arXiv:2604.14198v1 Announce Type: cross Abstract: Domain reweighting can improve sample efficiency and downstream generalization, but data-mixture optimization for multimodal midtraining remains large

model-releasesarxiv-cs-cl

17 Apr 2026

Safety

BROS: Bias-Corrected Randomized Subspaces for Memory-Efficient Single-Loop Bilevel Optimization

DGX agent

arXiv:2605.10288v1 Announce Type: new Abstract: Stochastic bilevel optimization (SBO) has become a standard framework for hyperparameter learning, data reweighting, representation learning, and data-m

safetyarxiv-cs-lg

12 May 2026

Model Releases

Mixture-Trained Merging for Unified Multi-Objective Models

DGX agent

arXiv:2610.01238v1 Announce Type: new Abstract: Unified language models are increasingly expected to combine heterogeneous capabilities, such as mathematics, code, instruction following, and controlla

model-releasesarxiv-cs-lg

2 Oct 2026

Safety

UniReason-Med: A Shared Grounded Reasoning Interface for 2D-to-3D Transfer in Medical VQA

DGX agent

arXiv:2606.11740v1 Announce Type: cross Abstract: We study whether grounded reasoning supervision from abundant 2D medical images can improve 3D medical VQA when both input types are aligned through a

safetyarxiv-cs-cl

11 Jun 2026

Research

MixAtlas: Uncertainty-aware Data Mixture Optimization for Multimodal LLM Midtraining

DGX agent

This paper was accepted at the Workshop on Navigating and Addressing Data Problems for Foundation Models (NADPFM) at ICLR 2026. Principled domain reweighting can substantially improve sample efficienc

researchapple-ml-research

16 Apr 2026

6 results