在LLM空间中骑行脑波:利用个体神经特征理解激活模式
Riding Brainwaves in LLM Space: Understanding Activation Patterns Using Individual Neural Signatures
AI总结:
研究通过训练个性化线性探针,发现冻结的LLM表示能编码个体特定的EEG信号,且在高伽马功率上表现显著优于群体探针。
摘要:
消费者级EEG设备正进入日常设备,如耳塞和头带,引发语言模型能否适应个体神经反应的问题。我们通过测试冻结LLM表示是否编码特定个体的EEG信号,即激活空间中预测某人脑活动但非另一人活动的方向。使用30名参与者阅读自然句子(ZuCo语料库)的词级EEG数据,我们为每个人训练独立的线性探针,将冻结的Qwen 2.5 7B隐藏状态映射到该个体的EEG功率。个体特定探针在所有测试的EEG特征上均优于单一群体探针;在高伽马功率上,个体特定探针的rho值为0.183,比群体探针(rho=0.020,p<10^-4)提高了九倍。负对照“注视计数”显示无个体特定优势(p=0.360);注视计数反映词长和频率而非个体认知。个体方向在时间上稳定(分半余弦=0.824),无法跨个体转移(自我rho=0.369 vs. 其他rho=0.143,p<10^-19),且与共享群体信号不同:在移除群体成分后,个体特定探针仍保留预测能力。个体信号集中在模型的深层,随着深度增加而持续上升,峰值出现在28层的第24层。结果在不同架构(LLaMA 3.1 8B)中一致,并在词级混杂控制下存活。冻结语言模型在其深层包含稳定的、个体特定的神经方向,为EEG驱动的个性化提供了几何基础。
英文摘要:
Consumer-grade EEG is entering everyday devices, from earbuds to headbands, raising the question of whether language models can be adapted to individual neural responses. We test this by asking whether frozen LLM representations encode person-specific EEG signals, directions in activation space that predict one person's brain activity but not another's. Using word-level EEG from 30 participants reading naturalistic sentences (ZuCo corpus), we train a separate linear probe for each person, mapping hidden states from a frozen Qwen 2.5 7B to that individual's EEG power. Person-specific probes outperform a single population probe on every EEG feature tested; for high-gamma power, the person-specific probe achieves rho = 0.183, a ninefold improvement over the population probe (rho = 0.020, p < 10^-4). A negative control, fixation count, shows no person-specific advantage (p = 0.360); fixation count reflects word length and frequency rather than individual cognition. The individual directions are temporally stable (split-half cosine = 0.824), non-transferable across people (self rho = 0.369 vs. other rho = 0.143, p < 10^-19), and distinct from the shared population signal: person-specific probes retain predictive power after the population component is removed. The person-specific signal concentrates in the model's deep layers, rising consistently with depth and peaking at Layer 24 of 28. The results are consistent across architectures (LLaMA 3.1 8B) and survive word-level confound controls. Frozen language models contain stable, person-specific neural directions in their deep layers, providing a geometric foundation for EEG-driven personalization.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。