acceptodds
Withdrawn from ICLR 2027

MEME: Modeling the evolutionary modes of financial markets

Withdrawn from ICLR 2027, so left out of the paper lists and the map.

Abstract

Large Language Models (LLMs) have demonstrated significant potential in quantitative finance by processing vast unstructured data to emulate human-like analytical workflows. However, current LLM-based methods primarily follow either an Asset-Centric paradigm focused on individual stock prediction or a Market-Centric approach for portfolio allocation, often remaining agnostic to the underlying reasoning that drives market movements. In this paper, we propose a Logic-Oriented perspective, modeling the financial market as a dynamic, evolutionary ecosystem of competing investment narratives, termed Modes of Thought. To operationalize this view, we introduce MEME (Modeling the Evolutionary Modes of Financial Markets), a framework designed to reconstruct market dynamics through the lens of evolving logics. MEME employs a multi-agent extraction module to transform noisy data into high-fidelity Investment Arguments and utilizes Gaussian Mixture Modeling (GMM) to uncover latent consensus within a semantic space. To mitigate “logical noise” and semantic drift, we also implement a temporal evaluation and alignment mechanism to track the lifecycle and historical profitability of these modes. By prioritizing enduring market wisdom over transient anomalies, MEME ensures that portfolio construction is guided by robust, verified reasoning. Extensive experiments on three heterogeneous Chinese stock pools (SSE 50, CSI 300, and CSI 500) from 2023 to 2025 demonstrate that MEME consistently outperforms eight state-of-the-art baselines. Further ablation studies, sensitivity analyses, and lifecycle visualizations validate MEME's capacity to identify and adapt to the evolving consensus of financial markets. Our implementation can be found at https://anonymous.4open.science/r/MEME-FA3C.

Primary area:
foundation or frontier models, including LLMs
Keywords:
llm agents, evolution modes, portfolio construction

Reviews

Among:
Shape:
1234this paper, Bottom 10%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 1.67 over 3 reviews, above 10% of the others.
Reviewer_U6gYOpenReview ↗
Rating
2
Confidence
3
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

The paper proposes MEME, an LLM-based stock selection framework motivated by Narrative Economics. An LLM extracts bullish/bearish "investment arguments" for each stock from fundamental, news and technical data. These arguments are embedded and clustered daily with a GMM into "Modes of Thought". Modes are matched across days via Hungarian assignment, and each mode is scored by an EMA of its realised, polarity-signed excess returns. A stock's signal is the mode-score-weighted difference between its bullish and bearish arguments, and the portfolio holds the top-K stocks with equal weights. Experiments on SSE 50, CSI 300, CSI 500 and Nasdaq 100 (2023Q4–2025Q3, plus a 2025Q4 out-of-sample quarter) report gains over seven baselines, together with ablations and a case study linking modes to value and trend-following styles.

Critical strengths
  • The motivation behind is interesting. Clustering LLM-generated investment arguments and asking which types of reasoning are currently profitable is an appealing alternative to black-box return prediction. The case study gives some qualitative insight, for example dividend and balance-sheet arguments working in the bear market and momentum arguments in the rally.

  • The evaluation covers four stock pools in two markets and seven baselines from ML, DL and LLM-agent families. It also includes ablations, sensitivity analysis, a signal-orthogonality analysis, an earnings-season split and a post-cutoff test.

  • Transparency. Code and prompts are released. The limitations section openly reports a case where MEME underperforms (CSI 1000, where MASTER beats it).

Critical weaknesses
  • The method is naive. Plugging Eq. 12 into Eq. 13 is basically a linear factor model on K text clusters, where each factor's return is forecast by an EMA of its own past returns. With λ = 0.5 the memory is about one period, and since K is fixed with a forced one-to-one matching, modes can never "emerge or decay". The narrative-lifecycle framing therefore has no real counterpart in the method.

  • Data leakage is not tackled. The main period (2023Q4–2025Q3) largely falls within DeepSeek V3.1's training data, and the post-cutoff test covers only one quarter. In that quarter MEME's advantage largely disappears (CSI 300 IC 4.12 vs. 4.09), and its IC drops by 39%, while DTML and LightGBM barely change.

  • The backtest is unrealistic. The transaction cost of 1.5e-4 is below the A-share stamp duty alone (5e-4 on sells since Aug 2023). With daily rebalancing and no reported turnover, the AR and SR results are not credible.

  • The GMM is ill-posed. It uses full covariance on 4096d embeddings with only about 40–200 samples per component, so each $ \Sigma_k $ has rank at most nk−1n_k − 1 and the remaining eigenvalues equal reg_covar = 1e-6. The "soft assignment" explanation for why PM beats k-means does not hold.

  • Minor. Missing references:

    • Narrative Asset Pricing: Interpretable Systematic Risk Factors from News Text. RFS, 2023.
    • Factor Momentum and the Momentum Factor. JF, 2022.
    • The Memorization Problem: Can We Trust LLMs' Economic Forecasts? arXiv, 2025.
    • Can LLM-based Financial Investing Strategies Outperform the Market in Long Run? KDD, 2026.
Questions for authors

Please see the weakness section

Reviewer_UWsxOpenReview ↗
Rating
2
Confidence
4
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

This paper proposes MEME, which extracts structured investment arguments through a multi-agent system, identifies latent investment modes using probabilistic clustering, and constructs stock portfolios through temporal alignment and historical return evaluation.

Critical strengths

I find the research motivation appealing. Rather than directly predicting stock returns, the authors attempt to understand financial markets from the perspective of evolving investment logic. The overall method is clear, the experiments cover different markets, and the paper provides fairly extensive ablation studies.

Critical weaknesses
  1. The paper interprets clusters of LLM-generated investment arguments as market consensus, but I find this interpretation insufficiently validated. Semantically similar arguments do not necessarily represent actual market consensus. The current case studies mainly demonstrate the interpretability of the clusters, rather than proving that these modes have independent economic significance.
  2. The authors emphasize the emergence, evolution, and disappearance of investment modes, but the actual method uses a fixed number of GMM clusters and one-to-one matching between consecutive days. I hope the authors can discuss whether this design can truly capture the emergence, splitting, or disappearance of modes.
  3. I hope the authors can more clearly explain the temporal handling of financial data, especially news publication times, financial statement disclosure dates, and historical stock index constituents. These details directly affect the credibility of the results.
Questions for authors

Among the issues above, my main request is for the authors to clarify the temporal alignment of the financial data.

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

The paper proposes MEME which models evolving financial market narratives by extracting investment arguments from multimodal data using LLMs. The goal is to capture the "why" behind price movements and try to exploit this information for portfolio construction.

Critical strengths

(1) I personally like the idea of narrative economics and appreciate seeing work in this direction. This is indeed a new thing which is possible due to the advent of LLMs.

(2) The portfolio results seems promising.

(3) They test their method on four stock pools with distinct styles across the Chinese and U.S. markets. Not just a single market.

Critical weaknesses

(1) Limited novelty

Investment argument extraction uses LLM agents, mode identification relies on Gaussian Mixture Models, and temporal evolution uses Euclidean distance matching. Mode evaluation and portfolio construction are also relatively conventional. While the integration could itself be novel, related work such as [1] has already explored LLM based extraction and tracking of financial narratives.

[1] Deborah Miori, Constantin Petrov, "Narratives from GPT-derived Networks of News, and a link to Financial Markets Dislocations," 2024.

(2) Incomplete modeling of narrative evolution

Equation 6 imposes one to one matching between clusters across time. This does not explicitly capture narrative births, deaths, splits, or mergers. I am therefore not convinced that the method adequately models an evolutionary ecosystem.

(3) Insufficient validation of investment reasoning

It is unclear whether the identified clusters genuinely represent meaningful investment reasoning or market consensus, rather than simply similarities in LLM generated text.

(4) Limited broader ML contribution

MEME demonstrates a useful application specific combination of existing techniques. However I struggle to identify a new ML algorithm, modeling principle, or generalizable insight. I might evaluate this differently at a finance focused venue such as ICAIF, but this is ICLR. I wanted to see the broader machine learning contributions.

Questions for authors

(1) The investment reasoning is largely assumed rather than validated. Is this understanding correct? Beyond the qualitative examples is there any independent validation that these clusters represent meaningful investment narratives?

(2) Could the authors clarify the methodological novelty compared to prior work such as [1]? What is fundamentally different beyond the choice of clustering method and portfolio construction?

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

Related papers

Discussion (0)

Sign in to comment.