Ningyuan Deng
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ningyuan Deng has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Bo Li
Tsinghua University, Beijing, China and Baidu Inc, Beijing, China
1 shared publication - Huacan Wang
Midea AIRC, Shanghai, China
1 shared publication - Jinyuan Wang1 shared publication
- Lijie Wen
Tsinghua University
1 shared publication - Qi Li1 shared publication
- Ronghao Chen
Peking University, Beijing, China
1 shared publication
Latest Publications
- Open Access
- 15 August 2026
- Published by Elsevier in SSRN Electronic Journal
MacroAllocAgent: From macro narratives to strategic asset allocation via a multi-agent LLM system
We propose MacroAllocAgent, a multi-agent Large Language Model (LLM) system designed to translate macro-financial narratives into strategic asset allocation weights for long-horizon investors. The system processes Chinese official macroeconomic communications and market outlook reports through five specialist agents. Each agent maps the evidence into Pring-cycle probabilities and asset-level return views. The views are aggregated, assigned probability-derived uncertainty, and enter into a Black-Litterman optimizer that produces allocation weights. Using 13 Chinese macroeconomic-exposure indices from 2015-2025, MacroAllocAgent achieves Sharpe ratios of 0.955 and 0.758 across unconstrained and bounded portfolios, respectively, representing 6.0% and 11.0% improvements over the best baseline.Bootstrap-based tests support the statistical significance of the improvements, while ablation studies and backbone comparisons demonstrate the robustness and generalizability of MacroAllocAgent and reveal the coordinated contribution of its components.
- 15 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMacroAllocAgent: From macro narratives to strategic asset allocation via a multi-agent LLM system
We propose MacroAllocAgent, a multi-agent Large Language Model (LLM) system designed to translate macro-financial narratives into strategic asset allocation weights for long-horizon investors. The system processes Chinese official macroeconomic communications and market outlook reports through five specialist agents. Each agent maps the evidence into Pring-cycle probabilities and asset-level return views. The views are aggregated, assigned probability-derived uncertainty, and enter into a Black-Litterman optimizer that produces allocation weights. Using 13 Chinese macroeconomic-exposure indices from 2015-2025, MacroAllocAgent achieves Sharpe ratios of 0.955 and 0.758 across unconstrained and bounded portfolios, respectively, representing 6.0% and 11.0% improvements over the best baseline.Bootstrap-based tests support the statistical significance of the improvements, while ablation studies and backbone comparisons demonstrate the robustness and generalizability of MacroAllocAgent and reveal the coordinated contribution of its components.
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 2533-2544
VaaWIT: Visual-Aware Adaptation of Large Language Models for Multilingual Web Image Translation
Translating text embedded in Web images is crucial for improving content accessibility and cross-lingual information retrieval, particularly within social media and e-commerce domains. Although Large Vision-Language Models (LVLMs) have advanced multimodal understanding, applying them to Web image translation remains challenging due to the visual representation gap: standard encoders often prioritize high-level semantics over the fine-grained visual details required for recognizing diverse character morphologies. To address this challenge, we propose VaaWIT, an end-to-end framework that adapts Large Language Models for multilingual Web image translation. The framework introduces two key technical contributions: (1) a Dual-Stream Attention Module (DSAM), which facilitates bidirectional interaction between multilingual semantic features and detailed visual representations, thereby synthesizing unified features robust to textual variations; and (2) a Visual-Aware Adapter (VAA), a parameter-efficient fine-tuning strategy that dynamically injects these fused visual cues into the frozen LLM backbone. This design enables the model to align the visual context with linguistic reasoning effectively while minimizing computational costs. Extensive experiments on eight tasks on three public benchmarks demonstrate that VaaWIT significantly outperforms state-of-the-art (SOTA) open-source baselines and achieves competitive performance against proprietary models. These results validate the efficacy of integrating fine-grained visual perception into LLMs for complex Web content analysis.
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 2533-2544
Open AccessVaaWIT: Visual-Aware Adaptation of Large Language Models for Multilingual Web Image Translation
Translating text embedded in Web images is crucial for improving content accessibility and cross-lingual information retrieval, particularly within social media and e-commerce domains. Although Large Vision-Language Models (LVLMs) have advanced multimodal understanding, applying them to Web image translation remains challenging due to the visual representation gap: standard encoders often prioritize high-level semantics over the fine-grained visual details required for recognizing diverse character morphologies. To address this challenge, we propose VaaWIT, an end-to-end framework that adapts Large Language Models for multilingual Web image translation. The framework introduces two key technical contributions: (1) a Dual-Stream Attention Module (DSAM), which facilitates bidirectional interaction between multilingual semantic features and detailed visual representations, thereby synthesizing unified features robust to textual variations; and (2) a Visual-Aware Adapter (VAA), a parameter-efficient fine-tuning strategy that dynamically injects these fused visual cues into the frozen LLM backbone. This design enables the model to align the visual context with linguistic reasoning effectively while minimizing computational costs. Extensive experiments on eight tasks on three public benchmarks demonstrate that VaaWIT significantly outperforms state-of-the-art (SOTA) open-source baselines and achieves competitive performance against proprietary models. These results validate the efficacy of integrating fine-grained visual perception into LLMs for complex Web content analysis.