Comparison of Deterministic Information Providers
研究通过使用同时后验匹配、构造性划分标准和共同知识组件，分析了不完全信息博弈中不同确定性信息提供者的表现。
0 citationsRead paper
Selected work
Representative Papers
From Centrality Discounts to Centrality Premia: Interoperability and Platform Competition in Social Networks
This study investigates how interoperability shapes personalized pricing competition between platforms in social networks. The authors develop a duopoly model of differentiated platforms, where consumers derive network externalities from neighbors using the same or interoperable platforms. By integrating game theory, network economics, and the Katz–Bonacich centrality measure from graph theory, they derive closed-form equilibrium prices for arbitrary network structures. Their analysis reveals that whether central users receive discounts or premiums depends on the relative magnitude of interoperability and product substitutability. A critical threshold exists at which equilibrium prices become independent of users’ network positions. Furthermore, interoperability attenuates price competition between platforms, incentivizing them to favor denser user networks and reversing the conventional direction of price discrimination—whereby highly central users, who typically benefit under standard models, may instead face higher prices.
0 citationsRead paper
Why Inference in Large Models Becomes Decomposable After Training
Large model inference relies on dense parameter matrices, leading to computational costs and system complexity that scale unsustainably with model size. This work is the first to reveal, from a post-training, model-agnostic structural perspective, that parameter dependencies in large models exhibit strong locality and selectivity: a substantial portion of parameter structures are statistically indistinguishable from their initialization distribution, indicating an inherent decomposability of inference systems. To exploit this property, the authors propose a structure annealing method based on statistical analysis of gradient update events, along with a significance criterion for parameter dependencies, enabling effective identification and removal of unsupported dependencies and extraction of stable, independent substructures. Without altering model functionality, this approach enables structured parallel inference, establishing a new paradigm for efficient large-model inference.
0 citationsRead paper
Recent publications
Latest Papers
Comparison of Deterministic Information Providers
研究通过使用同时后验匹配、构造性划分标准和共同知识组件，分析了不完全信息博弈中不同确定性信息提供者的表现。
0 citationsRead paper
From Centrality Discounts to Centrality Premia: Interoperability and Platform Competition in Social Networks
This study investigates how interoperability shapes personalized pricing competition between platforms in social networks. The authors develop a duopoly model of differentiated platforms, where consumers derive network externalities from neighbors using the same or interoperable platforms. By integrating game theory, network economics, and the Katz–Bonacich centrality measure from graph theory, they derive closed-form equilibrium prices for arbitrary network structures. Their analysis reveals that whether central users receive discounts or premiums depends on the relative magnitude of interoperability and product substitutability. A critical threshold exists at which equilibrium prices become independent of users’ network positions. Furthermore, interoperability attenuates price competition between platforms, incentivizing them to favor denser user networks and reversing the conventional direction of price discrimination—whereby highly central users, who typically benefit under standard models, may instead face higher prices.
0 citationsRead paper
Why Inference in Large Models Becomes Decomposable After Training
Large model inference relies on dense parameter matrices, leading to computational costs and system complexity that scale unsustainably with model size. This work is the first to reveal, from a post-training, model-agnostic structural perspective, that parameter dependencies in large models exhibit strong locality and selectivity: a substantial portion of parameter structures are statistically indistinguishable from their initialization distribution, indicating an inherent decomposability of inference systems. To exploit this property, the authors propose a structure annealing method based on statistical analysis of gradient update events, along with a significance criterion for parameter dependencies, enabling effective identification and removal of unsupported dependencies and extraction of stable, independent substructures. Without altering model functionality, this approach enables structured parallel inference, establishing a new paradigm for efficient large-model inference.
0 citationsRead paper