- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper asks if frozen EEG encoders keep individual-specific information on a new clinical dataset, and if that information depends on temporal order. TUEP windows go through two frozen encoders (a 1.46M-parameter CNN-Transformer and LUNA-Base), and global k-means puts them into states. State occupancy and the lag-1 transition vector are compared within and between subjects across two sessions () using a fingerprintability gap . A temporal-excess statistic (Eq. 8) compares of the original sequences with that of occupancy-preserving shuffles. Both occupancy and reach of about 0.10, but is and . The authors conclude that does not imply .
Critical strengths
-
Section 4 separates reproducible individual information () from information that depends on temporal order (). The toy sequence in Section 4.2 makes the shuffle null easy to follow. The point that a transition matrix inherits subject-specific occupancy is correct and matters for any latent-state analysis.
-
The paper reports the one setting that goes against its conclusion (LUNA-Base at , , Table 5 and Figure 6), low Top-1 retrieval and the dependence on . The limitations paragraph (Section 7, L437-447) names the single dataset, short recordings and two encoders. I recomputed as original minus permuted from Tables 1, 4 and 5; it matches the columns, so the numbers are internally consistent.
Critical weaknesses
-
I can't interpret the central negative result as it stands. The claim (abstract; L103-107; Section 8) rests on being about zero or negative (Table 1, Section 6.3). The only sanity check I found is "by construction" (Section 4.2), and that says nothing about whether the test can detect temporal individuality. A session needs only 20 windows (Section 5). So a session at that minimum gives 19 transitions for the 16 entries of . Also, uncentered cosine on non-negative vectors starts from a high baseline (between-subject 0.45, Table 1). A negative may just be noise in . The paper has no positive control (for example, synthetic sequences where subjects share occupancy but differ in transition matrices, run at the real session lengths) and gives no detectable effect size. There is also no uncertainty over subjects. The Mann-Whitney tests compare within- and between-subject similarities, and as far as I can tell the between-subject pairs reuse the same signatures, so they are not independent. The shuffle standard deviations in Table 1 (0.008 and 0.006) only cover permutation noise; one subject at moves a Top-1 rate by about 0.9 points. Without an equivalence margin and subject-level intervals, failing to reject a difference does not support "no temporal excess". The long-recording check (D.1, ) reports residual dwell, not .
-
The residual and persistence nulls still carry identity, and may partly reflect group or session similarity. In Table 4 the shuffled dwell-perp signature reaches of 0.157 and 0.194, higher than the original (0.106 and 0.118). acts the same way: 0.113 and 0.122 after shuffling, against 0.089 and 0.098. A residual near zero under shuffling should carry little identity. So I suspect finite-sample effects that depend on and session length make it fingerprintable; Appendix D.1 explains residual dwell in a similar way. If that is right, the negative values can't be read as "no temporal information". Appendix D also reports that a dwell-preserving block shuffle lowers fingerprintability, which is the opposite direction, and only describes it as preserving run lengths. Separately, Eqs. 5-6 average over all between-subject pairs, including pairs from different diagnostic groups, and 75 of the 112 subjects have epilepsy. In Table 6 (CNN-Transformer, all 198 labeled subjects) similarity is higher within epilepsy (0.605) than within non-epilepsy (0.492). So part of could come from diagnosis, age, medication or recording setup. This is the part that bothers me most. The paper also doesn't give the interval between sessions A and B.
A few smaller points. The title and abstract talk about EEG foundation models, but the audit covers two encoders on a 4-state discretisation. There is no raw-feature or randomly initialised baseline to show whether pretraining matters for the 0.10 gap. "demonstrates" (L105-107) is stronger than what a null at on one dataset supports. Appendix G (L845) calls pre-specified, even though peaks there in Table 5, and Appendix F only says that is not a unique optimum. "mLTD" (Appendix E, L775) is never defined, and the y-axis of Figure 4 reads "OUR".
Questions for authors
- What does the pipeline give on synthetic sequences with matched occupancy and subject-specific transitions, at your real session lengths?
- What are the subject-level bootstrap intervals for , and , with an equivalence margin for the claim of no temporal excess?
- How exactly is the dwell-preserving block shuffle defined? Why does it lower when the occupancy-preserving shuffle doesn't?
- If you keep only between-subject pairs from the same diagnostic group, does stay near 0.10? How does it change with the interval between sessions?
- Are any TUEP subjects or recordings in the pretraining data of either encoder? I'm not sure that TUEP and the TUEG pretraining corpus are disjoint, and Appendix A doesn't say what LUNA was pretrained on.