WARP recovers training domain mixtures from fine-tuned model weights using weight-space interpolation via model merging to generate pseudo-checkpoints and geometric features mapped to proportions.
hub
Efficient online data mixing for language model pre-training
11 Pith papers cite this work. Polarity classification is still indexing.
11
Pith papers citing it
hub tools
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
Power-law LLM scaling laws are largely the aggregate of stepwise token learning events whose heavy-tailed learning-time spectrum reconstructs loss derivatives along step, data, and model axes.
Under a shared-head/disjoint-tail assumption, multi-domain loss decomposes into a capacity-competition term c_i x_i^*(h)^{-b_i} plus a per-domain noise term A_i(Dh_i)^{-a_i}, and the fitted law extrapolates optimal mixtures to unseen scales.
Repetition rate mismatch between small-scale proxies and target budgets is the main reason data mixture experiments do not scale; a subsampling procedure that equalizes repetition rates recovers optimal mixtures from 1/16-scale experiments.
DRATS derives a minimax objective from a feasibility formulation of MTRL to adaptively sample tasks with the largest return gaps, leading to better worst-task performance on MetaWorld benchmarks.
MixAtlas uses CLIP-based decomposition and Gaussian process optimization on small proxies to discover data mixtures that improve multimodal benchmark performance by up to 17.6% and transfer to larger models with faster convergence.
DRIFT applies on-policy influence functions with signed weighting and debiasing to attribute and refine SFT data, raising performance on 7B instruction and reasoning models over prior curation methods.
DUET is a global-to-local method that optimizes LLM training data mixtures via Bayesian optimization guided by influence-based selection and feedback from unseen evaluation tasks, with a regret bound showing convergence to the optimal mixture.
DCLM-Baseline dataset lets a 7B model reach 64% 5-shot MMLU accuracy after 2.6T tokens, beating prior open-data models by 6.6 points on MMLU with 40% less compute.
HDS uses Soft Actor-Critic RL with a multi-objective reward (data quality, inter-domain loss influence, weight norms) for online data mixing in LLM pre-training, reaching target perplexity with 44% fewer iterations and 7.2% MMLU gain on The Pile.
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.
citing papers explorer
-
WARP: Weight-Space Analysis for Recovering Training Data Portfolios
WARP recovers training domain mixtures from fine-tuned model weights using weight-space interpolation via model merging to generate pseudo-checkpoints and geometric features mapped to proportions.
-
Smooth Scaling Laws Hide Stepwise Token Learning
Power-law LLM scaling laws are largely the aggregate of stepwise token learning events whose heavy-tailed learning-time spectrum reconstructs loss derivatives along step, data, and model axes.
-
Explaining Data Mixing Scaling Laws
Under a shared-head/disjoint-tail assumption, multi-domain loss decomposes into a capacity-competition term c_i x_i^*(h)^{-b_i} plus a per-domain noise term A_i(Dh_i)^{-a_i}, and the fitted law extrapolates optimal mixtures to unseen scales.
-
Repetition Mismatch: Why Data Mixture Experiments Don't Scale and How to Fix Them
Repetition rate mismatch between small-scale proxies and target budgets is the main reason data mixture experiments do not scale; a subsampling procedure that equalizes repetition rates recovers optimal mixtures from 1/16-scale experiments.
-
Distributionally Robust Multi-Task Reinforcement Learning via Adaptive Task Sampling
DRATS derives a minimax objective from a feasibility formulation of MTRL to adaptively sample tasks with the largest return gaps, leading to better worst-task performance on MetaWorld benchmarks.
-
MixAtlas: Uncertainty-aware Data Mixture Optimization for Multimodal LLM Midtraining
MixAtlas uses CLIP-based decomposition and Gaussian process optimization on small proxies to discover data mixtures that improve multimodal benchmark performance by up to 17.6% and transfer to larger models with faster convergence.
-
DRIFT: Refining Instruction Data via On-Policy Data Attribution
DRIFT applies on-policy influence functions with signed weighting and debiasing to attribute and refine SFT data, raising performance on 7B instruction and reasoning models over prior curation methods.
-
DUET: Optimizing Training Data Mixtures via Feedback from Unseen Evaluation Tasks
DUET is a global-to-local method that optimizes LLM training data mixtures via Bayesian optimization guided by influence-based selection and feedback from unseen evaluation tasks, with a regret bound showing convergence to the optimal mixture.
-
DataComp-LM: In search of the next generation of training sets for language models
DCLM-Baseline dataset lets a 7B model reach 64% 5-shot MMLU accuracy after 2.6T tokens, beating prior open-data models by 6.6 points on MMLU with 40% less compute.
-
Holistic Data Scheduler for LLM Pre-training via Multi-Objective Reinforcement Learning
HDS uses Soft Actor-Critic RL with a multi-objective reward (data quality, inter-domain loss influence, weight norms) for online data mixing in LLM pre-training, reaching target perplexity with 44% fewer iterations and 7.2% MMLU gain on The Pile.
-
Data Mixing for Large Language Models Pretraining: A Survey and Outlook
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.