Zhuofan Josh Ying

Headlines · 1

1. Oct 18
   BrainLoRA adapts 100B+ multimodal backbones for brain encoding via PEFT

   An arXiv paper submitted on 19 October 2026, "Scaling Parameter-Efficient Fine-Tuning for Brain Encoding" by Zhuofan Josh Ying, Nikolaus Kriegeskorte, Hossein Adeli and six co-authors, proposes adapting very large pretrained multimodal backbones to brain encoding with low-rank adapters (LoRA) plus a lightweight readout. The authors say the approach scales to backbones of up to 104B parameters, which they describe as the largest model adapted for brain encoding to their knowledge, and that it consistently improves performance across nine multimodal backbones. They report that BrainLoRA achieves the best published result on the Algonauts 2025 Challenge on the same Qwen3-Omni backbone used by the prior state of the art, and that an ensemble of 200+ adapted models built with a multi-LoRA kernel ranks first on the public Algonauts post-challenge leaderboard at the time of writing. These benchmark claims are self-reported and not independently verified in the supplied material.

   Hugging Face · Papers · 🔥 0

Experience and discussion from the community

Share my Zhuofan Josh Ying experienceAsk about Zhuofan Josh Ying