- Rating
- 2
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper studies malicious fine-tuning in a partially protected open-weight setting. It combines a null-space-based score, a cubic layer, an inverse adapter, and a protected custom backward gate. Most weights remain trainable, while the gate and output head are protected by a root of trust. Experiments show resistance to covered harmful fine-tuning, useful benign behavior, and clear limits under incomplete coverage and representation drift. The evidence supports selective gradient gating, but leaves the added value of the cubic construction and parts of the adaptive-attack argument unresolved.
Critical strengths
-
The paper studies a useful threat model and states its trust assumptions. Controls with the same forward computation support the benefit of blocking covered harmful gradients. The benign-task results also show that the gate can allow useful continued training.
-
The evaluation reports important limits, including poor detection of unseen harmful samples in some domains, threshold drift, and loss of separation after large updates. It also tests two surrogate families with eight seeds each and clearly marks severe generation collapse.
Critical weaknesses
-
The evidence does not isolate the benefit of the cubic-inverse construction. Where the exact inverse exists, the composite forward map is the identity; the practical blocking comes from the custom backward rule in Eq. 8. This does not invalidate the implemented gate, but the current baselines do not show what the cubic construction adds over a direct norm-ratio gate or another simple gate with the same protected interface, reference data, and budget.
-
The adaptive evaluation does not fully support the claims that attack routes are closed or that surrogate distillation is the sole remaining route. Appendix A specifies a hidden-state-to-logit interface, and the attacker controls training. The tested surrogates replace the protected head and fail during generation. This does not rule out using an approximate gradient to update the open weights while retaining the real protected head for inference. I do not claim this attack succeeds, but a focused test or a narrower security claim is needed.
Questions for authors
-
Under the actual protected interface and low-budget threat model, can an attacker use forward queries or a surrogate only to estimate gradients, while keeping the real protected head for inference? Please specify the relevant input restrictions, precision, and query budget, and evaluate this route or narrow the closure claims.
-
How does the method compare with a direct norm-ratio backward gate without the cubic-inverse pair, and with a suitable simple gate, using the same trust boundary, reference data, and budget? Which benefit specifically requires the cubic construction?
-
Please explain the BeaverTails-H differences between Tables 1 and 7 at the stated deployed operating point, report repeats or uncertainty for the main SFT results, and correct the claim that Table 6 never shows an ASR increase.
2 comments
Q3. Numerical and descriptive errors
We thank the reviewer for carefully identifying these inconsistencies. The BeaverTails-H entries in the 100%-coverage column of Table 7 were swapped due to a copying error during table preparation. The correct values should match Table 1: 0.07 for Llama-8B and 0.05 for Qwen-14B. We will correct Table 7 accordingly. This copying error does not affect the other experimental results or their conclusions.
We also agree that the claim in the main text and the Table 6 caption that the gate never increases ASR is inaccurate. The numerical data in Table 6 themselves do not require correction. We will remove overly strong statements such as “never increases ASR” and “reduces ASR in every cell,” and instead report the results as observed: some low-ASR cells exhibit small increases of 0.01–0.02.
In addition, we will report repeated runs of the main SFT experiments with multiple random seeds, including standard deviations.
| model | dataset | None Pre/Post | Booster Pre/Post | RepNoise Pre/Post | Antibody Pre/Post | Null Space Cubic Pre/Post |
|---|---|---|---|---|---|---|
| Llama | JBB | 0.02±0.000 / 0.90±0.005 | 0.02±0.000 / 0.97±0.008 | 0.02±0.000 / 0.37±0.033 | 0.02±0.000 / 0.01±0.005 | 0.02±0.000 / 0.03±0.005 |
| Llama | HB | 0.13±0.000 / 0.75±0.000 | 0.13±0.000 / 0.49±0.000 | 0.13±0.000 / — | 0.13±0.000 / 0.10±0.000 | 0.13±0.000 / 0.13±0.000 |
| Llama | BT | 0.07±0.005 / 0.24±0.000 | 0.07±0.004 / 0.71±0.005 | 0.07±0.002 / 0.09±0.000 | 0.07±0.002 / 0.10±0.000 | 0.07±0.002 / 0.07±0.000 |
| Qwen | JBB | 0.02±0.000 / 0.57±0.000 | 0.02±0.000 / 0.00±0.000 | 0.02±0.000 / 0.75±0.005 | 0.02±0.000 / 0.05±0.000 | 0.02±0.000 / 0.02±0.000 |
| Qwen | HB | 0.14±0.005 / 0.44±0.005 | 0.14±0.005 / 0.16±0.000 | 0.14±0.005 / 0.85±0.005 | 0.14±0.009 / 0.28±0.005 | 0.14±0.005 / 0.14±0.009 |
| Qwen | BT | 0.05±0.000 / 0.07±0.000 | 0.05±0.000 / 0.53±0.009 | 0.05±0.000 / 0.38±0.012 | 0.05±0.002 / 0.21±0.009 | 0.05±0.002 / 0.05±0.000 |
W1 + Q2. What benefit does the Cubic Layer provide?
In the threshold-based implementation used in this paper, a transparent forward pass combined with a direct norm-ratio backward gate is functionally equivalent to the Cubic Layer + Inverse Adapter. The main features of the cubic construction are its analytic derivation of the backward-blocking condition and its invertible decomposition of the forward computation.
From the backward perspective, compared with training a conventional linear classifier, our null-space construction provides a training-free way to derive the gradient-blocking condition. The fixed matrix W can be obtained through a one-time SVD of hidden states from harmful reference samples, without introducing a classification loss or training an additional harmfulness classifier. The Cubic Layer further provides an analytic connection between this null-space geometry and gradient blocking: for C(h)=(hWh⊤)h, the terms in its input Jacobian contain hW or hWh⊤, so the input gradient vanishes when hW=0. Thus, the null-space construction provides the training-free representation geometry, while the Cubic Layer provides the analytic derivation of the gradient-blocking condition.
From the forward perspective, compared with a transparent forward pass, the Cubic Layer + Inverse Adapter decomposes the safeguard into two functionally decoupled steps with clearly distinct roles: a safety-related representation transformation and restoration of the base model’s forward behavior. These steps can be analyzed and modified separately. Although we currently use this decomposition only to preserve the original forward behavior, it offers potential for future extensions. For example, adjusting the compensation strength of the Inverse Adapter could allow the cubic transformation to exert a controlled influence on hidden representations during the forward pass.
Importantly, both the Cubic Layer + Inverse Adapter and a transparent forward pass with a direct norm-ratio backward gate are concrete implementations of the abstract USG mechanism. Other structures capable of selective gradient blocking could also instantiate USG, although their decision rules and construction procedures would differ from our null-space construction. We will revise the paper to clarify the three levels of contribution: the abstract USG mechanism, the analytic cubic instantiation, and the PPOW protected-release structure. We will also avoid implying that the cubic computation itself is indispensable to the results reported in the current experiments.
W2 + Q1. Can an attacker fit a surrogate to approximate the gradient mapping and thereby obtain approximate ungated gradients?
We analyze this route in terms of the information available through the gated-backward interface.
The gradient returned by the protected component can be written as
G(h)=g(h)F(h),
where F(h) is the ungated backward mapping and g(h) is the gate induced by the cubic construction. The gate uses
r(h)=∥h∥∥hW∥,
blocking the gradient when r(h)≤τ and passing it when r(h)>τ. Thus,
\qquad G(h)= \begin{cases} F(h), & r(h)>\tau,\\ 0, & r(h)\leq\tau. \end{cases}$$ The attacker’s goal is to recover the gradient mapping without the gate, namely $F(h)$. For this analysis, we assume that the attacker does not know the protected architecture inside the TEE and therefore has no prior knowledge of $F$. We further consider the conservative case of an unlimited query budget, arbitrary inputs, and access to the complete gated-backward function $G$. For a blocked input, the response $G(h)=0$ provides no direct observation of the local value of $F(h)$ through that backward response. Under a backward-only interface, zero responses alone do not uniquely identify $F$ over the blocked region. An attacker attempting to reconstruct $F$ would therefore need to extrapolate from observations in the passing region $r(h)>\tau$. The feasibility of such extrapolation depends on the functional complexity of $F$ and on how well the available observations cover the relevant harmful input directions. In the extreme case where the behavior of $F$ in the blocked region is entirely unrelated to its behavior in the passing region, observations from the latter provide no basis for predicting the former. In practical deployments, using a more complex protected mapping—for example, stacked MLPs or a Transformer layer rather than a linear head—can increase the cost of surrogate fitting and gradient reconstruction. This is consistent with our distillation-hardening argument: increasing the complexity of the protected computation is expected to raise the attacker’s reconstruction cost, although we do not treat complexity alone as a proof that reconstruction is impossible.