LoRA fine-tuning scales brain encoding to 104B-parameter multimodal backbones
mindlab-research · hf · 2026-10-20
A paper from Columbia, Mind Lab, Tsinghua, and HKUST(GZ): "Scaling Parameter-Efficient Fine-Tuning for Brain Encoding" (arXiv cs.LG, Oct 19, 2026):
- Problem: brain encoding models predict neural responses to stimuli; the standard approach fits readouts on frozen backbones, but scarce, costly neural data makes training or fully fine-tuning large models hard
- Method: parameter-efficient fine-tuning with low-rank adapters (LoRA) plus a lightweight readout; exploits the backbone's causal structure to predict multiple timepoints in a single pass instead of sliding windows, slashing feature-extraction cost
- Results: adapts multimodal backbones up to 104B parameters — the largest model adapted for brain encoding to date — consistently improving performance across nine datasets
Takeaway: PEFT makes very large multimodal backbones efficient and interpretable for brain encoding. (ai4science)
More from Research
- HumanoidPose recovers humanoid robots' full-body joint states from RGB video alone — cholz · 2026-10-23
- Sophont's Holon-S1 predicts genes and biomarkers from routine pathology slides — iScienceLuvr · 2026-10-23
- Glasgow study: raised eyebrows and slow blinks make speakers look confident — Limor_Raviv · 2026-10-23
- IFM open-sources RL360, the agentic RL infra behind K2 Horizon's training — rupspace · 2026-10-23
- Wayfarer discovers RL options from a single experience stream, SOTA on hard Atari games — MarlosCMachado · 2026-10-23
- PhysVidBench accepted at EMNLP 2026: do video gen models grasp physical commonsense? — aykuterdemml · 2026-10-23