English
Related papers

Related papers: Compute Optimal Tokenization

200 papers

We explore the impact of parameter sparsity on the scaling behavior of Transformers trained on massive datasets (i.e., "foundation models"), in both vision and language domains. In this setting, we identify the first scaling law describing…

Machine Learning · Computer Science 2023-09-18 Elias Frantar , Carlos Riquelme , Neil Houlsby , Dan Alistarh , Utku Evci

The impact of subword tokenization on language model performance is well-documented for perplexity, with finer granularity consistently reducing this intrinsic metric. However, research on how different tokenization schemes affect a model's…

Computation and Language · Computer Science 2025-08-12 Nishant Luitel , Nirajan Bekoju , Anand Kumar Sah , Subarna Shakya

We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus…

Reasoning is an integral part of many tasks performed by language models (LMs). However, the effects of scaling model sizes and data on reasoning abilities at pretraining time remain understudied. To rigorously investigate this problem, we…

Artificial Intelligence · Computer Science 2025-09-30 Xinyi Wang , Shawn Tan , Shenbo Xu , Mingyu Jin , William Yang Wang , Rameswar Panda , Yikang Shen

We introduce compression laws for language language models (LLMs). While recent scaling laws have sought to understand how LLMs scale with respect to model size, pre-training data, and computational resources, we focus on understanding how…

Computation and Language · Computer Science 2025-04-08 Ayan Sengupta , Siddhant Chaudhary , Tanmoy Chakraborty

Recent work has identified simple empirical scaling laws for language models, linking compute budget, dataset size, model size, and autoregressive modeling loss. The validity of these simple power laws across orders of magnitude in model…

Machine Learning · Statistics 2021-09-27 Amélie Chatelain , Amine Djeghri , Daniel Hesslow , Julien Launay , Iacopo Poli

Byte-Pair Encoding (BPE) is a widely used method for subword tokenization, with origins in grammar-based text compression. It is employed in a variety of language processing tasks such as machine translation or large language model (LLM)…

Data Structures and Algorithms · Computer Science 2024-11-14 László Kozma , Johannes Voderholzer

Large language models with a huge number of parameters, when trained on near internet-sized number of tokens, have been empirically shown to obey neural scaling laws: specifically, their performance behaves predictably as a power law in…

Machine Learning · Computer Science 2022-11-01 Alexander Maloney , Daniel A. Roberts , James Sully

There are growing interests in adapting large-scale language models using parameter-efficient fine-tuning methods. However, accelerating the model itself and achieving better inference efficiency through model compression has not been…

Since compute grows much faster than web text available for language model pre-training, we ask how one should approach pre-training under fixed data and no compute constraints. We first show that existing data-constrained approaches of…

Machine Learning · Computer Science 2025-09-19 Konwoo Kim , Suhas Kotha , Percy Liang , Tatsunori Hashimoto

Tokenization is widely used in large language models because it significantly improves performance. However, tokenization imposes several disadvantages, such as performance biases, increased adversarial vulnerability, decreased…

Computation and Language · Computer Science 2024-10-08 Kevin Slagle

Adapting language models to new data distributions by simple finetuning is challenging. This is due to the rigidity of their subword tokenizers, which typically remain unchanged during adaptation. This inflexibility often leads to…

Computation and Language · Computer Science 2026-05-14 Abraham Toluwase Owodunni , Orevaoghene Ahia , Sachin Kumar

Training compute is increasingly outpacing the availability of high-quality data. This shifts the central challenge from optimal compute allocation to extracting maximum value from limited data. The widely adopted Chinchilla scaling law…

Machine Learning · Computer Science 2026-05-05 Justin Lovelace , Christian Belardi , Srivatsa Kundurthy , Shriya Sudhakar , Kilian Q. Weinberger

The performance of embodied agents has been shown to improve by increasing model parameters, dataset size, and compute. This has been demonstrated in domains from robotics to video games, when generative learning objectives on offline…

Machine Learning · Computer Science 2024-12-19 Tim Pearce , Tabish Rashid , Dave Bignell , Raluca Georgescu , Sam Devlin , Katja Hofmann

Recent works have shown that machine learning models improve at a predictable rate with the total amount of training data, leading to scaling laws that describe the relationship between error and dataset size. These scaling laws can help…

Machine Learning · Computer Science 2024-06-03 Ian Covert , Wenlong Ji , Tatsunori Hashimoto , James Zou

Learning, prediction, and compression are intimately connected: a model that accurately predicts the next symbol in a sequence can be coupled with a source coder to compress that sequence near its information-theoretic limit. When tokenized…

Information Theory · Computer Science 2026-05-05 Vishnu Teja Kunde , Jean-Francois Chamberland , Krishna R. Narayanan , Jamison Ebert

The assumption across nearly all language model (LM) tokenization schemes is that tokens should be subwords, i.e., contained within word boundaries. While providing a seemingly reasonable inductive bias, is this common practice limiting the…

Computation and Language · Computer Science 2025-08-28 Alisa Liu , Jonathan Hayase , Valentin Hofmann , Sewoong Oh , Noah A. Smith , Yejin Choi

This work investigates the optimal allocation of inference compute across three key scaling factors in video vision language models: language model size, frame count, and the number of visual tokens per frame. While prior works typically…

Computer Vision and Pattern Recognition · Computer Science 2025-05-27 Peiqi Wang , ShengYun Peng , Xuewen Zhang , Hanchao Yu , Yibo Yang , Lifu Huang , Fujun Liu , Qifan Wang

Generative language models define distributions over sequences of tokens that can represent essentially any combination of data modalities (e.g., any permutation of image tokens from VQ-VAEs, speech tokens from HuBERT, BPE tokens for…

The number of tokens it takes to encode parallel text in different languages is known to vary. These disparities are called token premiums. Having high token premiums leads to less throughput during training and increases costs at…

Computation and Language · Computer Science 2025-10-28 Catherine Arnett , Tyler A. Chang , Stella Biderman , Benjamin K. Bergen