SCORE: Subject Coordinate Recovery for Label-Free Cross-Subject EEG-to-Image Retrieval

Forum topic · 小凯 · 2026-08-21

Summary

SCORE (Subject Coordinate Recovery) is a target label-free framework for EEG-to-image retrieval that addresses the performance drop new users face without labeled calibration. The authors analyze cross-subject EEG features and find that different subjects preserve similar relationships among concepts but express them along different coordinate directions. SCORE combines recovery-aware source training with coordinate alignment at deployment: during training, source-subject EEG is aligned to a common image space while simulating unseen-subject recovery using only source data; at deployment, both encoders are frozen, reliable EEG-image landmarks are selected via hubness-corrected matching, and an orthogonal transform recovers the target subject's EEG coordinates—requiring no source data or target labels. On 200-way retrieval over two public benchmarks, SCORE outperforms unadapted baselines for every target subject and achieves the best overall accuracy: 53.23%/83.55% Top-1/Top-5 on THINGS-EEG2 and 12.01%/32.16% on Alljoined-1.6M, surpassing the strongest baselines by 17.45/15.70 and 3.08/4.62 percentage points. Paper: arXiv 2608.19134.

Paper Overview

Field: Machine Learning Authors: Zhenyao Cui, Siyuan Kan, Siyang Li, Ziwei Wang, Dongrui Wu arXiv: 2608.19134

Introduction

Accurate visual decoding can reveal how the brain represents visual information and recover perceived content from neural signals such as electroencephalography (EEG), with potential for neural communication. However, current EEG-to-image retrieval methods perform far below their within-subject counterparts for new users without labeled calibration, limiting real-world deployment.

Key Insight

Analyzing EEG features across subjects reveals that different subjects preserve similar relationships among concepts but express them along different coordinate directions. This motivates a coordinate-recovery approach rather than full retraining.

The SCORE Framework

SCORE (Subject Coordinate Recovery) is a target label-free framework combining recovery-aware source training with coordinate alignment at deployment:

Training: SCORE aligns source subject EEG with a common image space and simulates unseen-subject recovery using only source data.
Deployment: Both encoders are frozen. SCORE selects reliable EEG-image landmarks via hubness-corrected matching and estimates an orthogonal transform to recover the target EEG coordinates — no source data or target labels required.

Results (200-way retrieval)

| Benchmark | Top-1 | Top-5 | |---|---|---| | THINGS-EEG2 | 53.23% | 83.55% | | Alljoined-1.6M | 12.01% | 32.16% |

SCORE outperforms unadapted baselines for every target subject and achieves the best overall accuracy.
It surpasses the strongest baselines by 17.45/15.70 points (Top-1/Top-5) on THINGS-EEG2 and 3.08/4.62 points on Alljoined-1.6M.

Conclusion

Without target labels or encoder updates, SCORE brings brain-based visual decoding closer to robust, practical, low-latency cross-user deployment.