- Rating
- 1
- Confidence
- 2
- Informativeness
- 1
- Presentation
- 2
- Soundness
- 1
Read review
Paper summary
The paper looks at how LLM-based search engines rank products, and argues that models favor items they already know from pretraining over unfamiliar ones. To quantify this, the authors propose a simple score (PRS) based on a controlled ranking experiment. They then propose a decoding method, CARD, that contrasts predictions with and without the retrieved context. Depending on the sign of a parameter, it either amplifies the context to reduce this bias, or suppresses it to defend against some poisoning and SEO-style attacks.
Critical strengths
The question is relevant: LLM-based search engines do act as recommenders, and whether they favor well-known brands over new ones, or can be steered by injected content, matters in practice. The controlled ranking setup, which balances descriptions and context positions, is a reasonable way to probe this.
Critical weaknesses
-
The PRS metric comparison is between the model’s favorite real products and obviously fake names like “S Brand”, so the gap may just reflect that the names look fake(the appendix itself shows the gap shrinks with more realistic fictional names). The main result cannot support the main claims of the paper.
-
Why preferring a known brand, given the same description, should be seen as a bias.
-
The method seems to be close to CAD and CK-PLUG with an extra scaling term. The ablation does not re-tune the other parameter, so it does not really show that this term helps.
-
In the defense results: the attacked item is always a fake-named product, so suppressing the context pushes it down, whether it was attacked or not; The direction is also set by hand, which assumes one already knows an attack is present.
Questions for authors
-
If in a real deployment, how to decide whether to amplify or suppress the retrieved context, given that it is usually not known in advance whether the content has been attacked?
-
Why preferring a well-known brand, when the descriptions are the same, should be treated as a bias to remove rather than a reasonable choice from the user’s point of view.
3 comments
W3: Differences from CAD and CK-PLUG, and the Ablation
1. Differences from CAD and CK-PLUG. As stated in the Introduction and Appendix D, CARD is fundamentally different from CK-PLUG. Below, we further explain three core differences, also in relation to CAD. The output of CAD can be written as ; the output of CARD is (Eq. 9). The three differences are as follows:
- Adaptive intervention strength. CAD uses a fixed throughout; CK-PLUG determines, based on the distribution shift, whether the current decoding step involves a knowledge conflict, and then decides whether to intervene. In contrast, CARD continuously determines the intervention strength from the KL divergence at each step, upper-bounded by (Eq. 8): the greater the influence of the retrieved content on the prediction, the more gently the intervention grows, which avoids over-amplification.
- PRS calibration. is exactly the model's PRS, measured in the Latin-square controlled experiment, and is a model-level empirical value. CARD uses to calibrate the anchoring weight of : the larger the model's reliance bias, the lower the weight of in the final output. Neither CAD nor CK-PLUG has this model-level calibration.
- Designed for multi-candidate ranking. CK-PLUG targets knowledge conflicts in factual question answering, aiming to decide whether a single answer should rely on parametric knowledge or external knowledge. In recommendation, multiple candidate items compete simultaneously, and the influence of parametric knowledge manifests as a systematic shift of the entire ranking. This difference is reflected in the results: in Table 2 of the paper, CK-PLUG (Ada.) increases PRS on all three models (e.g., LLaMA 0.2568 → 0.3317), whereas CARD reduces it to 0.1041.
2. Ablation of . We ablate by setting it to different constants. As shown in Tables 2 and 3, clearly outperforms on both tasks; on both tasks, the optimal region lies around 0.25, and the measured by PRS falls right in this region, with no tuning required. This demonstrates the effectiveness of PRS calibration ().
Table 2: Fixed- ablation for bias mitigation (LLaMA3.1-8B).
| 0 | 0.125 | 0.25 | 0.2568 (PRS) | 0.375 | |
|---|---|---|---|---|---|
| PRS ↓ | 0.0974 | 0.0901 | 0.0873 | 0.0890 | 0.0902 |
Table 3: Fixed- ablation for TAP defense (LLaMA3.1-8B). : average ranking score of the target item.
| Method | ↓ | ↑ |
|---|---|---|
| Vanilla | 2.2225 | – |
| TAP | 4.5875 | 0% |
| CARD () | 3.0325 | 65.8% |
| CARD () | 2.6700 | 81.1% |
| CARD () | 2.5675 | 85.4% |
| CARD (, PRS) | 2.5925 | 84.4% |
| CARD () | 2.7700 | 76.8% |
W4 (first part): Does the Defense Merely Suppress Unfamiliar Item Names?
As shown in Table 4, after replacing the TAP attack target with real items that are familiar to the model, CARD still achieves the highest on all three models. This demonstrates the effectiveness of CARD's defense: its defense effect comes from suppressing the influence of external attack content, rather than merely from systematically pushing down unfamiliar item names.
Table 4: TAP defense with familiar item names as attack targets. : average ranking score of the target item.
| LLaMA3.1-8B | Qwen3-8B | Gemma3-12B | ||||
|---|---|---|---|---|---|---|
| Method | ↓ | ↑ | ↓ | ↑ | ↓ | ↑ |
| Vanilla | 2.3386 | – | 2.0531 | – | 1.8031 | – |
| TAP | 4.9286 | 0% | 4.9519 | 0% | 4.9156 | 0% |
| CAD | 3.6039 | 51.1% | 3.8069 | 39.5% | 3.9908 | 29.7% |
| CK-PLUG (Ada.) | 4.1522 | 30.0% | 4.5731 | 13.1% | 4.0939 | 26.4% |
| CK-PLUG (Fixed) | 3.2014 | 66.7% | 4.3186 | 21.8% | 3.7992 | 35.9% |
| CARD (Ours) | 2.8478 | 80.3% | 2.6547 | 79.2% | 3.1997 | 55.1% |
W4 (second part) & Q1: How to Choose the Direction in Deployment
CARD is designed for bidirectional control over the LLM's knowledge reliance (Introduction), targeting deployment environments where the retrieved content is untrustworthy or, conversely, must be trusted. As stated in Appendix D, the sign of is a deployment-time configuration, set by the deployer (user) according to the deployment setting and the retriever trust policy:
- When the retrieved content is reliable (e.g., using verified private documents): the deployer can use CARD to amplify reliance on the retrieved content, which is especially important when the model's parametric knowledge is known to be outdated.
- When the retrieved content is unreliable (e.g., retrieval results are noisy), or when the user wishes to rely on the LLM's own judgment: the deployer can suppress the influence of external content and rely on the model's robust parametric knowledge, effectively defending against adversarial manipulation.
Therefore, the choice of direction is not an unstated assumption or weakness; it is precisely CARD's defining feature.
We thank the reviewer again for the thoughtful feedback and would be grateful for further discussion. We hope that the clarifications and additional analyses address the concerns and may support a more positive assessment of our work.
We thank the reviewer for their time and insightful feedback. The comments are invaluable for improving our work and clarifying our contributions. We believe there may be some misunderstandings about our paper, and we clarify and respond to each point below.
W1: Placeholder Names
1. PRS under the two name constructions. In Appendix B of the revised version, we add quartile plots of PRS for LLaMA3.1-8B, Qwen3-8B, Gemma3-12B, and Gemini-3.1 when the non-parametric-knowledge items are represented by placeholders and by fictional-but-plausible names, respectively. Except for LLaMA, the interquartile ranges of PRS under the two kinds of names largely overlap, with no notable gap. We conjecture that the difference on LLaMA is related to its earlier release and weaker capability, and is an isolated case. This shows that the gap measured by PRS consistently exists under both kinds of names.
2. CARD under fictional-but-plausible names. We evaluate CARD on fictional-but-plausible names, directly reusing the M and β from the placeholder setting, without re-tuning. As shown in Table 1, with fictional-but-plausible names, the PRS of Vanilla remains clearly positive on all three models, showing that this gap does not arise merely because the names look fake; CARD's PRS is lower than that of Vanilla and all baselines on all three models, showing that CARD's debiasing effect does not depend on placeholders. Therefore, the main conclusions of the paper still hold under fictional-but-plausible names.
Table 1: PRS under fictional-but-plausible names.
3. Why the main experiments still use placeholders. We use placeholders to represent non-parametric-knowledge items. This is a deliberate design: it guarantees that these items are genuinely absent from the model's parametric knowledge, which is the prerequisite for isolating the factor of "whether an item belongs to parametric knowledge". LLM-generated fictional names that read more naturally cannot achieve this: because LLMs are trained on large-scale web corpora, even low-probability sampling may produce names that coincide with real entities. To quantify this, we prompted GPT-5.5 to generate 400 (50 × 8) natural-sounding fictional item names and searched each of them on the web: 214 (53.5%) of them have real-world counterparts. Therefore, names that look fictional cannot be guaranteed to lie outside the model's parametric knowledge, whereas placeholders provide a clean, verifiable control.
W2 & Q2: Why Favoring Well-Known Items Counts as Bias
The bias in this paper refers to the systematic preference in LLM recommendation rankings caused by over-reliance on parametric knowledge. This reliance is implicit: users see only the final ranking, and can neither know how much of it comes from the model's parametric knowledge nor control it. This amounts to the LLM making the decision to "favor well-known items" on the user's behalf. We agree that, for some users, favoring well-known items can be a reasonable choice. Therefore, CARD does not forcibly eliminate this reliance, but makes it controllable: with M>0, the ranking relies more on the retrieved content; with M<0, the ranking relies more on the model's internal knowledge. Whether to favor well-known items is decided by users according to their own needs, rather than implicitly by the model.