- Rating
- 1
- Confidence
- 2
- Informativeness
- 1
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The authors propose the residual SAE (RSAE), a new activation function for sparse autoencoders that aims to outperform the activation functions currently used in SAEs.
Critical strengths
I find the idea interesting and promising. I am not familiar with every recently published SAE variant, so I will not make claims about its novelty. The theoretical results provide some support for the approach.
Critical weaknesses
- Clarity. I found the paper hard to follow. For example (and this is not the only case), Section 3 opens directly with the exact equation for RSAE, without explaining why it was constructed this way or what each part is meant to do. Theorem 1 then states that “for p ≥ 1 and q ≥ 0, the infimum of RSAE reconstruction risk is no greater than that of ReLU at any feasible expected...”. It took me several minutes to find where p and q are defined, and more time to understand why this result matters. It may well be important for the paper, but the text does not explain its significance. Why should the reader care about this infimum, and why is a theoretical comparison with ReLU the right one to make? What does this result tell us in practice? Theorem 2 relies on material from the appendix, so the main text is not self-contained, and it introduces terms that have not been defined before. Similar issues appear throughout the rest of the paper. I would encourage the authors to give some motivation before each formal statement and to define all notation in the main text.
- Experiments. This is my main concern and the one that determines my score. The evaluation is limited to the SAEBench core metrics, such as cross-entropy loss and the number of live features. From these results alone, I cannot tell whether the method is useful in practice. SAEs are a practical tool for interpretability, and practitioners use them for downstream tasks such as probing. SAEBench already includes benchmarks for these tasks, but they are not reported. As it stands, the results show a gain on one metric, and it is unclear whether this gain leads to meaningful improvements on the tasks that actually matter.
- Significance. Following from the previous point, the paper currently reads as one more SAE variant among many. A large number of SAE variants have been proposed, and only a small fraction of them hold up under practical evaluation or are worth adopting. A better cross-entropy loss at a given sparsity level is not enough, on its own, to convince me that this variant will have a significant impact. Based on the current results, I am not convinced that it will.
- Inaccurate references. Some citations appear to misdescribe the cited work. For example, the paper states that “data-free SAEs (Laptev et al., 2025) fit dictionaries directly from model weights without streaming activations.” This is not what the cited paper does. It applies more sophisticated methods to standard SAEs trained on activations. I suspect there may be other similar cases in the paper.
- Some figures are not very informative. For example, Figure 2 shows that there are configurations where RSAE gives large improvements. I do not doubt this, but such configurations can likely be found for almost any method, and selected comparisons can show nearly any size of improvement, even “+100%” in some scenarios. Without a broader view, such as results across all configurations, it is hard to draw conclusions from this figure.
Questions for authors
See weaknesses
4 comments
Response to Reviewer 9Gh3 (part 2)
Probing results beyond the core metrics. We would like to point out that a probing result is in the submission: Section 5 ("Sanity check on feature quality", lines 461–464) and Appendix B.3 / Table 4 report the 8-task SAEBench sparse-probing panel for the RSAE and its teacher (93.71% vs. 93.69%). For this response we extended it, following the other reviewers' suggestions, to SAEBench's default -sparse probing (top- latents per class, ), for all three teacher families, using the already-trained RSAEs (Pythia-160m, level).
| SAE | all latents | ||||||
|---|---|---|---|---|---|---|---|
| ReLU teacher () | 76.9 | 82.1 | 86.0 | 88.5 | 90.3 | 91.2 | 93.2 |
| RSAE (ReLU-init) | 76.3 | 80.3 | 85.5 | 88.3 | 90.1 | 91.5 | 93.3 |
| (RSAE − teacher) | -0.6 | -1.8 | -0.5 | -0.2 | -0.2 | +0.3 | +0.1 |
| JumpReLU teacher () | 75.8 | 81.3 | 86.3 | 88.3 | 89.7 | 91.2 | 93.4 |
| RSAE (JumpReLU-init) | 79.0 | 82.4 | 86.3 | 88.6 | 90.0 | 91.4 | 93.6 |
| (RSAE − teacher) | +3.2 | +1.0 | 0.0 | +0.3 | +0.3 | +0.1 | +0.2 |
| TopK teacher () | 79.2 | 82.7 | 85.8 | 88.1 | 89.7 | 91.3 | 93.5 |
| RSAE (TopK-init) | 79.6 | 82.6 | 85.8 | 88.0 | 89.9 | 91.4 | 93.5 |
| (RSAE − teacher) | +0.4 | -0.1 | 0.0 | -0.2 | +0.2 | +0.1 | 0.0 |
| ReLU teacher (, the pair of Table 4) | 76.0 | 81.2 | 84.9 | 87.3 | 89.2 | 91.1 | 93.7 |
| RSAE (ReLU-init) | 77.9 | 80.5 | 85.2 | 88.3 | 90.1 | 91.2 | 93.7 |
| (RSAE − teacher) | +1.9 | -0.6 | +0.2 | +1.0 | +0.9 | +0.1 | +0.1 |
| residual stream (no SAE), top- dimensions | 64.9 | 73.1 | 81.8 | 84.9 | 87.3 | 90.0 | 93.2 |
Mean test accuracy (%) over the 8 SAEBench datasets; the last row is the SAEBench baseline that probes the top- dimensions of the residual stream itself. Across the four pairs and the six values of , the RSAE is within 2 points of its teacher in 23 of 24 cells, and the one larger difference is in its favour (+3.2 at for the JumpReLU teacher). There is no systematic direction: the RSAE is slightly below the ReLU teacher for (−0.2 to −1.8), above the JumpReLU teacher for , indistinguishable from the TopK teacher, and level with or slightly above the ReLU teacher. For scale: probes on one or two latents are noisy (on single datasets the RSAE–teacher difference reaches 10–20 points at and stays below about 4 points for ; re-running the probe on the same teacher moves its mean accuracy by up to 0.4 points), and every SAE, teacher or RSAE, is 11–15 points above the top-1 residual-stream dimension.
We want to be precise about what this shows and what it does not. We have no theoretical guarantee on the interpretability of the learned codes, and we do not claim one. Sparse-probing accuracy is also not a direct measurement of how interpretable a dictionary is: it measures whether a small number of latents is linearly predictive of a handful of labelled concepts. We report it as the standard protocol, to show that the RSAE produces sparse features that score as well as its teacher's under that protocol. The claim of the paper is the one that our theory can support and that the experiments verify: better fidelity at matched sparsity.
Completeness of Figure 2. Figure 2 is not a selection of favourable configurations. It contains every SAEBench teacher that exists at on Pythia-160m: all six sparsity levels for JumpReLU and TopK and all five for ReLU, 17 teachers in total, including the ones where the RSAE gains nothing (the JumpReLU curve sits at 0–1% for , and one CE point is slightly negative and is shown as such). Table 5 gives the same comparison for all 51 (sparsity, width, family) settings, again with no omission. Since submission we went further: with better fine-tuned RSAEs we now have results for all 54 SAEBench teachers on Pythia-160m and all 54 on Gemma-2-2B (3 widths 6 sparsity levels 3 families). The RSAE lowers MSE for all 36 ReLU teachers, for 34 of 36 TopK teachers and for 29 of 36 JumpReLU teachers. The first is what Theorem 1 leads one to expect, and for the gated teachers the reduction grows with the number of active latents, which is the regime Theorem 2 identifies.
Response to Reviewer 9Gh3
Significance of the contribution. We respectfully disagree with the criterion implied here, namely that an SAE contribution is only worth publishing if it wins by a large margin on every existing benchmark at once. We see three problems with it.
(i) The existing metrics are partial, and none of them is "the" target. No current benchmark measures interpretability directly. Sparse probing (above) tests linear decodability of a few labelled concepts; automated interpretability scores depend on the explainer model and the prompt; absorption, SCR/TPP and RAVEL each probe one specific failure mode on specific datasets. SAEBench itself reports that gains on proxy metrics do not reliably translate into better practical performance, and that architectures rank differently on different metrics. A method can be "better" on one and "worse" on another without either verdict being wrong. Requiring a uniform win on all of them selects for benchmark coverage, not for understanding. By contrast, the quantities we target, reconstruction error and cross-entropy degradation at matched , are well-defined, architecture-independent, and are the ones for which we can prove statements (Theorems 1 and 2). We consider it a strength, not a weakness, that the paper claims exactly what it can prove and verify.
(ii) Which metric matters depends on the downstream use, and for a growing class of uses fidelity is the binding constraint. SAEs are no longer used only to browse features of a chat model. They are now applied to scientific foundation models in chemistry, materials science, biology and medicine, physics and astrophysics [1-3]. In natural-language settings one can often tolerate a small loss of fidelity: after splicing in the reconstruction or ablating a latent, the model still produces fluent text, and the experiment remains interpretable. In chemistry and materials models this is not the case. The model's inputs and outputs are constrained strings (e.g. polymer SMILES, pSMILES); a reconstruction error in the hidden state that would be harmless for text can turn the output into a string that is chemically invalid, i.e. not a molecule at all. An intervention study (ablating or steering a sparse feature) on top of such a reconstruction is then physically meaningless, whatever the probing accuracy of the dictionary. For these applications the first requirement on an SAE is that it reproduces the model's computation at the sparsity one can afford to inspect, and an RSAE, which lowers reconstruction error and CE at the same , is better suited than the SAE it starts from.
[1] Cohen, Jaron, Alexander G. Hasson, and Sara Tanovic. "Unveiling latent knowledge in chemistry language models through sparse autoencoders." arXiv preprint arXiv:2512.08077 (2025).
[2] Varadi, Kristof, Mark Marosi, and Peter Antal. "Circuits, features, and heuristics in molecular transformers." arXiv preprint arXiv:2512.09757 (2025).
[3] Simon, Elana, and James Zou. "InterPLM: discovering interpretable features in protein language models via sparse autoencoders." Nature methods 22.10 (2025): 2107-2117.
(iii) What we propose is an upgrade of SAEs people already have, and we say plainly what it does and does not buy. The RSAE starts from a released ReLU, JumpReLU or TopK SAE, adds 20 scalars, and is trained for minutes on one consumer GPU. For the most widely released family (ReLU) the gain over the released SAE is not marginal: reconstruction error drops by 27–98% and CE by 34–97% at matched on the completed Pythia-160m grid (25–87% and 34–93% in Table 5 of the submission), by 14–69% and 43–75% on Gemma-2-2B (all three widths and all sparsity levels, completed since submission; see the general response), and both improve at every one of the 12 layers of GPT-2 small. For JumpReLU and TopK teachers the gains are small, and the paper says so ((C2), lines 399–416). The new controls requested by the other reviewers (see the general response) also show that an equally fine-tuned JumpReLU reaches the same fidelity or slightly better, so we do not claim that the RSAE beats the best gated SAE; we claim that one smooth, 20-parameter, gate-free activation reaches (or, from a ReLU teacher, comes within about 10% of) that level from any of the three starting families, with a proof that it can never be worse than ReLU. We think a method with a proof of non-inferiority, a clearly identified regime where it helps a lot, and an honest statement of where it only matches existing designs is a useful contribution to a field that the reviewer describes as crowded with variants of unclear value.
We believe that several of the concerns raised stem from material that is already in the submission but was overlooked, and from a narrower view of the purpose of SAEs than the one the paper takes. In light of the clarifications above and the additional results, we would appreciate it if the reviewer would re-evaluate the submission and reconsider the score.
Response to Reviewer 9Gh3 (part 1)
We thank the reviewer for the time devoted to our submission and for the candid feedback. We address each of the reviewer's concerns below.
Clarity of Section 3. We respectfully disagree that Section 3 leaves its construction unexplained. Its layout is deliberate, and the items the reviewer found missing are in the submitted text.
We nevertheless want the section to be easy to read for someone who goes straight to it, so we will adopt the following clarification:
Correction of references. We thank the reviewer for the careful reading of the related work. Laptev et al. (2025) analyse how features of SAEs trained on activations evolve across layers, using a data-free (decoder cosine-similarity) matching between layers. Our sentence describes this work as fitting dictionaries directly from model weights; we will reword it to describe an analysis and steering method built on standard SAEs. The sentence is part of a short list of work orthogonal to ours (Section 2, lines 108–109), and no claim or result of the paper depends on it. We are also re-checking the other citations in Section 2 against the cited papers and will correct any description we find imprecise.