Pith. sign in

REVIEW 17 cited by

Dolma: an Open Corpus of Three Trillion Tokens for Language Model Pretraining Research

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2402.00159 v2 pith:45QXWJA7 submitted 2024-01-31 cs.CL

classification cs.CL
keywords datadolmalanguageresearchcurationmodelmodelspretraining
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Information about pretraining corpora used to train the current best-performing language models is seldom discussed: commercial models rarely detail their data, and even open models are often released without accompanying training data or recipes to reproduce them. As a result, it is challenging to conduct and advance scientific research on language modeling, such as understanding how training data impacts model capabilities and limitations. To facilitate scientific research on language model pretraining, we curate and release Dolma, a three-trillion-token English corpus, built from a diverse mixture of web content, scientific papers, code, public-domain books, social media, and encyclopedic materials. We extensively document Dolma, including its design principles, details about its construction, and a summary of its contents. We present analyses and experimental results on intermediate states of Dolma to share what we have learned about important data curation practices. Finally, we open-source our data curation toolkit to enable reproduction of our work as well as support further research in large-scale data curation.

Discussion (0). Sign in to comment.

Forward citations

Cited by 17 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Domain-Aware Scaling Laws Uncover Data Synergy

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Domain-aware scaling laws with fitted γ and σ synergy terms recover stable code-math interactions from observational LLM mixtures and correctly predict mixture rankings in controlled small-scale trainings.

  2. Small edits, large models: How Wikipedia advocacy shapes LLM values

    cs.CL 2026-04 unverdicted novelty 7.0 of 10

    Wikipedia edits by animal welfare advocates measurably influence LLM outputs on animal welfare topics, shown via retrieval and gradient attribution plus fine-tuning experiments.

  3. OLMoASR: Open Models and Data for Training Robust Speech Recognition Models

    cs.SD 2025-08 conditional novelty 7.0 of 10

    An open 1M-hour English speech dataset plus Whisper-architecture models trained on it match Whisper's word error rates on short and long-form benchmarks.

  4. Moir: Let the Model Direct Its Own Story for Robust Cross-Domain Knowledge Editing

    cs.CL 2026-05 reject novelty 6.0 of 10

    MOIR: estimating the preservation covariance from a model's own random-token generations reduces collapse of math/code capabilities in some knowledge-editing settings, but the claimed consistency is not supported by t...

  5. From Context Shift to Stylistic Collapse: Why Training Objectives Matter More Than Scale

    cs.CL 2026-04 conditional novelty 6.0 of 10

    Stylistic collapse in LLMs is largely pre-alignment and scale-independent; strong entropy regularization reduces it far more than weak smoothing or RLHF.

  6. LLM generation novelty through the lens of semantic similarity

    cs.LG 2025-10 conditional novelty 6.0 of 10

    An output is called novel when no semantically similar pretraining context can be retrieved; using this embedding-based measure, SmolLM models show longer-range reuse and instruction tuning increases novelty.

  7. User Privacy and Large Language Models: An Analysis of Frontier Developers' Privacy Policies

    cs.CY 2025-09 conditional novelty 6.0 of 10

    All six leading U.S. AI chatbot developers, as of May 2025, appear to train their models on users' chat data by default, often without clear opt-out options.

  8. BlockFFN: Towards End-Side Acceleration-Friendly Mixture-of-Experts with Chunk-Level Activation Sparsity

    cs.LG 2025-07 conditional novelty 6.0 of 10

    A ReLU-routed MoE with chunk-level sparsity training objectives and custom kernels combining activation sparsity with speculative decoding achieves over 70% 8-token chunk sparsity and up to 3.67x end-side speedup.

  9. FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language

    cs.CL 2025-06 conditional novelty 6.0 of 10

    An adaptive, per-language data filtering and deduplication pipeline produces multilingual LLM pre-training corpora that beat prior public datasets on 11 of 14 evaluated languages, and a 20TB, 1,868 language-script dat...

  10. Thunder-LLM: Efficiently Adapting LLMs to Korean with Minimal Resources

    cs.CL 2025-06 conditional novelty 6.0 of 10

    A cost-effective recipe consisting of tokenizer extension, continual pretraining, FP8 training, and SFT/DPO post-training yields Korean-English bilingual 8B models with top Korean benchmark scores.

  11. Essential-Web v1.0: 24T tokens of organized web data

    cs.CL 2025-06 conditional novelty 6.0 of 10

    A 24T-token web corpus with automatic document-level taxonomy labels enables competitive domain-specific datasets via simple filters.

  12. BehaviorBox: Automated Discovery of Fine-Grained Performance Differences Between Language Models

    cs.CL 2025-06 conditional novelty 6.0 of 10

    A pipeline that combines contextual embeddings with two LMs' per-word probabilities and sparse autoencoders to automatically find interpretable slices where one model outperforms another.

  13. Using Scaling Laws for Data Source Utility Estimation in Domain-Specific Pre-Training

    cs.LG 2025-07 conditional novelty 5.0 of 10

    Running multiple short annealing runs at different token scales can reveal per-source utility scaling curves that change data-source rankings compared with single point estimates.

  14. A New Pair of GloVes

    cs.CL 2025-07 conditional novelty 5.0 of 10

    New 2024 GloVe embeddings, trained on updated Wikipedia, Gigaword, and a Dolma subset, match the 2014 models on structural benchmarks while improving NER on temporally recent and non-Western data.

  15. FEVO: Financial Knowledge Expansion and Reasoning Evolution for Large Language Models

    cs.AI 2025-07 reject novelty 5.0 of 10

    FEVO applies continued pretraining, supervised fine-tuning, and reinforcement learning to a 32B model, claiming state-of-the-art financial benchmark scores, but the evaluation appears contaminated by training data.

  16. CCI4.0: A Bilingual Pretraining Dataset for Enhancing Reasoning in Large Language Models

    cs.CL 2025-06 reject novelty 5.0 of 10

    A new 35 TB bilingual pretraining dataset with 4.5 billion chain-of-thought templates is described, but the evidence for its benefits is marginal, confounded, and contradicted by the paper's own tables.

  17. From Data to Device: ELMOD An Efficient German-First 2.7B Language Model for Mobile Inference

    cs.CL 2026-07 conditional novelty 4.0 of 10

    A 2.7B German-first LLM trained cheaply on public data with language-specific quality filtering matches larger 7B models on German reasoning benchmarks and runs on-device.

Pith tools