- Rating
- 2
- Confidence
- 3
- Informativeness
- 1
- Presentation
- 2
- Soundness
- 1
Read review
Paper summary
The authors propose CARS as an inference time defense for MLLMs They first prepare two sets of text instructions containing benign and harmful requests These instructions are fed into the target MLLM The hidden state of the last input token is extracted at each layer The difference between the mean hidden states of the two groups gives the safety steering direction for each layer A logistic regression classifier is also trained on text hidden states to distinguish benign from harmful content Classification F1 scores across different intervention layer ranges and generation positions are compared to select the target layers and a fixed detection position During inference the model first applies steering at the base strength to the target layers At the preset generation position the classifier reads the hidden state of the current token across the target layers and averages the predictions to obtain a harmfulness probability If the input is classified as benign further steering stops and the generated content is retained If it is classified as harmful the steering strength increases and the output before the detection position is discarded Generation then continues
Critical strengths
-
The method considers when detection and intervention become effective during generation It starts with steering at a lower strength performs classification at a later position and increases steering when harm is detected This design follows the observed failure pattern
-
The experiments cover three models and four safety benchmarks including both text and image attacks
-
CARS operates during inference without updating the MLLM parameters It uses direction vectors and logistic regression which gives it some deployment value
Critical weaknesses
-
The method modifies hidden states using text safety directions and then uses a text safety classifier to assess their classifiability and guide further intervention These operations rely on signals from a single modality and do not establish cross modal alignment or convergence of the representation distributions The improvement may also come from safety related content in the generated prefix that the classifier can recognize more easily The reasons for using only text safety directions and a text classifier are also not fully supported The poor results from image and text mean difference directions only show that this construction has limited effect in the tested setting They do not rule out the value of image and text data The paper lacks a comparison of detectors trained on text data image and text data and a mixture of both under the same conditions It also does not sufficiently test directions extracted from image and text representations at later generation positions
-
The detection position is a fixed hyperparameter chosen separately for each model. The 10th or 15th generation position is selected using offline classification performance. The method does not check whether the current input has reached convergence during inference. Figure 4 shows that changes in early detection positions can affect defense performance. It remains unclear whether these fixed positions transfer across tasks, response lengths, and input distributions
-
The ablations on intervention layers and steering strength are limited. All three models use 12 intervention layers with a base strength of 3.5 and an enhanced strength of 6.0. Although the paper compares two strength settings, it does not test a wider range or fully report the effects of changing the starting layer and the number of intervention layers. The experiments therefore do not establish how stable these settings are or how much they depend on calibration for each model
-
CARS makes a decision only once at a fixed position. It does not check again after a benign decision or verify the later output after a harmful decision. Its ability to recover from an initial missed detection has therefore not been evaluated separately. Later detection also means longer intervention and a longer discarded prefix. These factors may explain part of the defense improvement, so the gain cannot be attributed entirely to convergence
Questions for authors
-
Can the authors provide evidence of convergence that is independent of the text safety directions and text safety classifier? How can they rule out shifts relative to the classification boundary and the effect of safety-related words in the generated prefix as explanations for the F1 improvement?
-
Under the same sample budget, how do detectors trained on text data, image and text data, and a mixture of both compare? Do safety directions extracted from image and text representations at later generation positions remain ineffective?
-
What calibration data and selection rule are used to choose the 10th or 15th position? How are these data separated from the final test set? Do the selected positions transfer across datasets?
-
How much does performance change when the starting layer or the number of intervention layers changes? Can the authors provide a joint sensitivity analysis of the intervention layer range and detection position?
-
Can the authors test a wider range of steering strengths and intervention layer settings?
-
How does CARS check whether later content is harmful after the first detection? What are the benefits and costs of repeated detection or a final output check?
10 comments
Response to Reviewer CifH (2/4)
We next examine whether modality and harmfulness remain separable during unsteered generation. At every token index , we train a new classifier using only representations from that same index. The modality classifier separates from , while the semantic classifier separates from . We use 15% of the data for training and 85% for testing, so this comparison measures modality versus semantic separability at the same token position rather than transferring a fixed classifier across positions. Higher accuracy indicates that the corresponding distinction remains more separable. The complete curves are provided in Appendix A.2, Fig. 7, and selected token positions are reported below.
| Model | Category (accuracy, %) | Token 0 | Token 1 | Token 2 | Token 3 | Token 4 | Token 5 | Token 10 | Token 15 |
|---|---|---|---|---|---|---|---|---|---|
| LLaVA-1.5-7B | Modality classifier | 100.0 | 95.4 | 95.4 | 98.4 | 99.2 | 98.4 | 93.3 | 81.4 |
| LLaVA-1.5-7B | Semantic classifier | 98.1 | 95.4 | 91.7 | 94.4 | 82.3 | 87.1 | 92.2 | 88.4 |
| Qwen2.5-VL | Modality classifier | 100.0 | 96.0 | 95.7 | 96.8 | 97.0 | 97.9 | 93.0 | 76.1 |
| Qwen2.5-VL | Semantic classifier | 98.7 | 95.7 | 97.9 | 93.8 | 94.4 | 89.3 | 90.6 | 89.7 |
The modality classifier is nearly perfect at the input position, even though the examples within a modality contain different semantic labels. Its accuracy then falls from 100.0% to 81.4% for LLaVA-1.5-7B and to 76.1% for Qwen2.5-VL by token 15. Meanwhile, semantic classification remains comparatively stable and becomes more competitive than modality classification at later positions. Together with the cosine results above, this shows that the representation becomes progressively less modality-specific without any base steering. It also complements the original experiment in Section 4.1 and Fig. 2, where a classifier trained only on the final input prompt representations yields increasing F1 on visual representations over generation steps. Therefore, convergence is a naturally occurring phenomenon rather than an effect induced by steering.
Response to Reviewer CifH (3/4)
Q2: Under the same sample budget, how do text-only, image--text, and mixed-modality safety signals compare?
Thanks for your constructive feedback. The purpose of the base text-representation steering in CARS is not to claim that visual representations are intrinsically unsafe. Rather, the backbone LLM exhibits much stronger safety alignment in the text representation space. Without early guidance, a harmful visual input can undergo a modality shift while already generating harmful tokens, and later convergence cannot undo those tokens. The original Fig. 2 shows that applying text-derived steering makes the image-representation F1 curve rise faster and reach a higher level. We further test this explanation using the matched datasets under the same sample budget.
For each output position, we train a new semantic classifier and compare three conditions: no steering, text-representation steering, and image-representation steering. All steering directions are extracted from the matched datasets above, using and . The complete curves are provided in Appendix A.3, Fig. 8, and selected positions are reported below.
| Model | Category (accuracy, %) | Token 0 | Token 1 | Token 2 | Token 3 | Token 4 | Token 5 | Token 10 | Token 15 |
|---|---|---|---|---|---|---|---|---|---|
| LLaVA-1.5-7B | Baseline semantic classifier | 98.1 | 95.4 | 91.7 | 94.4 | 82.3 | 87.1 | 92.2 | 88.4 |
| LLaVA-1.5-7B | Text-rep steering classifier | 98.7 | 97.9 | 97.3 | 98.1 | 85.8 | 90.6 | 91.7 | 91.1 |
| LLaVA-1.5-7B | Image-rep steering classifier | 99.5 | 97.6 | 96.5 | 95.7 | 83.9 | 89.5 | 90.1 | 87.1 |
| Qwen2.5-VL | Baseline semantic classifier | 98.7 | 95.7 | 97.9 | 93.8 | 94.4 | 89.3 | 90.6 | 89.7 |
| Qwen2.5-VL | Text-rep steering classifier | 98.9 | 96.0 | 98.1 | 94.1 | 94.6 | 89.5 | 91.4 | 92.0 |
| Qwen2.5-VL | Image-rep steering classifier | 98.1 | 96.2 | 98.9 | 96.0 | 92.5 | 88.2 | 85.8 | 84.1 |
Text-representation steering improves the average accuracy over positions 0--15 from 90.2% to 92.8% for LLaVA-1.5-7B and from 92.0% to 93.4% for Qwen2.5-VL. Image-representation steering is not consistently better than the baseline and declines at later positions, especially for Qwen2.5-VL. These results explain why CARS uses a lightweight text-derived intervention before the detection point: it makes the visual representation more compatible with the text-trained safety direction early enough to prevent harmful-prefix generation, while still allowing the model to transition to refusal after harmfulness is detected.
We also run the complete image-representation CARS variant under the same setting. Following the CARS$_{\mathrm{avg}}$ ablation, both variants average predictions over output tokens 5--15, so the only changed component is the source modality of the safety direction. The results are reported in Appendix A.3, Table 6.
| Model | Method | FigStep | MM-SafetyBench |
|---|---|---|---|
| LLaVA-1.5 | Baseline | 43.0% | 64.8% |
| LLaVA-1.5 | CARS$_{\mathrm{avg}}$ (Text Rep.) | 92.7% | 95.8% |
| LLaVA-1.5 | CARS$_{\mathrm{avg}}$ (Image Rep.) | 64.0% | 76.7% |
| Qwen2.5-VL | Baseline | 67.8% | 76.3% |
| Qwen2.5-VL | CARS$_{\mathrm{avg}}$ (Text Rep.) | 97.0% | 93.9% |
| Qwen2.5-VL | CARS$_{\mathrm{avg}}$ (Image Rep.) | 65.0% | 87.6% |
Image-representation CARS improves over the undefended baseline in three of the four model--benchmark settings, but is 2.8 points below the Qwen2.5-VL baseline on FigStep. It is consistently weaker than text-representation CARS: the gaps are 28.7 and 32.0 points on FigStep for LLaVA-1.5 and Qwen2.5-VL, respectively, and 19.1 and 6.3 points on MM-SafetyBench. This supports our interpretation that the backbone's text safety alignment is the useful transferable signal, while an image-derived direction may retain modality-related features and therefore encode semantic refusal behavior less reliably.
Response to Reviewer CifH (1/4)
Q1: Can convergence be demonstrated independently of the text safety direction and classifier boundary?
Thanks for your constructive feedback. To address this concern, we construct matched datasets following the basic idea of FigStep. We sample 128 harmful prompts from AdvBench and 128 benign prompts from Alpaca. For every prompt, we create one text-only example and one (image,text) pair by rendering the prompt itself as text in an image. All image examples use the same instruction. This produces four datasets, Dvlharm, Dtextharm, Dvlbenign, and Dtextbenign, so comparisons across modalities keep the semantics fixed, while comparisons across harmfulness labels keep the modality fixed. Examples are provided in Appendix A.1, Fig. 5.
We first analyze representational proximity without applying any steering on two architecturally different models, LLaVA-1.5-7B and Qwen2.5-VL. At each output position, we compute cosine similarity across layers for four dataset pairings. The two within-modality pairings compare harmful and benign semantics, whereas the two cross-modal pairings compare matched semantics. Higher cosine similarity indicates greater representational proximity. The complete curves are provided in Appendix A.2, Fig. 6, and selected token positions are reported below.
At token 0, the highest cosine similarities occur between harmful and benign inputs within the vision modality, even though their semantics differ: 0.943 for LLaVA-1.5-7B and 0.992 for Qwen2.5-VL. In contrast, matched semantics across modalities have substantially lower similarity. During generation, the similarities between harmful and benign visual inputs decrease to 0.209 and 0.212 by token 15, whereas cross-modal pairs with matched semantics vary comparatively less after the initial generation steps. This shows that representations become progressively less modality-specific as generation proceeds, while semantic information is comparatively preserved.