lab NVIDIA Research · 1h ago · research

Source introduces approximate unrolled differentiation for training data attribution

from English

Researchers introduce Source, a new training data attribution (TDA) method that combines the computational efficiency of influence functions with the accuracy of unrolling-based approaches. By using an influence-function-like formula to approximate unrolled differentiation, Source addresses limitations in implicit differentiation methods, such as their failure to account for optimization bias or multi-stage pipelines.

Source is computationally efficient compared to full unrolling-based TDA methods.
It remains suitable for non-converged models and multi-stage training pipelines where implicit-differentiation approaches struggle.
Empirical results show Source outperforms existing TDA techniques in counterfactual prediction, particularly in settings where traditional methods fall short.

Importance 1/3 Trust 3/3 NVIDIA Research Research paper Training methods

Read original

lab NVIDIA Research · just now · 24 views Live

LLMs match or beat Bayesian optimization for hyperparameter optimization

A new paper explores using foundational large language models (LLMs) to automate hyperparameter optimization (HPO), a process that typically relies on manual approaches in limited-budget settings. The authors developed a methodology where LLMs suggest hyperparameter configurations based on dataset and model descriptions, which are then iteratively refined according to model performance.

lab NVIDIA Research · 1h ago Live

JacNet learns functions by directly modeling structured Jacobians

The paper introduces JacNet, a neural network architecture that directly learns the Jacobian of an input-output function rather than the mapping itself. This approach allows for precise control over derivative properties, enabling the enforcement of structural priors such as invertibility and k-Lipschitz continuity.

arxiv arXiv cs.CL · 8h ago

MiLMMT-46-v1.0 improves multilingual translation via reference-free post-training

Researchers have released MiLMMT-46-v1.0, an open large language model for multilingual machine translation that utilizes reference-free post-training. Starting from the supervised-finetuned MiLMMT-46-v0.1, the team applied Group Relative Policy Optimization (GRPO) with a reward based on two reference-free quality estimation models gated by language identification.

lab NVIDIA Research · 1d ago · 2 views

Masters distills vision-language models via masking teacher and reinforcing student

Researchers propose Masters, a mask-progressive reinforcement learning framework for distilling large-scale vision-language models into compact versions suitable for edge deployment. The method addresses the instability caused by the size gap between teacher and student models.

lab NVIDIA Research · 1d ago · 2 views

ZPPO uses prompts instead of gradients to improve small vision-language model training

Researchers introduce Zone of Proximal Policy Optimization (ZPPO), a method that injects teacher knowledge into prompts rather than policy gradients to address brittleness in knowledge distillation and drift in reinforcement learning. ZPPO constructs Binary Candidate-included Questions (BCQ) and Negative Candidate-included Questions (NCQ) for hard questions, recirculating them via a replay buffer until the student's accuracy improves or they are evicted.