Leveraging Large Language Models for Sentiment Analysis: Multi-Modal Analysis of Decentraland's MANA Token

# Leveraging Large Language Models for Sentiment Analysis: Multi-Modal Analysis of Decentraland’s MANA Token
Source: [https://arxiv.org/html/2605.20192](https://arxiv.org/html/2605.20192)
Peiting Tsai, Jing Yuan, Michael Yu, and Greg SunLuyao Zhang\*\*Corresponding author: Luyao Zhang \(lz183@duke\.edu\), Digital Innovation Research Center and Social Science Division, Duke Kunshan University\. Address: Duke Avenue No\.8, Kunshan, Suzhou, Jiangsu, China, 215316\.Acknowledgments: Luyao Zhang acknowledges support from the National Science Foundation of China \(NSFC\) for the project titled ”Trust Mechanism Design on Blockchain: An Interdisciplinary Approach of Game Theory, Reinforcement Learning, and Human\-AI Interactions” \(Grant No\. 12201266\)\. Xintong Wu is grateful for the support provided by the Office of Academic Services, the Summer Research Scholar Program, and the Student Experiential Learning Fellow \(SELF\) Program at Duke Kunshan University, supervised by Prof\. Luyao Zhang\. Xintong also extends appreciation to all members of the One\-Person Entrepreneur \(OPE\) group at Microsoft China for their valuable inspiration\.

###### Abstract

Decentraland, a decentralized virtual reality platform operating within the expanding Metaverse ecosystem, utilizes its native MANA token to facilitate virtual asset transactions and governance\. This study investigates the integration of Discord community sentiment with multi\-modal financial data to enhance cryptocurrency price prediction within virtual world economies\. We address: \(1\) identifying sentiment patterns within Decentraland’s Discord community, and \(2\) evaluating the impact of multi\-modal features on token return forecasting\. Using a BERT\-based large language model for sentiment analysis, we develop two LSTM architectures: a baseline incorporating historical prices and a multi\-modal variant integrating sentiment scores, trading volume, and market capitalization\. Results indicate predominantly neutral community sentiment with a positive skew\. The multi\-modal model significantly outperforms the price\-only baseline in prediction accuracy\. These findings demonstrate the predictive value of community\-derived signals for virtual economy forecasting and establish a foundation for future research at the intersection of immersive virtual environments, natural language processing, and cryptocurrency market analysis\.

## IIntroduction

The emergence of blockchain\-enabled immersive virtual environments—the Metaverse—has catalyzed interest in persistent digital economies where users transact virtual assets through native cryptocurrencies\[[28](https://arxiv.org/html/2605.20192#bib.bib81),[50](https://arxiv.org/html/2605.20192#bib.bib82)\]\. Decentraland represents a paradigmatic case of this convergence, combining virtual reality with decentralized infrastructure to create economic spheres with unique market dynamics\[[20](https://arxiv.org/html/2605.20192#bib.bib29)\]\. Its native token, MANA, facilitates transactions for virtual goods and land parcels while conferring governance rights through a decentralized autonomous organization \(DAO\), rendering it ideal for examining sentiment dynamics in nascent virtual economies\.

StartData PrepPreprocessLoad RoBERTaSentimentInferenceTemporalAggregationEndStS\_\{t\}LSTM InputSt=1nt​∑si,t​γi,tS\_\{t\}=\\frac\{1\}\{n\_\{t\}\}\\sum s\_\{i,t\}\\gamma\_\{i,t\}TrainingEvaluationFine\-tuningEvaluationDiscordTokenRoBERTasi,t∈\{\+1,0,−1\}s\_\{i,t\}\\in\\\{\+1,0,\-1\\\}τt,Vt\\tau\_\{t\},V\_\{t\}𝐱t=\(τt,Vt,St\)\\mathbf\{x\}\_\{t\}=\(\\tau\_\{t\},V\_\{t\},S\_\{t\}\)Figure 1:LLM\-based sentiment analysis pipeline\.Red circle highlights sentiment outputStS\_\{t\}feeding into downstream LSTM alongside conventional technical indicators such as priceτt\\tau\_\{t\}and volumeVtV\_\{t\}\.Investor sentiment from social media constitutes a significant predictor of market behavior, prompting extensive applications of sentiment analysis in financial forecasting\[[22](https://arxiv.org/html/2605.20192#bib.bib10)\]\. Within Decentraland, Discord serves as the primary communication channel where user discussions may influence MANA price dynamics\. Recent advances in large language models \(LLMs\) have enhanced sentiment analysis capabilities, with multi\-agent frameworks improving financial classification accuracy\[[54](https://arxiv.org/html/2605.20192#bib.bib79)\]and specialized models advancing emotion detection\[[37](https://arxiv.org/html/2605.20192#bib.bib80)\]\.

This study constructs a multi\-modal dataset integrating historical price data with Discord\-derived sentiment, trading volume, and market capitalization to forecast token returns\. We investigate two research questions \(RQs\):

- •\(RQ1\)What are the prevailing sentiment patterns among Decentraland Discord users, and to what extent do these reflect platform dynamics?
- •\(RQ2\)How does integrating multi\-modal features affect token return prediction accuracy relative to price\-only models?

Employing a BERT\-based architecture for sentiment extraction and LSTM networks for time\-series prediction, we evaluate the incremental value of community sentiment for cryptocurrency forecasting\. Figure[1](https://arxiv.org/html/2605.20192#S1.F1)illustrates the LLM\-based sentiment analysis pipeline\.111All data and code are publicly accessible at:https://github\.com/Xintong1122/Decentraland\_Senti

## IIData

### II\-AData Collection

#### II\-A1Numerical Data

Daily OHLCV data for MANA \(ERC\-20 utility token\[[11](https://arxiv.org/html/2605.20192#bib.bib1),[44](https://arxiv.org/html/2605.20192#bib.bib2),[26](https://arxiv.org/html/2605.20192#bib.bib3)\]\) spanningt∈\[1,T\]t\\in\[1,T\]whereT=366T=366days \(June 15, 2023–June 14, 2024\) comprise open, high, low, close prices \(PtO,PtH,PtL,PtCP\_\{t\}^\{O\},P\_\{t\}^\{H\},P\_\{t\}^\{L\},P\_\{t\}^\{C\}\), trading volumeVtV\_\{t\}, and market capitalizationMtM\_\{t\}\(USD\), obtained fromCoinMarketCap\.222https://coinmarketcap\.com/

#### II\-A2Textual Data

Discord\#generalchannel messagesℳ=\{mi\}i=1n\\mathcal\{M\}=\\\{m\_\{i\}\\\}\_\{i=1\}^\{n\}\(n=5,883n=5,883\) were harvested viaDiscordChatExporter,333https://github\.com/Tyrrrz/DiscordChatExporteryielding timestampstit\_\{i\}, anonymized author IDsαi\\alpha\_\{i\}, contentcic\_\{i\}, and engagement metrics\. Post\-filtering invalid entries,n′=5,513n^\{\\prime\}=5,513messages \(ℳ′⊂ℳ\\mathcal\{M\}^\{\\prime\}\\subset\\mathcal\{M\}\) comprised the final corpus\.

### II\-BData Processing

#### II\-B1Price Metric

The typical priceτt=13​\(PtH\+PtL\+PtC\)\\tau\_\{t\}=\\frac\{1\}\{3\}\(P\_\{t\}^\{H\}\+P\_\{t\}^\{L\}\+P\_\{t\}^\{C\}\)serves as the daily indicator, smoothing intraday fluctuations following standard technical analysis practice\[[42](https://arxiv.org/html/2605.20192#bib.bib7)\]\.

#### II\-B2Return Calculation

Continuous returnsrt=ln⁡\(τt\+1/τt\)=ln⁡\(τt\+1\)−ln⁡\(τt\)r\_\{t\}=\\ln\(\\tau\_\{t\+1\}/\\tau\_\{t\}\)\\\\ =\\ln\(\\tau\_\{t\+1\}\)\-\\ln\(\\tau\_\{t\}\)stabilize variance and ensure temporal comparability for ERC\-20 token markets\[[24](https://arxiv.org/html/2605.20192#bib.bib13)\]\.

## IIIMethodology

### III\-ASentiment Analysis

#### III\-A1LLM Selection

We employ the RoBERTa\-based classifiercardiffnlp/twitter\-roberta\-base\-sentiment\-latest444https://huggingface\.co/cardiffnlp/twitter\-roberta\-base\-sentiment\-latestfrom Hugging Face, pretrained on social media corpora for informal text classification\[[15](https://arxiv.org/html/2605.20192#bib.bib14),[38](https://arxiv.org/html/2605.20192#bib.bib15)\]\.

#### III\-A2Aggregation

Each messageiiat timettmaps to sentimentsi,t∈\{\+1,0,−1\}s\_\{i,t\}\\in\\\{\+1,0,\-1\\\}\(positive, neutral, negative\) with confidenceγi,t\\gamma\_\{i,t\}\. Daily sentiment aggregates asSt=1nt​∑i=1ntsi,t​γi,tS\_\{t\}=\\frac\{1\}\{n\_\{t\}\}\\sum\_\{i=1\}^\{n\_\{t\}\}s\_\{i,t\}\\gamma\_\{i,t\}, wherentn\_\{t\}denotes daily message volume\[[61](https://arxiv.org/html/2605.20192#bib.bib16)\]\. Discretization thresholds defineSt∈\[−1,−0\.1\)S\_\{t\}\\in\[\-1,\-0\.1\)as negative,\[−0\.1,0\.1\]\[\-0\.1,0\.1\]as neutral, and\(0\.1,1\]\(0\.1,1\]as positive\.

### III\-BReturn Prediction

#### III\-B1LSTM Architecture

Long Short\-Term Memory networks model sequential dependencies\{\(𝐱t,rt\)\}t=1T\\\{\(\\mathbf\{x\}\_\{t\},r\_\{t\}\)\\\}\_\{t=1\}^\{T\}via gated mechanisms\[[58](https://arxiv.org/html/2605.20192#bib.bib17),[46](https://arxiv.org/html/2605.20192#bib.bib19),[48](https://arxiv.org/html/2605.20192#bib.bib18)\]:

it=σ​\(Wi⋅\[ht−1,𝐱t\]\+bi\),ot=σ​\(Wo⋅\[ht−1,𝐱t\]\+bo\)i\_\{t\}=\\sigma\(W\_\{i\}\\cdot\[h\_\{t\-1\},\\mathbf\{x\}\_\{t\}\]\+b\_\{i\}\),\\quad o\_\{t\}=\\sigma\(W\_\{o\}\\cdot\[h\_\{t\-1\},\\mathbf\{x\}\_\{t\}\]\+b\_\{o\}\)\(1\)whereit,oti\_\{t\},o\_\{t\}denote input and output gates,σ\\sigmathe sigmoid function, andW,bW,blearned parameters\. Model specifications:

- •Baseline:𝐱t=τt\\mathbf\{x\}\_\{t\}=\\tau\_\{t\}\(scalar, typical price only\)
- •Multi\-modal:𝐱t=\(τt,Vt,Mt,St\)′∈ℛ4\\mathbf\{x\}\_\{t\}=\(\\tau\_\{t\},V\_\{t\},M\_\{t\},S\_\{t\}\)^\{\\prime\}\\in\\mathcal\{R\}^\{4\}

#### III\-B2Evaluation Metrics

Predictive performance is quantified viaMSE=T−1​∑t=1T\(rt−r^t\)2\\text\{MSE\}=T^\{\-1\}\\sum\_\{t=1\}^\{T\}\(r\_\{t\}\-\\hat\{r\}\_\{t\}\)^\{2\},MAE=T−1​∑t=1T\|rt−r^t\|\\text\{MAE\}=T^\{\-1\}\\sum\_\{t=1\}^\{T\}\|r\_\{t\}\-\\hat\{r\}\_\{t\}\|, andR2=1−∑t=1T\(rt−r^t\)2/∑t=1T\(rt−r¯\)2R^\{2\}=1\-\{\\sum\_\{t=1\}^\{T\}\(r\_\{t\}\-\\hat\{r\}\_\{t\}\)^\{2\}\}/\{\\sum\_\{t=1\}^\{T\}\(r\_\{t\}\-\\bar\{r\}\)^\{2\}\}\[[10](https://arxiv.org/html/2605.20192#bib.bib20),[31](https://arxiv.org/html/2605.20192#bib.bib21),[52](https://arxiv.org/html/2605.20192#bib.bib22),[5](https://arxiv.org/html/2605.20192#bib.bib23),[4](https://arxiv.org/html/2605.20192#bib.bib24)\]\.

## IVResults

Figure[2](https://arxiv.org/html/2605.20192#S4.F2)presents the dual analysis of community sentiment dynamics and predictive model performance\. The bar plot illustrates the distribution of Discord sentiment over the 366\-day observation period: 26\.94% positive \(si,t=\+1s\_\{i,t\}=\+1\), 60\.96% neutral \(si,t=0s\_\{i,t\}=0\), and 12\.10% negative \(si,t=−1s\_\{i,t\}=\-1\)\. This predominance of neutral sentiment with a positive skew indicates stable community affect, with daily aggregate scoresStS\_\{t\}concentrating within\[0,1\]\[0,1\]\. Notably, negative sentiment spikes correlate primarily with user\-level technical issues rather than exogenous market events\[[18](https://arxiv.org/html/2605.20192#bib.bib27),[19](https://arxiv.org/html/2605.20192#bib.bib25)\], suggesting these data reflect platform user experience rather than macroeconomic signals\.

The time\-series plots compare the baseline LSTM \(price\-only feature𝐱t=τt\\mathbf\{x\}\_\{t\}=\\tau\_\{t\}\) against the multi\-modal architecture \(𝐱t=\(τt,Vt,Mt,St\)′∈ℛ4\\mathbf\{x\}\_\{t\}=\(\\tau\_\{t\},V\_\{t\},M\_\{t\},S\_\{t\}\)^\{\\prime\}\\in\\mathcal\{R\}^\{4\}\)\. The multi\-modal model improves prediction accuracy, reducingMSEfrom0\.0021000\.002100to0\.0015280\.001528andMAEfrom0\.02970\.0297to0\.02410\.0241\. However, both models exhibit negativeR2R^\{2\}values \(baseline:−0\.4185\-0\.4185; multi\-modal:−0\.0321\-0\.0321\), indicating limited explanatory power for directional trends\. While auxiliary features enhance short\-term precision, they inadequately capture broader market dynamics, consistent with\[[33](https://arxiv.org/html/2605.20192#bib.bib86)\]\.

![Refer to caption](https://arxiv.org/html/2605.20192v1/fig/new.png)Figure 2:Price Prediction and Sentiment Categories\.
## VConclusion

Our study contributes to metaverse and blockchain\-empowered virtual economy research\[[16](https://arxiv.org/html/2605.20192#bib.bib47),[22](https://arxiv.org/html/2605.20192#bib.bib10),[32](https://arxiv.org/html/2605.20192#bib.bib83),[53](https://arxiv.org/html/2605.20192#bib.bib48),[12](https://arxiv.org/html/2605.20192#bib.bib30),[62](https://arxiv.org/html/2605.20192#bib.bib4),[56](https://arxiv.org/html/2605.20192#bib.bib5),[7](https://arxiv.org/html/2605.20192#bib.bib6)\]and the literature on user\-generated content and community sentiment\[[9](https://arxiv.org/html/2605.20192#bib.bib49),[3](https://arxiv.org/html/2605.20192#bib.bib50),[41](https://arxiv.org/html/2605.20192#bib.bib51),[57](https://arxiv.org/html/2605.20192#bib.bib52)\]\. By applying large language models to financial sentiment\[[30](https://arxiv.org/html/2605.20192#bib.bib37),[6](https://arxiv.org/html/2605.20192#bib.bib36),[27](https://arxiv.org/html/2605.20192#bib.bib38),[34](https://arxiv.org/html/2605.20192#bib.bib39),[36](https://arxiv.org/html/2605.20192#bib.bib33),[55](https://arxiv.org/html/2605.20192#bib.bib58),[60](https://arxiv.org/html/2605.20192#bib.bib59),[49](https://arxiv.org/html/2605.20192#bib.bib34),[13](https://arxiv.org/html/2605.20192#bib.bib35)\], we extend methodological approaches in sentiment analysis\[[39](https://arxiv.org/html/2605.20192#bib.bib31),[51](https://arxiv.org/html/2605.20192#bib.bib54),[29](https://arxiv.org/html/2605.20192#bib.bib55),[14](https://arxiv.org/html/2605.20192#bib.bib56),[21](https://arxiv.org/html/2605.20192#bib.bib57),[59](https://arxiv.org/html/2605.20192#bib.bib60),[43](https://arxiv.org/html/2605.20192#bib.bib61)\]to the cryptocurrency domain\[[17](https://arxiv.org/html/2605.20192#bib.bib68),[25](https://arxiv.org/html/2605.20192#bib.bib69),[8](https://arxiv.org/html/2605.20192#bib.bib65),[63](https://arxiv.org/html/2605.20192#bib.bib62),[40](https://arxiv.org/html/2605.20192#bib.bib63),[1](https://arxiv.org/html/2605.20192#bib.bib64),[45](https://arxiv.org/html/2605.20192#bib.bib67)\]\. Results show that incorporating sentiment enhances price\-only models\[[47](https://arxiv.org/html/2605.20192#bib.bib72),[2](https://arxiv.org/html/2605.20192#bib.bib71),[23](https://arxiv.org/html/2605.20192#bib.bib70)\], though gains remain limited as Discord sentiment reflects platform\-specific user experience rather than macro\-market conditions\. Future work should integrate additional data sources\[[35](https://arxiv.org/html/2605.20192#bib.bib77)\], improve domain\-specific modeling, and consider scalability and security aspects\[[28](https://arxiv.org/html/2605.20192#bib.bib81)\]\.555Detailed contributions to each literature stream are provided in Appendix[A](https://arxiv.org/html/2605.20192#A1)\.

## References

- \[1\]A\. S\. M\. Alharbi and E\. de Doncker\(2019\)Twitter sentiment analysis with a deep neural network: an enhanced approach using user behavioral information\.Cognitive Systems Research54,pp\. 50–61\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[2\]S\. Athey and G\. W\. Imbens\(2019\)Machine learning methods that economists should know about\.Annual Review of Economics11\(1\),pp\. 685–725\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p5.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[3\]A\. Z\. Bahtar and M\. Muda\(2016\)The impact of user–generated content \(ugc\) on product reviews towards online purchasing–a conceptual framework\.Procedia Economics and Finance37,pp\. 337–342\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p2.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[4\]A\. C\. Cameron and F\. A\. Windmeijer\(1997\)An r\-squared measure of goodness of fit for some common nonlinear regression models\.Journal of econometrics77\(2\),pp\. 329–342\.Cited by:[§III\-B2](https://arxiv.org/html/2605.20192#S3.SS2.SSS2.p1.3)\.
- \[5\]T\. Chai and R\. R\. Draxler\(2014\)Root mean square error \(rmse\) or mean absolute error \(mae\)?–arguments against avoiding rmse in the literature\.Geoscientific model development7\(3\),pp\. 1247–1250\.Cited by:[§III\-B2](https://arxiv.org/html/2605.20192#S3.SS2.SSS2.p1.3)\.
- \[6\]Y\. Chang, X\. Wang, J\. Wang, Y\. Wu, L\. Yang, K\. Zhu, H\. Chen, X\. Yi, C\. Wang, Y\. Wang,et al\.\(2024\)A survey on evaluation of large language models\.ACM Transactions on Intelligent Systems and Technology15\(3\),pp\. 1–45\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p4.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[7\]N\. Chemaya, L\. W\. Cong, E\. Joergensen, D\. Liu, and L\. Zhang\(2025\)A dataset of uniswap daily transaction indices by network\.Scientific Data12\(1\),pp\. 93\.External Links:[Document](https://dx.doi.org/10.1038/s41597-024-04042-0),[Link](https://www.nature.com/articles/s41597-024-04042-0)Cited by:[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[8\]C\. Y\. Chen and C\. M\. Hafner\(2019\)Sentiment\-induced bubbles in the cryptocurrency market\.Journal of Risk and Financial Management12\(2\),pp\. 53\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[9\]H\. J\. Cheong and M\. A\. Morrison\(2008\)Consumers’ reliance on product information and recommendations found in ugc\.Journal of interactive advertising8\(2\),pp\. 38–49\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p2.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[10\]D\. Chicco, M\. J\. Warrens, and G\. Jurman\(2021\)The coefficient of determination r\-squared is more informative than smape, mae, mape, mse and rmse in regression analysis evaluation\.Peerj computer science7,pp\. e623\.Cited by:[§III\-B2](https://arxiv.org/html/2605.20192#S3.SS2.SSS2.p1.3)\.
- \[11\]Decentraland Foundation\(2024\)Decentraland white paper 2\.0\.Technical reportDecentraland Foundation\.Note:Accessed: 2024External Links:[Link](https://decentraland.org/whitepaper2.pdf)Cited by:[§II\-A1](https://arxiv.org/html/2605.20192#S2.SS1.SSS1.p1.5)\.
- \[12\]E\. Den Yeoh, T\. Chung, and Y\. Wang\(2023\)Metaverse in investment using sentiment analysis and machine learning\.InStrategies and Opportunities for Technology in the Metaverse World,pp\. 78–113\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p2.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[13\]X\. Deng, V\. Bashlovkina, F\. Han, S\. Baumgartner, and M\. Bendersky\(2023\)What do llms know about financial markets? a case study on reddit market sentiment analysis\.InCompanion Proceedings of the ACM Web Conference 2023,pp\. 107–110\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p4.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[14\]M\. D\. Devika, C\. Sunitha, and A\. Ganesh\(2016\)Sentiment analysis: a comparative study on different approaches\.Procedia Computer Science87,pp\. 44–49\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[15\]J\. Devlin, M\. Chang, K\. Lee, and K\. Toutanova\(2019\)BERT: pre\-training of deep bidirectional transformers for language understanding\.InProceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 \(Long and Short Papers\),pp\. 4171–4186\.External Links:[Link](https://aclanthology.org/N19-1423/),[Document](https://dx.doi.org/10.18653/v1/N19-1423)Cited by:[§III\-A1](https://arxiv.org/html/2605.20192#S3.SS1.SSS1.p1.1)\.
- \[16\]H\. Duan, J\. Li, S\. Fan, Z\. Lin, X\. Wu, and W\. Cai\(2021\)Metaverse for social good: a university campus prototype\.InProceedings of the 29th ACM international conference on multimedia,pp\. 153–161\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p2.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[17\]A\. ElBahrawy, L\. Alessandretti, A\. Kandler, R\. Pastor\-Satorras, and A\. Baronchelli\(2017\)Evolutionary dynamics of the cryptocurrency market\.Royal Society open science4\(11\),pp\. 170623\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p5.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[18\]D\. Foundation\(2023\)Decentraland’s weekly newsletter \- jan 4\.External Links:[Link](https://decentraland.beehiiv.com/p/weekly-newsletter-january-4)Cited by:[§IV](https://arxiv.org/html/2605.20192#S4.p1.5)\.
- \[19\]L\. Gibbons\(2024\-03\)Decentraland art week 2024 kicks off\.External Links:[Link](https://www.blockleaders.io/news/decentraland-art-week-2024-kicks-off)Cited by:[§IV](https://arxiv.org/html/2605.20192#S4.p1.5)\.
- \[20\]C\. Goanta\(2020\)Selling land in decentraland: the regime of non\-fungible tokens on the ethereum blockchain under the digital content directive\.Disruptive technology, legal innovation, and the future of real estate,pp\. 139–154\.Cited by:[§I](https://arxiv.org/html/2605.20192#S1.p1.1)\.
- \[21\]P\. Gonçalves, M\. Araújo, F\. Benevenuto, and M\. Cha\(2013\)Comparing and combining sentiment analysis methods\.InProceedings of the first ACM conference on Online social networks,pp\. 27–38\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[22\]B\. Guidi and A\. Michienzi\(2022\)Social games and blockchain: exploring the metaverse of decentraland\.In2022 IEEE 42nd International Conference on Distributed Computing Systems Workshops \(ICDCSW\),pp\. 199–204\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p2.1),[§I](https://arxiv.org/html/2605.20192#S1.p2.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[23\]J\. D\. Hamilton\(2020\)Time series analysis\.Princeton university press\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p5.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[24\]H\. T\. Heinonen, A\. Semenov, and V\. Boginski\(2020\)Collective behavior of price changes of erc\-20 tokens\.InComputational Data and Social Networks: 9th International Conference, CSoNet 2020, Dallas, TX, USA, December 11–13, 2020, Proceedings 9,pp\. 487–498\.Cited by:[§II\-B2](https://arxiv.org/html/2605.20192#S2.SS2.SSS2.p1.1)\.
- \[25\]B\. M\. Henrique, V\. A\. Sobreiro, and H\. Kimura\(2019\)Literature review: machine learning techniques applied to financial market prediction\.Expert Systems with Applications124,pp\. 226–251\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p5.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[26\]Y\. Hu, Z\. Li, J\. Chen, Y\. Gao, and Y\. Liu\(2025\)Exploring the decentraland economy: multifaceted parcel attributes, key insights, and benchmarking\.arXiv preprint arXiv:2404\.07533\.External Links:[Link](https://arxiv.org/abs/2404.07533)Cited by:[§II\-A1](https://arxiv.org/html/2605.20192#S2.SS1.SSS1.p1.5)\.
- \[27\]H\. Huang, O\. Zheng, D\. Wang, J\. Yin, Z\. Wang, S\. Ding, H\. Yin, C\. Xu, R\. Yang, Q\. Zheng,et al\.\(2023\)ChatGPT for shaping the future of dentistry: the potential of multi\-modal large language model\.International Journal of Oral Science15\(1\),pp\. 29\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p4.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[28\]H\. Huang, Z\. Yin, Q\. Yang, T\. Li, X\. Luo, L\. Zhou, and Z\. Zheng\(2024\)Scalability and security of blockchain\-empowered metaverse: a survey\.IEEE Open Journal of the Computer Society\.Cited by:[§I](https://arxiv.org/html/2605.20192#S1.p1.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[29\]D\. M\. E\. M\. Hussein\(2018\)A survey on sentiment analysis challenges\.Journal of King Saud University\-Engineering Sciences30\(4\),pp\. 330–338\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[30\]E\. Kasneci, K\. Seßler, S\. Küchemann, M\. Bannert, D\. Dementieva, F\. Fischer, U\. Gasser, G\. Groh, S\. Günnemann, E\. Hüllermeier,et al\.\(2023\)ChatGPT for good? on opportunities and challenges of large language models for education\.Learning and individual differences103,pp\. 102274\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p4.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[31\]O\. Köksoy\(2006\)Multiresponse robust design: mean square error \(mse\) criterion\.Applied Mathematics and Computation175\(2\),pp\. 1716–1729\.Cited by:[§III\-B2](https://arxiv.org/html/2605.20192#S3.SS2.SSS2.p1.3)\.
- \[32\]H\. D\. Le, V\. T\. Truong, and L\. B\. Le\(2024\)Blockchain\-empowered metaverse: decentralized crowdsourcing and marketplace for trading machine learning data and models\.IEEE Access\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p2.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[33\]T\. Lee\(2024\)A test method for the convergence of the metaverse and blockchain\.In2024 26th International Conference on Advanced Communications Technology \(ICACT\),pp\. 321–326\.Cited by:[§IV](https://arxiv.org/html/2605.20192#S4.p2.11)\.
- \[34\]H\. Li\(2018\)Deep learning for natural language processing: advantages and challenges\.National Science Review5\(1\),pp\. 24–26\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p4.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[35\]Q\. Li, J\. Tan, J\. Wang, and H\. Chen\(2020\)A multimodal event\-driven lstm model for stock prediction using online news\.IEEE Transactions on Knowledge and Data Engineering33\(10\),pp\. 3323–3337\.Cited by:[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[36\]Y\. Liu, M\. Ott, N\. Goyal, J\. Du, M\. Joshi, D\. Chen, O\. Levy, M\. Lewis, L\. Zettlemoyer, and V\. Stoyanov\(2019\)Roberta: a robustly optimized bert pretraining approach\.arXiv preprint arXiv:1907\.11692\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p4.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[37\]Z\. Liu, K\. Yang, Q\. Xie, T\. Zhang, and S\. Ananiadou\(2024\)EmoLLMs: a series of emotional large language models and annotation tools for comprehensive affective analysis\.InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining,KDD ’24,New York, NY, USA,pp\. 5487–5496\.External Links:ISBN 9798400704901,[Link](https://doi.org/10.1145/3637528.3671552),[Document](https://dx.doi.org/10.1145/3637528.3671552)Cited by:[§I](https://arxiv.org/html/2605.20192#S1.p2.1)\.
- \[38\]M\. Masala, S\. Ruseti, and M\. Dascalu\(2020\)Robert–a romanian bert model\.InProceedings of the 28th International Conference on Computational Linguistics,pp\. 6626–6637\.Cited by:[§III\-A1](https://arxiv.org/html/2605.20192#S3.SS1.SSS1.p1.1)\.
- \[39\]W\. Medhat, A\. Hassan, and H\. Korashy\(2014\)Sentiment analysis algorithms and applications: a survey\.Ain Shams engineering journal5\(4\),pp\. 1093–1113\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[40\]C\. A\. Melton, O\. A\. Olusanya, N\. Ammar, and A\. Shaban\-Nejad\(2021\)Public sentiment analysis and topic modeling regarding covid\-19 vaccines on the reddit social media platform: a call to action for strengthening vaccine confidence\.Journal of Infection and Public Health14\(10\),pp\. 1505–1512\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[41\]A\. Micu, A\. E\. Micu, M\. Geru, and R\. C\. Lixandroiu\(2017\)Analyzing user sentiment in social media: implications for online marketing strategy\.Psychology & Marketing34\(12\),pp\. 1094–1100\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p2.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[42\]J\. J\. Murphy\(1999\)Technical analysis of the financial markets: a comprehensive guide to trading methods and applications\.Penguin\.Cited by:[§II\-B1](https://arxiv.org/html/2605.20192#S2.SS2.SSS1.p1.1)\.
- \[43\]F\. Neri, C\. Aliprandi, F\. Capeci, and M\. Cuadros\(2012\)Sentiment analysis on social media\.In2012 IEEE/ACM international conference on advances in social networks analysis and mining,pp\. 919–926\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[44\]E\. Ordano, A\. Meilich, Y\. Jardi, and M\. Araoz\(2017\)Decentraland: a blockchain\-based virtual world\.Technical reportDecentraland\.External Links:[Link](https://decentraland.org/whitepaper.pdf)Cited by:[§II\-A1](https://arxiv.org/html/2605.20192#S2.SS1.SSS1.p1.5)\.
- \[45\]Y\. Quan, X\. Wu, W\. Deng, and L\. Zhang\(2024\)Decoding social sentiment in dao: a comparative analysis of blockchain governance communities\.In2024 IEEE 24th International Conference on Software Quality, Reliability, and Security Companion \(QRS\-C\),pp\. 216–224\.External Links:[Document](https://dx.doi.org/10.1109/QRS-C63300.2024.00037),[Link](https://doi.org/10.1109/QRS-C63300.2024.00037)Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[46\]A\. Sherstinsky\(2020\)Fundamentals of recurrent neural network \(rnn\) and long short\-term memory \(lstm\) network\.Physica D: Nonlinear Phenomena404,pp\. 132306\.Cited by:[§III\-B1](https://arxiv.org/html/2605.20192#S3.SS2.SSS1.p1.1)\.
- \[47\]S\. Siami\-Namini and A\. S\. Namin\(2018\)Forecasting economics and financial time series: arima vs\. lstm\.arXiv preprint arXiv:1803\.06386\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p5.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[48\]R\. C\. Staudemeyer and E\. R\. Morris\(2019\)Understanding lstm–a tutorial into long short\-term memory recurrent neural networks\.arXiv preprint arXiv:1909\.09586\.Cited by:[§III\-B1](https://arxiv.org/html/2605.20192#S3.SS2.SSS1.p1.1)\.
- \[49\]P\. Törnberg, D\. Valeeva, J\. Uitermark, and C\. Bail\(2023\)Simulating social media using large language models to evaluate alternative news feed algorithms\.arXiv preprint arXiv:2310\.05984\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p4.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[50\]V\. T\. Truong, L\. Le, and D\. Niyato\(2023\)Blockchain meets metaverse and digital asset management: a comprehensive survey\.Ieee Access11,pp\. 26258–26288\.Cited by:[§I](https://arxiv.org/html/2605.20192#S1.p1.1)\.
- \[51\]M\. Wankhade, A\. C\. S\. Rao, and C\. Kulkarni\(2022\)A survey on sentiment analysis methods, applications, and challenges\.Artificial Intelligence Review55\(7\),pp\. 5731–5780\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[52\]C\. J\. Willmott and K\. Matsuura\(2005\)Advantages of the mean absolute error \(mae\) over the root mean square error \(rmse\) in assessing average model performance\.Climate research30\(1\),pp\. 79–82\.Cited by:[§III\-B2](https://arxiv.org/html/2605.20192#S3.SS2.SSS2.p1.3)\.
- \[53\]B\. Wu and B\. Wu\(2022\)Metaverse: the world reimagined\.InBlockchain for teens: With case studies and examples of blockchain across various industries,pp\. 267–313\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p2.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[54\]F\. Xing\(2024\-08\)Designing heterogeneous llm agents for financial sentiment analysis\.ACM Trans\. Manage\. Inf\. Syst\.\.Note:Just AcceptedExternal Links:ISSN 2158\-656X,[Link](https://doi.org/10.1145/3688399),[Document](https://dx.doi.org/10.1145/3688399)Cited by:[§I](https://arxiv.org/html/2605.20192#S1.p2.1)\.
- \[55\]H\. Xu, B\. Liu, L\. Shu, and P\. S\. Yu\(2019\)BERT post\-training for review reading comprehension and aspect\-based sentiment analysis\.InProceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 \(Long and Short Papers\),pp\. 2324–2335\.External Links:[Link](https://aclanthology.org/N19-1242/),[Document](https://dx.doi.org/10.18653/v1/N19-1242)Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p4.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[56\]T\. Yan, S\. Li, B\. Kramer, L\. Zhang, and C\. J\. Tessone\(2025\)A data engineering framework for ethereum beacon chain rewards: from data collection to decentralization metrics\.Scientific Data12\(1\),pp\. 519\.External Links:[Document](https://dx.doi.org/10.1038/s41597-025-04623-7),[Link](https://www.nature.com/articles/s41597-025-04623-7)Cited by:[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[57\]S\. Y\. Yang, S\. Y\. K\. Mo, and A\. Liu\(2015\)Twitter financial community sentiment and its predictive relationship to stock market movement\.Quantitative Finance15\(10\),pp\. 1637–1656\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p2.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[58\]Y\. Yu, X\. Si, C\. Hu, and J\. Zhang\(2019\)A review of recurrent neural networks: lstm cells and network architectures\.Neural computation31\(7\),pp\. 1235–1270\.Cited by:[§III\-B1](https://arxiv.org/html/2605.20192#S3.SS2.SSS1.p1.1)\.
- \[59\]L\. Yue, W\. Chen, X\. Li, W\. Zuo, and M\. Yin\(2019\)A survey of sentiment analysis in social media\.Knowledge and Information Systems60,pp\. 617–663\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[60\]T\. Zhan, C\. Shi, Y\. Shi, H\. Li, and Y\. Lin\(2024\)Optimization techniques for sentiment analysis based on llm \(gpt\-3\)\.arXiv preprint arXiv:2405\.09770\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p4.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[61\]B\. Zhang, H\. Yang, T\. Zhou, M\. Ali Babar, and X\. Liu\(2023\)Enhancing financial sentiment analysis via retrieval augmented large language models\.InProceedings of the fourth ACM international conference on AI in finance,pp\. 349–356\.Cited by:[§III\-A2](https://arxiv.org/html/2605.20192#S3.SS1.SSS2.p1.9)\.
- \[62\]L\. Zhang, Y\. Quan, J\. Cao, K\. Z\. Zhou, and X\. Tong\(2024\)Leveraging social media sentiments and ethical signals for NFT valuation\.In2024 IEEE 24th International Conference on Software Quality, Reliability, and Security Companion \(QRS\-C\),pp\. 206–215\.External Links:[Document](https://dx.doi.org/10.1109/QRS-C63300.2024.00036),[Link](https://ieeexplore.ieee.org/document/10727085)Cited by:[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.
- \[63\]D\. Zimbra, A\. Abbasi, D\. Zeng, and H\. Chen\(2018\)The state\-of\-the\-art in twitter sentiment analysis: a review and benchmark evaluation\.ACM Transactions on Management Information Systems \(TMIS\)9\(2\),pp\. 1–29\.Cited by:[Appendix A](https://arxiv.org/html/2605.20192#A1.p3.1),[§V](https://arxiv.org/html/2605.20192#S5.p1.1)\.

## Appendix ADetailed Contributions to Related Work

This study advances four interconnected research streams\.

Metaverse and Decentraland Research\.While prior work has established the technological foundations of the Metaverse\[[16](https://arxiv.org/html/2605.20192#bib.bib47)\]and Decentraland’s specific implementation\[[22](https://arxiv.org/html/2605.20192#bib.bib10),[12](https://arxiv.org/html/2605.20192#bib.bib30)\], this study is among the first to empirically link platform\-specific community sentiment to token price dynamics \(see Table[III](https://arxiv.org/html/2605.20192#A2.T3)for performance metrics\)\. We demonstrate that DAO\-governed virtual worlds\[[53](https://arxiv.org/html/2605.20192#bib.bib48)\]generate user\-generated content\[[9](https://arxiv.org/html/2605.20192#bib.bib49),[3](https://arxiv.org/html/2605.20192#bib.bib50)\]with measurable but bounded predictive signal for MANA returns, extending research on blockchain\-empowered metaverse economies\[[32](https://arxiv.org/html/2605.20192#bib.bib83)\]and the importance of monitoring user sentiment for platform governance\[[41](https://arxiv.org/html/2605.20192#bib.bib51),[57](https://arxiv.org/html/2605.20192#bib.bib52)\]\. The daily sentiment trajectory shown in Figure[3](https://arxiv.org/html/2605.20192#A2.F3)illustrates the temporal dynamics of community engagement\.

Sentiment Analysis Methodology\.Our work bridges lexicon\-based and machine\-learning approaches\[[14](https://arxiv.org/html/2605.20192#bib.bib56),[21](https://arxiv.org/html/2605.20192#bib.bib57),[39](https://arxiv.org/html/2605.20192#bib.bib31)\]with modern LLM applications\. By validating that general\-domain sentiment surveys\[[51](https://arxiv.org/html/2605.20192#bib.bib54),[29](https://arxiv.org/html/2605.20192#bib.bib55),[59](https://arxiv.org/html/2605.20192#bib.bib60),[43](https://arxiv.org/html/2605.20192#bib.bib61)\]require domain adaptation for metaverse contexts, we contribute to methodological refinements in social media sentiment analysis\[[63](https://arxiv.org/html/2605.20192#bib.bib62),[40](https://arxiv.org/html/2605.20192#bib.bib63),[1](https://arxiv.org/html/2605.20192#bib.bib64)\]and crypto\-specific opinion mining\[[8](https://arxiv.org/html/2605.20192#bib.bib65),[45](https://arxiv.org/html/2605.20192#bib.bib67)\]\. The textual data structure summarized in Table[II](https://arxiv.org/html/2605.20192#A2.T2)enables these analyses\.

LLM Applications in Financial Forecasting\.We advance the deployment of large language models\[[30](https://arxiv.org/html/2605.20192#bib.bib37),[6](https://arxiv.org/html/2605.20192#bib.bib36),[27](https://arxiv.org/html/2605.20192#bib.bib38),[34](https://arxiv.org/html/2605.20192#bib.bib39)\]for specialized financial sentiment tasks\. Utilizing architectures like BERT and RoBERTa\[[36](https://arxiv.org/html/2605.20192#bib.bib33),[55](https://arxiv.org/html/2605.20192#bib.bib58),[60](https://arxiv.org/html/2605.20192#bib.bib59)\], we test their efficacy on Discord data, contributing to research on LLM\-based social media simulation\[[49](https://arxiv.org/html/2605.20192#bib.bib34)\]and Reddit\-style market sentiment analysis\[[13](https://arxiv.org/html/2605.20192#bib.bib35)\]within decentralized virtual communities\.

Cryptocurrency Prediction and Volatility\.Addressing the evolutionary dynamics of crypto markets\[[17](https://arxiv.org/html/2605.20192#bib.bib68)\], our LSTM\-based multi\-modal framework responds to calls for advanced prediction methodologies\[[25](https://arxiv.org/html/2605.20192#bib.bib69)\]\. We contribute to time series forecasting literature\[[23](https://arxiv.org/html/2605.20192#bib.bib70),[47](https://arxiv.org/html/2605.20192#bib.bib72)\]by quantifying the marginal value of sentiment features beyond technical indicators\[[2](https://arxiv.org/html/2605.20192#bib.bib71)\], demonstrating both the potential and limitations of community\-specific data for crypto\-volatility modeling\.

## Appendix BData Specifications and Empirical Results

We detail the dataset structures and empirical benchmarks supporting the multi\-modal analysis\.

### B\-AMarket Data Structure

Table[I](https://arxiv.org/html/2605.20192#A2.T1)defines the quantitative financial variables extracted from Decentraland’s MANA token markets\. These time\-series features constitute the baseline price\-only model inputs\.

TABLE I:Numerical Data: MANA Price, Volume, Market CapNote:All price and market cap values are denominated in US Dollars\.

### B\-BSentiment Corpus Structure

Table[II](https://arxiv.org/html/2605.20192#A2.T2)outlines the Discord message schema\. The sentiment labels and scores are derived via LLM\-based classification of the Content field\.

TABLE II:Textual Data: Users’ Messages \(Discord Corpus\)Note:Sentiment labels and scores derived via LLM\-based classification\.

### B\-CModel Performance Benchmarks

Table[III](https://arxiv.org/html/2605.20192#A2.T3)presents the eight\-fold cross\-validation results comparing price\-only LSTM baselines against multi\-modal variants \(Price \+ Sentiment\)\. The average improvements indicate that sentiment features provide marginal but consistent predictive enhancement, though negativeR2R^\{2\}values in both configurations suggest substantial unmodeled volatility drivers\.

TABLE III:Comparison of Prediction Models \(8 Independent Runs\)Note:MSE scaled by10410^\{4\}for readability\. Legend:↓\\downarrow=Lower error/higher R2\(better\);↑\\uparrow=Worse\.

### B\-DTemporal Visualization

Figure[3](https://arxiv.org/html/2605.20192#A2.F3)displays the daily averaged sentiment scores derived from the Discord corpus \(Table[II](https://arxiv.org/html/2605.20192#A2.T2)\), revealing temporal fluctuations in community mood\. Figure[4](https://arxiv.org/html/2605.20192#A2.F4)overlays predicted returns from the multi\-modal model against realized returns, visualizing the systematic underestimation of volatility discussed\.

![Refer to caption](https://arxiv.org/html/2605.20192v1/fig/sa2.png)

Figure 3:Daily Average Sentiment Score \(Decentraland Discord Community\)\.![Refer to caption](https://arxiv.org/html/2605.20192v1/fig/n0.png)

Figure 4:Daily Token Return\.