- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes SWAG (Steering With Augmented Guidelines), a framework for constructing activation-steering directions without behavior-labeled examples. By combining contrasting natural-language guidelines with task-independent random-token or natural-text substrates, SWAG extracts behavioral representations that can be integrated into existing steering methods, including Linear Steering, K-Steering, and Angular Steering. Experiments across multiple models and behavioral tasks demonstrate competitive performance against supervised steering approaches in several settings.
Critical strengths
- The paper addresses a meaningful problem by reducing the dependence of activation steering on behavior-specific labeled datasets. Constructing steering directions from natural-language guidelines and task-independent substrates offers a simple and potentially useful alternative.
- The evaluation covers multiple model families, behavioral tasks, and steering methods, demonstrating the applicability of SWAG across different intervention techniques. The substrate-related ablations also provide useful empirical evidence supporting the proposed approach.
Critical weaknesses
- Figure 12 shows that certain steering configurations degrade MMLU accuracy, while Table 1 demonstrates SWAG's ability to substantially weaken safety refusal. However, the method does not explicitly constrain degradation in non-target capabilities or systematically characterize the trade-offs between behavioral control, factual correctness, and safety preservation. Although refusal suppression is intentional in the safety experiment, unintended side effects when steering other behaviors remain insufficiently investigated.
- Several experiments select optimal steering parameters directly on the evaluation data. Figures 2 and 3 choose favorable steering multipliers on the same test questions, while Table 1 reports peak attack success rates over 36 rotation angles, separately optimized for each evaluator. Although these procedures are disclosed, they may lead to optimistic performance estimates and leave generalization under independently calibrated settings unclear.
Questions for authors
- Although SWAG constructs steering directions without behavior-labeled examples, some experiments rely on supervised development data or previously selected intervention layers, while steering strengths require task-specific evaluation feedback. How would SWAG determine these critical parameters in a fully label-free deployment setting, and could it maintain comparable performance without such supervision?
- Why are direct prompting results not reported for the open-ended generation experiments in Figure 2, despite the prompting protocol being described in Appendix C.3 and prompting baselines being included in other settings? Without this comparison, it is difficult to assess whether SWAG offers meaningful advantages over directly applying the same guidelines in open-ended behavioral control.
8 comments
Results for sycophancy and myopic-reward (MCQ P-Match metric) when conducting angular steering for refusal on Qwen-2.5-7B-Instruct are below. In the experiments below, guidelines (single-pair) and heldout sets are the same from the submission. The results show that sycophancy score changes by less than 1%, and myopic-reward changes by 5.4% (note that mean separation on OLMO myopic-reward MCQ was 0.10 when steering with random directions, and that there is potential cross-behavior interaction between refusal and myopic-reward). The results below are taken with the full-data layer/position and angle that we report later in our response (layer 19 post-attention residual at the angle of 170 degrees), and are across seeds 0, 1, 2.
Change in MCQ performance of angular-steered Qwen model.
| Behavior | Holdout questions | Unsteered p-match (%) | SWAG p-match (%) | Change (pp) |
|---|---|---|---|---|
| Myopic reward | 100 | 73.63 | 79.04 ± 0.35 | +5.40 |
| Sycophancy | 100 | 89.54 | 88.63 ± 0.08 | −0.91 |
2. Steering hyperparameters are selected on evaluation data
We acknowledge the reviewer's concern that multipliers in the Section 4.1 experiments and angles in the Section 4.4 experiments are selected on the reported questions. Below, we present new results where we select multipliers and angles based on a separate development-set prior to final testing on the heldout set, and demonstrate that the multiplier- or angle-selection process generally does not have a large impact on final scores. We will include these results in the revised version of the paper. While these results do not change the conclusions of our paper, we appreciate the reviewer's feedback, which helps make our experimental results more robust.
We note that this protocol is applied identically to every steered method, including the supervised baselines, and the multiplier selection has very few degrees of freedom: one of two magnitudes {0.5, 1.0} per side (Lines 262-263). AxBench, our largest evaluation, already uses disjoint development and held-out data (Lines 406-407), and our Sentiment Continuation on Yelp experiments report all multipliers in the range [-2, 2] (Figure 4).
The guidelines, development, and heldout sets used in our experiments below are the same from the submission (Lines 267-269). Results of SWAG on OLMO with the development-set multiplier-selection method, along with the corresponding CAA results, are presented below. Eight of twelve settings see a test-score change of less than 1% relative to the original results in our paper. SWAG retains its advantage over supervised CAA in the open-ended settings. Note that in our new experiments, open-ended outputs are newly generated and judged.
OLMO MCQ
| │ | Sycophancy | │ | Myopic-Reward | |||||
|---|---|---|---|---|---|---|---|---|
| Method | │ | New Results | Paper Results | Change | │ | New Results | Paper Results | Change |
| Supervised CAA | │ | 0.2855 ± 0.0000 | 0.2845 ± 0.0000 | +0.35% | │ | 0.6945 ± 0.0000 | 0.6939 ± 0.0000 | +0.09% |
| SWAG-Rand linear | │ | 0.1074 ± 0.0032 | 0.1068 ± 0.0046 | +0.56% | │ | 0.3846 ± 0.0157 | 0.3849 ± 0.0139 | −0.08% |
| SWAG-Text linear | │ | 0.0504 ± 0.0052 | 0.0498 ± 0.0048 | +1.20% | │ | 0.3885 ± 0.0329 | 0.3900 ± 0.0330 | −0.38% |
OLMO OE
| │ | Sycophancy | │ | Myopic-Reward | |||||
|---|---|---|---|---|---|---|---|---|
| Method | │ | New Results | Paper Results | Change | │ | New Results | Paper Results | Change |
| Supervised CAA | │ | 80.68 ± 0.00 | 84.55 ± 1.25 | −4.58% | │ | 148.63 ± 0.00 | 140.17 ± 3.02 | +6.04% |
| SWAG-Rand linear | │ | 118.82 ± 1.90 | 123.47 ± 3.51 | −3.77% | │ | 183.97 ± 0.86 | 183.50 ± 1.47 | +0.26% |
| SWAG-Text linear | │ | 125.54 ± 0.45 | 125.64 ± 1.60 | −0.08% | │ | 181.34 ± 1.05 | 181.38 ± 2.13 | −0.02% |
We additionally ran experiments for angular steering on AdvBench test splits, where we select the best angle on a development set. In the following experiments, the layer and position is fixed based on the process documented in the original Angular Steering paper (Section A.1.1, Lines 753-757), using the training split (either the SWAG guidelines + substrates, or the paper's training set). Once the layer and position are selected, angle is fixed based on the highest HarmBench score on a development set (mean of three seeds: 0, 1, 2). Note that in these experiments, we sample a new development set from the same training set in our submission, but the guidelines (single-pair) and heldout set are retained from the submission. In our original submission, 104 test prompts from AdvBench selected with random_state 42 (Section F.1 Line 1630) were kept as the heldout test set. The other 416 AdvBench prompts, and the full 512 Alpaca (harmless) prompts formed the training set for the supervised method. In the following experiments, we use the same 512 Alpaca prompts and 312 of the 416 original AdvBench training prompts, sampled with random state 11092026, as the new training set for full-data supervised. The 104 excluded AdvBench prompts from the original training set are used as the new development set, and we use the same 104 test prompts from the submission. As we can see below, SWAG is competitive with the full-data supervised method on all settings, and is particularly strong when the layer/position are fixed based on the full-data training split and the angle is then fixed by SWAG on the development set.
Refusal on Qwen (Angular Steering)
| Mode | Steering layer/position | Fixed angle | HarmBench ASR (%) | LlamaGuard 3 ASR (%) | Substring non-refusal (%) |
|---|---|---|---|---|---|
| Unsteered | — | — | 1.92 | 0.00 | 2.88 |
| Full-data supervised | 19/mid | 170° | 88.46 | 96.15 | 100.00 |
| SWAG: own layer / own angle | 21/mid | 210° | 80.13 ± 2.00 | 86.22 ± 2.00 | 99.04 ± 0.00 |
| SWAG: full-data layer / own angle | 19/mid | 220° | 90.71 ± 2.42 | 93.91 ± 1.11 | 100.00 ± 0.00 |
| SWAG: full-data layer / full-data angle | 19/mid | 170° | 89.10 ± 0.56 | 92.31 ± 1.67 | 100.00 ± 0.00 |
| SWAG: own layer / full-data angle | 21/mid | 170° | 77.24 ± 1.47 | 82.37 ± 1.11 | 97.12 ± 0.00 |
Table: When selecting angles using a development set, SWAG is competitive with the full-data supervised method. "Own layer" refers to the setting where SWAG selects the best layer and position with its own guideline + substrate activations, using the process from the Angular Steering paper. "Own angle" refers to SWAG choosing the best angle by HarmBench performance on the development set, after the layer and position are set. "Full-data layer" refers to the setting where SWAG uses the same fixed layer and position as the supervised method. "Full-data angle" refers to SWAG using the angle selected by the supervised method.
We thank the reviewer for the thorough assessment and useful insights regarding our paper.
To begin, we appreciate the reviewer's commendation of SWAG for addressing the meaningful problem of needing to create behavior-specific labeled datasets for activation steering. This, in addition to our evaluation across model families, behavioral tasks, and steering methods, were each stated as strengths across each of our reviewers. We also thank the reviewer for acknowledging the substrate-related ablations as further empirical evidence that justifies the SWAG method. The remainder of our response includes justifications for several of the reviewer's concerns and results from additional experiments which further strengthen our central claims.
1. Side effects on non-target capabilities are insufficiently investigated
We appreciate the reviewer's concern regarding non-target capabilities and would like to offer some clarifications and a new experimental result to address this concern.
While SWAG does not explicitly constrain degradation in non-target capabilities, this is true of the baseline methods (CAA, K-Steer, Angular Steering) we build on as well, since SWAG changes how the direction is extracted, not how it is applied.
Related papers have used certain metrics to track of the impact of steering on non-target capabilities (e.g. CAA and K-Steer papers evaluate MMLU results and Angular Steering paper evaluates model perplexity as well as results on the TinyBenchmark general capabilities suite). Similarly, we ran MMLU evaluation for SWAG, which is shown in Figure 12.
More importantly, the final scores for most of our paper's experiments account for side effects by considering qualities of a response beyond exhibiting the target behavior, such as instruction-following and coherence.
First, most open-ended steering scores in our paper penalize degraded outputs. Section 4.1 "Open-Ended" (Lines 258-261) and Section 4.2 "Overall Steering Scores" (Lines 365-367) take the harmonic mean of behavioral change and coverage (the fraction of usable, scorable generations). The AxBench score (Lines 404-406) is a harmonic mean of three metrics: concept-inclusion, instruction-following, and fluency. A direction that achieves the target behavior without following the instruction or producing a coherent output thus scores poorly.
Second, Figure 12 measures general capabilities directly. At multipliers of ±1, SWAG preserves OLMo MMLU accuracy better than its supervised counterpart in twelve of sixteen comparisons. For example, under linear sycophancy suppression at α = −1, supervised steering drops MMLU from 61.3% to roughly 40%, while both SWAG variants stay near 56–58%.
Third, as stated by the reviewer, the refusal experiment (Table 1) is an intentional demonstration that SWAG can reproduce an established capability, as discussed in our ethics statement. In this case, it is not a side effect of steering another behavior.
In addition to what is already included in our paper, we have run experiments that further investigate impacts on non-target capabilities, which are detailed below. We will include these results in our revised submission. In the following experiments, guidelines and heldout sets are the same from the submission. We have measured safety preservation under unrelated steering by evaluating the sycophancy and myopic-reward steering directions on held-out AdvBench prompts and reporting refusal rates at multipliers up to α = ±1. We have also measured the impact on myopic-reward and sycophancy when conducting angular steering for refusal.
Results for OLMO on HarmBench are shown below, where the baseline model has 1.0% ASR. We can observe that HarmBench ASR rises by at most 0.9% (one prompt). The configuration and settings used are the same as in the MCQ section of the paper, and we report results for seed 0 only, across the same 104 test prompts in AdvBench that we report in Section 4.4.
OLMO Peak ASR (HarmBench) by setting