pith. sign in

arxiv: 2504.13161 · v2 · pith:7F3HT4XHnew · submitted 2025-04-17 · 💻 cs.CL

Nemotron-CLIMB: CLustering-based Iterative Data Mixture Bootstrapping for Language Model Pre-training

classification 💻 cs.CL
keywords datamixturepre-trainingdatasetsdomainmodelnemotron-climboptimal
0
0 comments X
read the original abstract

Pre-training datasets are typically collected from web content and lack inherent domain divisions. For instance, widely used datasets like Common Crawl do not include explicit domain labels, while manually curating labeled datasets such as The Pile is labor-intensive. Consequently, identifying an optimal pre-training data mixture remains a challenging problem, despite its significant benefits for pre-training performance. To address these challenges, we propose CLustering-based Iterative Data Mixture Bootstrapping (Nemotron-CLIMB), an automated framework that discovers, evaluates, and refines data mixtures in a pre-training setting. Specifically, Nemotron-CLIMB embeds and clusters large-scale datasets in a semantic space and then iteratively searches for optimal mixtures using a smaller proxy model and a predictor. When continuously trained on 400B tokens with this mixture, our 1B model exceeds the state-of-the-art Llama-3.2-1B by 2.0%. Moreover, we observe that optimizing for a specific domain (e.g., Social Sciences) yields a 5% improvement over random sampling. Finally, we introduce Nemotron-ClimbLab, a filtered 1.2-trillion-token corpus with 20 clusters as a research playground, and Nemotron-ClimbMix, a compact yet powerful 400-billion-token dataset designed for efficient pre-training that delivers superior performance under an equal token budget. We analyze the final data mixture, elucidating the characteristics of an optimal data mixture. Our data is available at: https://research.nvidia.com/labs/lpr/climb/

This paper has not been read by Pith yet.

discussion (0)

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

Forward citations

Cited by 21 Pith papers

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

  1. DataComp-VLM: Improved Open Datasets for Vision-Language Models

    cs.CV 2026-06 conditional novelty 8.0

    DataComp-VLM benchmark shows instruction-heavy data mixing outperforms filtering for VLM training, with DCVLM-Baseline achieving 63.6% on 33 tasks for 8B models (+5.4pp over FineVision).

  2. HERMES: A Multi-Granularity Labeling Substrate for Pre-training Data Mixtures

    cs.LG 2026-07 unverdicted novelty 7.0

    HERMES provides a reusable hierarchical labeling substrate for pre-training data that reveals granularity-specific effects in data mixing rules during model training.

  3. MinGram: A Minimalist Unigram Tokenizer with High Compression and Competitive Morphological Alignment

    cs.CL 2026-06 unverdicted novelty 7.0

    MinGram is a simplified Unigram tokenizer training method that prioritizes token count minimization to deliver higher compression than BPE and standard Unigram while retaining competitive morphological alignment and s...

  4. Natural Ungrokking: Asymmetric Control of Which Rules Survive Pretraining

    cs.LG 2026-06 unverdicted novelty 7.0

    During pretraining, language models exhibit natural ungrokking where learned rules are forgotten based on their support frequency in the corpus, with asymmetric editability of rule survival.

  5. Tokenization with Split Trees

    cs.CL 2026-05 unverdicted novelty 7.0

    ToaST uses split trees and integer programming to cut token counts by over 11% versus BPE on English text at 40k+ vocab sizes, yielding higher CORE scores in 1.5B-parameter language model training.

  6. Tokenization with Split Trees

    cs.CL 2026-05 unverdicted novelty 7.0

    ToaST uses vocabulary-independent split trees and integer programming to produce tokenizers with over 11% fewer tokens than BPE, WordPiece, and UnigramLM while improving 1.5B-parameter LM scores on CORE.

  7. X-Token: Projection-Guided Cross-Tokenizer Knowledge Distillation

    cs.LG 2026-05 conditional novelty 7.0

    X-Token proposes projection-guided P-KL and H-KL losses to fix uncommon-token suppression and over-conservative matching in logit-based cross-tokenizer distillation, yielding gains over GOLD on Llama-3.2-1B.

  8. Scaling Laws for Mixture Pretraining Under Data Constraints

    cs.LG 2026-05 conditional novelty 7.0

    Repetition-aware scaling laws show scarce target data in pretraining mixtures can be repeated 15-20 times optimally, with the best count depending on data size, compute, and model scale.

  9. Data Mixing Agent: Learning to Re-weight Domains for Continual Pre-training

    cs.LG 2025-07 unverdicted novelty 7.0

    An RL agent learns domain re-weighting policies from evaluation feedback to improve balanced performance in continual pre-training of LLMs across source and target domains.

  10. DataComp-VLM: Improved Open Datasets for Vision-Language Models

    cs.CV 2026-06 unverdicted novelty 6.0

    DataComp-VLM benchmark shows instruction-heavy data mixtures outperform caption-heavy ones for VLM training, with DCVLM-Baseline reaching 63.6% on 33 tasks using 200B tokens, +5.4pp over FineVision.

  11. KletterMix: Climbing Toward High-Quality German Pretraining Data

    cs.CL 2026-06 unverdicted novelty 6.0

    KletterMix is a translated German corpus from English pretraining data that yields measurable gains on German downstream tasks in controlled pretraining experiments.

  12. Do Value Vectors in Deep Layers Need Context from the Residual Stream?

    cs.CL 2026-06 unverdicted novelty 6.0

    Deeper transformer layers benefit from context-free token-specific value vectors in a Bank of Values lookup table, improving performance over standard attention with less compute.

  13. Most Transformer Modifications Still Do Not Transfer at 1-3B: A 2020-2026 Update to Narang et al. (2021) with Downstream Evaluation and a Noise Floor

    cs.LG 2026-05 conditional novelty 6.0

    Empirical update to prior work shows most of 20 recent Transformer modifications do not transfer at 1-3B scales when measured with downstream CLIMB-12 tasks, multi-seed noise floor, and cross-scale stability.

  14. Scale-Invariant Neural Network Optimization: Norm Geometry and Heavy-Tailed Noise

    math.OC 2026-05 unverdicted novelty 6.0

    Establishes matching Ω and O(min{m,n} ε^-(3p-2)/(p-1)) bounds for scale-invariant spectral-norm methods under heavy-tailed noise, plus an improved O(min{m,n} ε^-(5p-3)/(2p-2)) rate via transported Scion under Hessian ...

  15. Scale-Invariant Neural Network Optimization: Norm Geometry and Heavy-Tailed Noise

    math.OC 2026-05 unverdicted novelty 6.0

    Establishes matching lower and upper oracle complexity bounds for scale-invariant methods with spectral norm under heavy-tailed noise, plus improved rates with higher-order smoothness, and practical tests on neural networks.

  16. Scaling Laws for Mixture Pretraining Under Data Constraints

    cs.LG 2026-05 unverdicted novelty 6.0

    Empirical study shows mixture pretraining tolerates higher target data repetition than single-source training, with a new repetition-aware scaling law enabling principled mixture selection based on data size, compute,...

  17. Scaling-Aware Data Selection for End-to-End Autonomous Driving Systems

    cs.LG 2026-04 unverdicted novelty 6.0

    MOSAIC is a scaling-aware data selection framework that outperforms baselines in training end-to-end autonomous driving planners, achieving comparable or better EPDMS scores with up to 80% less data.

  18. Capacity-Aware Mixture Law Enables Efficient LLM Data Optimization

    cs.LG 2026-03 unverdicted novelty 6.0

    CAMEL is a scaling law capturing nonlinear model-size and mixture interactions to extrapolate optimal data mixtures for large LLMs from small-model experiments, reducing optimization cost by 50% and improving benchmar...

  19. AdaptiveLoad: Towards Efficient Video Diffusion Transformer Training

    cs.DC 2026-05 unverdicted novelty 5.0

    AdaptiveLoad cuts computational imbalance in video DiT training from 39% to 18.9% and raises throughput 27.2% via memory-compute constraints and a custom LayerNorm-Modulate kernel.

  20. Small Language Models are the Future of Agentic AI

    cs.AI 2025-06 unverdicted novelty 5.0

    Small language models are sufficiently capable, more suitable, and far more economical than large models for the repetitive tasks that dominate agentic AI systems.

  21. Data Mixing for Large Language Models Pretraining: A Survey and Outlook

    cs.CL 2026-03 accept novelty 4.0

    A survey that taxonomizes data mixing strategies for LLM pretraining into static rule-based, learning-based, and dynamic adaptive families while highlighting transferability challenges and evaluation gaps.