- Rating
- 3(+1)
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 3
Read review
Paper summary
This paper concerns the topic of machine unlearning where by a model is trained on a dataset and then one wants to remove knowledge obtained by the model on some subset of the dataset, known as the forget set. The natural question that develops from this motivation is what is a good way to remove that knowledge from the model without affecting the models ability outside the forget set. In this paper, the authors approach this problem through the lens of quantization. Their approach is to view quantization as an actual unlearning operator and in this way they develop a quantization map that enables the model to forget a subset of the dataset while keeping capability outside the forget set. They then validate their approach on standard unlearning benchmarks MUSE and TOFU.
Critical strengths
-
The paper is written clearly with the authors taking the time to clearly explain the motivation of the problem and their methodology. One thing I really liked about the paper was the fact that the authors, in section 4, clearly explained how their quantization approach works and how it affects the forget and retain statistics (see 4.2).
-
There is some novelty in this approach in that while it is natural to apply quantization after applying an unlearning operator it is very interesting to consider whether quantization can be used as an unlearning operator itself.
Critical weaknesses
-
My main weakness with the paper was that the authors didn't really give any insight into why quantization can behave as an unlearning operator. They explain their method nicely in section 4 but why does this lead to decent unlearning results as shown in section 5? In some sense it is not surprising that such a method would work as unlearning itself can be thought of as trying to compress knowledge on the forget set, as perfect removal is probably impossible without high degradation on the retain set. However, understanding how their quantization as an unlearning operator does this via some analysis would have been much more insightful.
-
The authors predominately focused on 4-bit quantization but it would have been good if they considered other bit quantization as a form of comparison and giving more granularity to their approach.
-
I would have liked to see some limitations of the method especially since the paper doesn't explain the key driver of why this method seems to work.
Questions for authors
-
Could you give some insight into why your method works?
-
What happens when you try other post-training quantization methods possibly with other bit quantization?
13 comments
W2: Results at other bit widths
At 3-, 4-, 6-, and 8-bit precision, our method lowers the memorization metrics on the forget data to near the level of the retrained model. The full-precision target model has KnowMem and VerbMem of 57.89 and 64.43 on News and 47.12 and 99.70 on Books. In contrast, across all bit widths, our method yields KnowMem of 30.88–34.66 and VerbMem of 20.37–22.57 on News and 26.10–30.42 and 13.23–21.19 on Books. These values are close to those of the retrained model, which has KnowMem and VerbMem of 32.42 and 20.50 on News and 30.33 and 14.53 on Books. The NRD of our method also ranges from 0.07 to 0.17 at every bit width, lower than that of the target model (0.29 on News and 0.33 on Books).
At 3-bit precision, the quantization error itself is large, so even the retrained model reaches an NRD of 0.09 on News and 0.11 on Books from quantization alone, higher than at the other bit widths. Our method obtains the results above even at 3-bit precision, where the quantization error grows. We attribute this to its separate control of the ordinary quantization error through reconstruction-based compensation on retain data. We leave extending this principle to quantization methods that support extremely low precision of 3 bits or fewer for future work.
Results of our method on MUSE at 3-, 4-, 6-, and 8-bit precision.
| News KnowMem | News VerbMem | News PrivLeak | News MU | News NRD | Books KnowMem | Books VerbMem | Books PrivLeak | Books MU | Books NRD | |
|---|---|---|---|---|---|---|---|---|---|---|
| Retrain (FP) | 32.42 | 20.50 | −0.06 | 55.69 | – | 30.33 | 14.53 | −0.14 | 68.98 | – |
| Target (FP) | 57.89 | 64.43 | −99.81 | 55.52 | 0.29 | 47.12 | 99.70 | −57.34 | 69.13 | 0.33 |
| 3bit retrain | 21.15 | 20.32 | −10.47 | 37.35 | 0.09 | 18.53 | 14.30 | −6.32 | 41.91 | 0.11 |
| 4bit retrain | 34.65 | 17.47 | −0.82 | 43.37 | 0.05 | 27.85 | 14.14 | −4.64 | 59.20 | 0.04 |
| 6bit retrain | 33.94 | 20.50 | −0.57 | 53.65 | 0.01 | 29.34 | 14.08 | −0.48 | 69.88 | 0.01 |
| 8bit retrain | 31.70 | 20.56 | −0.17 | 55.25 | 0.00 | 29.02 | 14.12 | −0.33 | 68.75 | 0.01 |
| 3bit ours | 30.88 | 22.57 | −94.27 | 35.32 | 0.17 | 30.42 | 21.19 | −57.36 | 40.92 | 0.16 |
| 4bit ours | 32.97 | 20.68 | −67.20 | 37.30 | 0.13 | 26.10 | 16.03 | 4.11 | 47.30 | 0.07 |
| 6bit ours | 34.66 | 20.37 | −51.91 | 42.66 | 0.10 | 27.13 | 13.23 | −59.19 | 47.17 | 0.15 |
| 8bit ours | 31.25 | 20.74 | −46.18 | 38.29 | 0.10 | 28.93 | 17.18 | −55.38 | 48.82 | 0.14 |
W3: Limitation under re-quantization of the unlearned model to a lower bit width
Our method constructs the unlearned model directly at a given bit width and thus removes the post-unlearning quantization stage. Quantizing this model again to a lower bit width, however, can reduce its agreement with the retrained model. The table below reports the results of further quantizing the 8-bit model obtained by our method to 4-bit precision.
Results on MUSE when the 8-bit model of our method is further quantized to 4-bit precision.
| News KnowMem | News VerbMem | News PrivLeak | News MU | Books KnowMem | Books VerbMem | Books PrivLeak | Books MU | |
|---|---|---|---|---|---|---|---|---|
| Ours 4-bit | 32.97 | 20.68 | −67.23 | 37.30 | 26.10 | 16.03 | 4.11 | 47.30 |
| Ours 8-bit → 4-bit | 12.16 | 18.72 | −50.00 | 31.68 | 20.32 | 16.23 | −58.39 | 37.29 |
After the additional quantization, the memorization metrics stay at or below their 8-bit values, but the decrease in KnowMem and MU moves the model farther from the retrained model, raising NRD from 0.10 to 0.17 on News and from 0.14 to 0.19 on Books. Our method, however, quantizes a model to its final bit width in a single stage, so applying it directly at the desired bit width can avoid this degradation.
Q2: Other post-training quantization methods (AWQ)
To examine whether our method is limited to the sequential quantization of GPTQ-family methods, we implemented it on top of AWQ (Lin et al., 2024). Our method consists of two components.
- (1) Code assignment: we displace the codes of forget-sensitive weights according to the intervention score.
- (2) Error compensation: we compensate the error introduced by this displacement so that the layer output on retain data is preserved, while reducing compensation along forget-sensitive directions.
AWQ absorbs per-input-channel scales determined from activation statistics into the weights and then quantizes them with group-wise round-to-nearest (RTN), and unlike GPTQ, it has no per-weight error compensation step. We therefore implement (1) by applying the intervention score and the code displacement directly in the RTN step. For (2), we move the compensation to the group scales, which are the quantization parameters of AWQ. Specifically, with the codes fixed, we refit the group scales by least squares with respect to the retain statistic , and attenuate the forget-related component of this correction with the same attenuation factors as in the GPTQ version. We keep the scale and clipping search of AWQ unchanged. The table below reports the results on MUSE News.
Results on MUSE News with our method implemented on AWQ and GPTQ.
| Method | KnowMem | VerbMem | PrivLeak | MU |
|---|---|---|---|---|
| Retrain (FP) | 32.42 | 20.50 | −0.06 | 55.69 |
| Target (4bit) | 55.95 | 49.23 | −99.81 | 49.49 |
| Ours (AWQ 4bit) | 31.10 | 25.89 | −95.49 | 34.17 |
| Ours (AWQ 8bit) | 32.14 | 22.07 | −30.81 | 29.95 |
| Ours (GPTQ 4bit) | 32.97 | 20.68 | −67.23 | 37.30 |
| Ours (GPTQ 8bit) | 31.25 | 20.74 | −46.18 | 38.29 |
Our method also removes knowledge of the forget data when built on AWQ. At 4-bit precision, KnowMem and VerbMem decrease from 55.95 and 49.23 under AWQ alone to 31.10 and 25.89, respectively, approaching the retrained reference, and NRD falls from 0.26 to 0.18. At 8-bit precision, the AWQ version attains an NRD of 0.11, similar to the 0.10 of the GPTQ version, and a PrivLeak of −30.81, closer to the retrained reference than the −46.18 of the GPTQ version. In contrast, MU is lower than that of the GPTQ version at both bit widths, and at 4-bit precision PrivLeak remains at the level of the target model (−95.49). We attribute this to the AWQ version performing compensation only by refitting the group scales. This compensation has fewer degrees of freedom than that of GPTQ, which updates all remaining unquantized weights. We leave refining the compensation scheme for the structure of AWQ to future work.
References
- Lin et al. AWQ: Activation-aware Weight Quantization for On-Device LLM Compression and Acceleration. MLSys 2024.
We thank the reviewer for the careful and constructive comments. Below we provide a point-by-point response. Where the reviewer requested additional evidence, we include new experiments: results of our method at 3-, 4-, 6-, and 8-bit precision, further quantization of our 8-bit model to 4-bit precision, and an implementation of our method on top of AWQ. Throughout, the retrained model (trained from scratch with the forget data removed) is the reference that unlearning aims to approximate, and NRD is our aggregate distance to this reference across the four metrics (KnowMem, VerbMem, PrivLeak, MU).
W1 / Q1: Why quantization can act as an unlearning operator
Quantization is a form of knowledge compression. It aims to compress the knowledge a model acquires during training into fewer bits while retaining it with minimal loss. This compression inevitably introduces quantization error. Prior quantization methods have accordingly focused on how to handle this error.
AWQ identifies salient weights from activation magnitudes and protects them by reducing their quantization error through scaling. GPTQ applies error compensation, which propagates the error from quantizing each weight to the remaining unquantized weights. This compensation keeps the layer output close to that of the original model. These two methods yield two insights. First, weights differ in their importance for retaining knowledge, and this importance can be identified from activation statistics. Second, controlling where quantization error remains and how it is compensated is central to knowledge compression.
We adopt both insights and extend each by distinguishing forget data from retain data.
Extending weight importance to forget data. We compute activation statistics separately on forget and retain data to identify forget-sensitive weights. We displace the codes of these weights away from their nearest codes, intentionally increasing their quantization error (code displacement).
Extending quantization error control to forget and retain data. The existing GPTQ compensation can also offset the code displacement. We therefore separate the error into the quantization error of the nearest code and the code displacement error, and control the two separately. Only the code displacement error is propagated along a direction whose forget-related component is attenuated. The layer output on retain data is therefore preserved, whereas the error on forget data remains.
Our method thus builds directly on the two insights by which conventional quantization handles error, but amplifies quantization error only on the forget side. Applying the compression of quantization asymmetrically to forget and retain data in effect implements the reviewer's view of unlearning as trying to “compress knowledge on the forget set.”
In conventional unlearning, forgetting is induced by small continuous parameter updates, which can be removed by rounding. In our method, forgetting is encoded as integer changes in the assigned codes and therefore persists after rounding. The ablation study in Table 3 of the paper also shows that both mechanisms are necessary. Replacing our compensation with the standard GPTQ compensation offsets the code displacement and thereby undoes the forgetting. Applying the same amount of code displacement to randomly selected weights severely degrades model utility.
This insight extends beyond GPTQ-based quantization. We applied it to AWQ and confirmed that AWQ-based quantization is also sufficient to achieve unlearning. We describe these results in detail in our response to Q2.