
Generative Models Reimagined: Review into Resource-Efficient Training
Teke, A., & Pise, N. N. (2025) · International Conference on Control, Communication and Computing (ICCC) · doi:10.1109/iccc64910.2025.11077170
Investigating Multimodal Empathetic Conversational AI Utilizing Ensemble Learning and HumorPreprint
Mehta, D. B. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5356655
Gjorgjevikj, A., Mishev, K., Trajanov, D., & Kocarev, L. (2025) · IEEE Access · doi:10.1109/access.2025.3595894
Tapajós, G., de Melo, T., Guedes, E. B., & da Siva, F. S. (2025) · Revista Eletrônica de Iniciação Científica em Computação · doi:10.5753/reic.2025.6077
Proposing an NLP-Based Model for Multi-Class Classification of Holland Personality
Anggraini, E. K., Alamsyah, A., & Dudija, N. (2025) · International Conference on Electronics Representation and Algorithm (ICERA) · doi:10.1109/icera66156.2025.11087300
The Semantic Underpinnings of Speech Disorganization in SchizophreniaPreprint
Fradkin, I., Moran, R., Adams, R. A., Jefferies, E., & Dolan, R. J. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.08.10.669589
Development of Microcontroller-Based AI Robot Tour Guide Utilizing Custom Language Models
Jackson, I. S., Ballard, A. W., Hefeida, M., Srivastava, A. K., & Gyawali, P. (2025) · IEEE World AI IoT Congress (AIIoT) · doi:10.1109/aiiot65859.2025.11105214 · Cited by 1
Fine-Tuned Mistral Model for Multi-Agent Mental Health Counseling System
Patel, A., Lohumi, P., Shah, V., Dash, M., & Arya, M. S. (2025) · International Conference on Computing and Artificial Intelligence (ICCAI) · doi:10.1109/iccai66501.2025.00056
Multi-View Echocardiographic Embedding for Accessible AI DevelopmentPreprint
Tohyama, T., Han, A., Yoon, D., Paik, K., Gow, B. J., Izath, N., Kpodonu, J., & Celi, L. A. (2025) · medRxiv · doi:10.1101/2025.08.15.25333725 · Cited by 1
Classifier Performance on Long‐Tail Distributions
Pak, A., Tezekbayev, M., Bolatov, A., Melnykov, I. V., Dauletbek, K., & Assylbekov, Z. (2025) · Statistical Analysis and Data Mining The ASA Data Science Journal · doi:10.1002/sam.70037
Vronka, D. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-02406-0_17
Real-Time News Story IdentificationPreprint
Škvorc, T., Ivačič, N., Hribar, S., & Robnik‐Šikonja, M. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5404603
Fuzzy-DistilBERT: A Novel Framework for Opinion Mining Using Fuzzified Input Features
Es-sabery, F., Es-sabery, I., Es-sabery, K., et al. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-031-94623-3_20
Women are too easily offended, but robots aren't: a feminist critique of sentiment analysis
Doña-Falcón, D., & Bravo, P. M. (2025) · AI & Society · doi:10.1007/s00146-025-02582-6
Metadata
- Authors
- Sanh, V., Debut, L., Chaumond, J., & Wolf, T.
- Venue
- arXiv
- Publisher
- arXiv
- Published
- 2019
- ISSN
- 2331-8422 (ISSN-L)
- arXiv
- 1910.01108
- OpenAlex
- W2978017171
- Cited by
- 4,653
- References
- 18
- Citations by year
- 2026: 67 · 2025: 443 · 2024: 801 · 2023: 1136 · 2022: 736 · 2021: 968 · 2020: 486 · 2019: 15
- Type
- posted content
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Transactions on Graph Data and Knowledge (cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- DistilBERT, knowledge distillation, BERT, pre-training, model compression, on-device computation, language modeling, transfer learning
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-23, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-23 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
