acceptodds
Under review as a conference paper at ICLR 2027

Diffusion-based Uncertainty-Utility Alignment for Portfolio Optimization

Abstract

Portfolio optimization requires investment decisions that balance return and risk under uncertainty. Probabilistic forecasting provides an explicit representation of future return uncertainty. However, predictive uncertainty is not inherently aligned with investment utility: portfolio decisions can be sensitive to scenario-sampling variability, while forecasting-only training does not directly account for the investment consequences of the resulting weights. In this paper, we propose diffusion-based uncertainty–utility alignment (DualPO) to align both the exploitation and generation of predictive uncertainty with investment utility. Specifically, DualPO trains an uncertainty-conditioned allocator using differentiable distributional summaries of return scenarios generated by a pretrained diffusion forecaster, aligning portfolio decisions with investment utility. Meanwhile, DualPO backpropagates utility gradients through the allocator to adapt the forecaster, aligning scenario generation with investment utility while retaining denoising supervision. Experimental results on the S&P and CSI benchmarks show that DualPO outperforms state-of-the-art baselines.

Primary area:
learning on time series and dynamical systems
Keywords:
Portfolio optimization, diffusion models

Reviews

Among:
Shape:
1234this paper, Bottom 2%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 1.33 over 3 reviews, above 2% of the others.
Reviewer_BYV5OpenReview ↗
Rating
1
Confidence
2
Informativeness
2
Presentation
2
Soundness
1
Read review
Paper summary

This paper proposes DualIPO, a system that decides how much to allocate to each stock conforming a portfolio. It is composed by two networks. The first is a diffusion model, which generates many plausible future returns trajectories for each of the stock in the portfolio. The generated paths then are summarized into a few key statistics (average return, spread in the outcomes, etc.). The second component of the system is a neural network that takes as an input these statistics and outputs the corresponding portfolio weights. These weights are always non-negative and sum to one; meaning no short-selling or borrowing is allowed.

Training has two parts. First, the diffusion model trains on its own to just produce forecasting results. Then, both parts are trained together on the actual investment outcome. Training alternates between the two parts: in each step the forecaster is updated first, and the allocator is then updated using the forecaster's new scenarios. This pushes the forecaster to generate scenarios that lead to better investment choices.

The main claim of the paper is that training the proposed system on investment performance, rather than just on forecasting accuracy, leads to better portfolio allocations. The method is then tested on the 50 and 100 largest stocks of the US S&P 500 and the Chinese CSI 300, where the authors report the best Sharpe ratio in all four settings and compare it against some baselines: an equal-weight portfolio, a momentum strategy, classical mean–variance optimization, and two diffusion-based methods (STABLE and Diffusion-DFL).

Critical strengths
  • The experiments are careful about leaking information from the future , which is a common problem in these approaches. I did not find that it happens here.
  • The main idea makes sense here. Instead of solving an optimization problem to build. portfolio every time, you just take a neural network which is trained to learn to turn forecasting scenarios into portfolio weights. This is simpler and also faster once it has been trained. Also, training both the forecaster and allocator networks together on the final investment outcomes is the way to go here.
Critical weaknesses

This paper should be rejected because: (1) the testing period is too short to tell methods apart; (2) DualPO barely beats a simple equal-weight portfolio, while the other baselines look poorly tuned; (3) there is no proof that the diffusion model or the uncertainty it gives is what drives the results.

Reasons for rejection

  • The test period is only around six months (which is 120 trading days), and the results come from 3 seeds without any reported variance. Note the following: a Sharpe ratio is the mean of daily returns divided by their standard deviation, and estimating it over so few days can be very noisy. The daily Sharpe ratio is SR=μ/σ\mathrm{SR} = \mu / \sigma. The standard deviation of the estimated mean over TT days is σ/T\sigma / \sqrt{T}, so the standard deviation of the estimated Sharpe ratio is σ/Tσ=1T\frac{\sigma / \sqrt{T}}{\sigma} = \frac{1}{\sqrt{T}}. After annualizing, this becomes 252/T=252/T\sqrt{252} / \sqrt{T} = \sqrt{252 / T}, which is about 1.51.5 for T=120T = 120. This means that a strategy with a true Sharpe ratio of, say, 1.51.5 could give anything between −1.4-1.4 and 4.44.4 over this period of time just by pure chance. The gaps of 1.741.74 vs 1.731.73 against equal weight on S&P-100, or 1.821.82 vs 1.811.81 against the frozen ablation, are therefore clearly within this noise and not meaningful.

  • One of the paper’s central claims is not really supported: nothing is showing that the forecaster will produce meaningful uncertainty after training. More concretely, the method is presented as using the predictive uncertainty given by the forecaster multiple outputs, but then, in the second training phase, the forecasting (denoising) loss is heavily down-weighted, making the forecaster being rewarded by just producing scenarios that make the allocator network output weights that are profitable in the training period rather than producing realistic forecasts. This makes overfitting to the training period likely. Furthermore, forecast quality after this training stage is never measured.

Questions for authors
  • Could you put in the paper the per-seed means and standard deviations as well as to evaluate all methods on a longer test period? This would address the first weakness above.
  • How were the baselines tuned? Did they get a tuning budget similar to DualPO's one?
Reviewer_gg8ROpenReview ↗
Rating
1
Confidence
4
Informativeness
1
Presentation
2
Soundness
1
Read review
Paper summary

This paper proposes DualPO, a diffusion-based uncertainty-utility alignment framework for portfolio optimization, aiming to align predictive uncertainty with downstream investment utility. Specifically, DualPO employs an uncertainty-conditioned allocator that leverages distributional summaries of generated return scenarios, while utility gradients are backpropagated to jointly optimize the diffusion forecaster and portfolio allocation.

Critical strengths
  1. The motivation is meaningful. Instead of relying on point forecasts of future asset returns, the paper adopts distributional forecasting to explicitly capture predictive uncertainty.

  2. The proposed framework integrates probabilistic forecasting and downstream allocation in a coherent end-to-end pipeline.

Critical weaknesses
  1. The method is lack of novelty. The proposed framework primarily integrates existing techniques, the diffusion-based forecasting.

  2. The ablation results do not consistently demonstrate substantial improvements from generation alignment. For example, the improvement on S&P-50 is marginal, with ASR increasing only from1.81 to 1.82.

  3. The experiments are conducted on relatively small asset universes, containing only 50 or 100 stocks. Given that the S&P 500 and CSI 300 indices cover substantially larger asset universes, it would be valuable to evaluate DualPO on larger portfolios like 300 or 500 stocks.

  4. The manuscript contains inconsistencies in notation, terminology, and formatting. In particular, the inconsistent use of hyphens and dashes in technical terms affects the overall presentation quality and readability of the paper.

Questions for authors
  1. The split of training, validation and test datasets seems unreasonable. The test period covers only six months, why it is relatively short?

  2. The proposed utility-guided training directly modifies the diffusion forecaster using downstream investment objectives. Could this process distort the predicted return distributions? How do the authors ensure that the generated scenarios remain well-calibrated after utility alignment?

  3. Since the allocator only uses nine distributional statistics extracted from the generated scenarios, why is it necessary to generate complete future return trajectories using a diffusion model? Could these statistics be predicted directly from historical market information with a simple neural network?

Reviewer_VwFeOpenReview ↗
Rating
2
Confidence
3
Informativeness
2
Presentation
2
Soundness
1
Read review
Paper summary

The paper intended to utilize diffusion-based model as portfolio optimizer. The idea is to utilize the inference process of diffusion-based model to align investment risk. The method is evaluated on both CSI and S&P with top-k equities.

Critical strengths
  1. Alignment with diffusion-based model is an interesting direction to work with.
  2. The paper provides meaningful solution to address the problem.
Critical weaknesses
  1. The major concern is on the evaluation. The seperation of training, validation and testing appears to be strange. Additionally, the reviewer is not sure if the setup contains enough data instances to prevent overfitting in the model.
Questions for authors

Please address the reviewer's major concern in cons.

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Loading the map…

Discussion (0)

Sign in to comment.