REVIEW 13 cited by
RegMix: Data Mixture as Regression for Language Model Pre-training
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
read the original abstract
The data mixture for large language model pre-training significantly impacts performance, yet how to determine an effective mixture remains unclear. We propose RegMix to automatically identify a high-performing data mixture by formulating it as a regression task. RegMix trains many small models on diverse data mixtures, uses regression to predict performance of unseen mixtures, and applies the best predicted mixture to train a large-scale model with orders of magnitude more compute. To empirically validate RegMix, we train 512 models with 1M parameters for 1B tokens to fit the regression model and predict the best data mixture. Using this mixture we train a 1B parameter model for 25B tokens (i.e. 1000x larger and 25x longer) which we find performs best among 64 candidate 1B parameter models with other mixtures. Furthermore, RegMix consistently outperforms human selection in experiments involving models up to 7B models trained on 100B tokens, while matching or exceeding DoReMi using just 10% of the computational resources. Our experiments also show that (1) Data mixtures significantly impact performance; (2) Web corpora rather than data perceived as high-quality like Wikipedia have the strongest positive correlation with downstream performance; (3) Domains interact in complex ways often contradicting common sense, thus automatic approaches like RegMix are needed; (4) Data mixture effects transcend scaling laws. Our code is available at https://github.com/sail-sg/regmix.
Forward citations
Cited by 13 Pith papers
-
The Kinetics of Training: A Driven-Nucleation Rate Law for Emergence, Plasticity Loss, and Circuit Control in Language Models
Capability formation in small transformers is claimed to obey a driven-nucleation rate law J = Nνσ(c)e^{−βK} − D, read forward as emergence, backward as plasticity loss, and completed as circuit control.
-
RSIBench-Data: Benchmarking Data-Centric Research for Recursive Self-Improvement
A controlled benchmark shows LLM agents can sometimes discover better training-data strategies through feedback, but their improvements are fragile and usually not sustained.
-
GradAlign: Gradient-Aligned Data Selection for LLM Reinforcement Learning
GradAlign selects RL training problems whose policy gradients align with validation-set gradients, claiming more stable and better-performing LLM post-training.
-
Multi-Task GRPO: Reliable LLM Reasoning Across Tasks
MT-GRPO reweights tasks by reward and improvement and enforces those weights after zero-gradient filtering, improving worst-task accuracy by 6–28% over GRPO/DAPO baselines on 3- and 9-task setups.
-
Language Models Improve When Pretraining Data Matches Target Tasks
Ranking pretraining documents by similarity to benchmark training examples (BETR) yields consistent benchmark gains and a 2.1x compute multiplier over DCLM-Baseline.
-
Hunyuan-MT Technical Report
Hunyuan-MT and Chimera, a 7B open-source translation model and its multi-candidate fusion variant, claim state-of-the-art multilingual translation including Mandarin to minority languages, with open weights.
-
DynamixSFT: Dynamic Mixture Optimization of Instruction Tuning Collections
DynamixSFT adaptively reweights instruction-tuning datasets during training using a prior-anchored bandit and one-step loss-reduction reward, giving small but consistent benchmark gains.
-
ADMIRE-BayesOpt: Accelerated Data MIxture RE-weighting for Language Models with Bayesian Optimization
Using Bayesian optimization over Gaussian-process surrogates, data mixtures for LLM training can be found much faster than with linear or exponential regression baselines, including across model sizes.
-
Using Scaling Laws for Data Source Utility Estimation in Domain-Specific Pre-Training
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.
-
Learning Task Mixtures from Task Affinities: A Probabilistic Graphical Model for Supervised Fine-Tuning
TaskPGM optimizes a quadratic energy over task mixtures using PMI/JSD behavioral affinities, yielding mixtures that outperform naive sampling on several 7B LLM fine-tuning benchmarks.
-
BlueLM-2.5-3B Technical Report
BlueLM-2.5-3B is a small multimodal model with a switchable thinking mode that reportedly matches larger models like Qwen3-4B and comes close to Kimi-VL-A3B-16B on many benchmarks.
-
AgentScope 1.0: A Developer-Centric Framework for Building Agentic Applications
AgentScope 1.0 packages the components needed to build, evaluate, and deploy LLM agent applications into one developer framework.
-
LLM Data Selection and Utilization via Dynamic Bi-level Optimization
A bi-level-learned per-batch reweighting model improves average downstream accuracy of LLMs pre-trained on selected data by about one to two points, but without error bars or code the effect size is uncertain.
Discussion (0). Sign in to comment.