SMP 2026｜大会倒计时 6 天！Poster 录用名单公布

一、会议简介

全国社会媒体处理大会（China National Conference on Social Media Processing）由中国中文信息学会社会媒体处理专业委员会主办，专注于以社会媒体处理为主题的科学研究与工程开发，为传播社会媒体处理最新学术研究与技术成果提供广泛的交流平台，旨在构建社会媒体处理领域的产学研生态圈，成为中国乃至世界社会媒体处理的风向标。全国社会媒体处理大会创办于 2012 年，每年举办一次，现已成为社会媒体处理领域的重要学术活动。

第十四届全国社会媒体处理大会（SMP 2026）将于 2026 年 11 月 12 ~ 15 日在北京召开，由中国传媒大学媒体融合与传播国家重点实验室承办。大会主题是“大模型驱动人机共生”，旨在探索大模型日益普及的今天，大模型与人类社会深度共生的途径，聚焦大模型和智能体技术如何赋能人机协同的价值释放。

二、会议安排

会议时间：2026 年 11 月 12 日-15 日

2026 年 11 月 12 日-13 日：前沿技术讲习班
2026 年 11 月 14 日-15 日：主会议（特邀报告、分论坛等）

会议地点：北京辽宁大厦

会议官网：http://smp2026.cips-smp.org/

三、Poster 录用名单

	电子海报
1	题目：WALKSAFE: Risk-aware Graph Random Walk with Bi-GRPO for LLM Safety 讲者单位：国防科技大学讲者：潘世龙

	纸质海报
1	题目：Surfacing the Unsaid: CUE-Bench for Affective Stance in Chinese Discourse 讲者单位：华中科技大学讲者：郑贞言
2	题目：Preserving General Capability in Personality-Aligned LLMs via Orthogonal Gradient Projection 讲者单位：首都师范大学讲者：魏琳悦
3	题目：Deep-EMO: Dual-Graph Event-to-Cognition Evolutionary Modelingand Cognitive-Traction Optimization for LLM-based Empathetic Response Generation（IEEE Transactions on Affective Computing 投稿中）讲者单位：南京信息工程大学讲者：段彦俭
4	题目：Logical Phase Transitions: Understanding Collapse in LLM Logical Reasoning 讲者单位：华中科技大学讲者：张云耀
5	题目：Semantic-Aware Logical Reasoning via a Semiotic Framework 讲者单位：华中科技大学讲者：张云耀
6	题目：MePe: Rethinking Multimodal Chinese Idiom Reading Comprehensionfrom a Metaphorical Perspective 讲者单位：华中农业大学讲者：王同官
7	题目：MemeScope: Probing and Causally Localizing Pragmatic Inferencein VLM Meme Understanding 讲者单位：华中农业大学讲者：王同官
8	题目：面向分布式推荐场景的个性化模型聚合技术研究讲者单位：北京交通大学讲者：张洪磊
9	题目：StabCF: A Stabilized Training Method for Collaborative Filtering 讲者单位：燕山大学讲者：刘思奇
10	题目：Beyond Words: Enhancing Desire, Emotion, and Sentiment Recognition with Non-Verbal Cues 讲者单位：华中农业大学讲者：陈维
11	题目：Semantic-Aligned Structural Abstraction for Multimodal Sentiment Analysis 讲者单位：华中农业大学讲者：陈维
12	题目：Euphemism Identification via Feature Fusion and Individualization 讲者单位：湖北工业大学讲者：胡玉雪
13	题目：Keyword-Oriented Multimodal Modeling for Euphemism Identification 讲者单位：湖北工业大学讲者：胡玉雪
14	题目：Strong yet Controllable LLM Abstention without Learning to Abstain 讲者单位：中国科学院大学讲者：张钧玮
15	题目：TrustDABench: Benchmarking Reliability and Robustness of LLMs for Structured Data Analysis 讲者单位：中国移动数智事业部九天研究院讲者：史铂深
16	题目：FairTCD: Dual-Teacher Temporal Contrastive Distillation for Twofold Fair Dynamic Graph Embedding 讲者单位：南京航空航天大学讲者：李逸聪
17	题目：MIDSim: Simulating Multi-Channel Information Diffusion in Social Media with LLM-Powered Multi-Agent System 讲者单位：中国科学院计算技术研究所讲者：刘乐熙
18	题目：HieRuProL: Hierarchical Rumor Propagation Control by Coupling Adversarial Team-based MARL with Local-to-Global Network Dynamics 讲者单位：南京信息工程大学讲者：孙杰
19	题目：Debiased Multimodal Personality Understandingthrough Dual Causal Intervention 讲者单位：首都师范大学讲者：冯航
20	题目：Diff-VLN: Brain-Inspired Efficient Vision-Language Navigation by Differentiable Arbitration on Goal-Directed and Habitual Policy 讲者单位：南京信息工程大学讲者：李知非
21	题目：CompTab: A Comprehensive Benchmark for Real-World TableQA with Complex Reasoning and Irregular Tables 讲者单位：安徽大学讲者：杨震
22	题目：SIP: Aligning LLM-based TableQA with Table Semantics via SQL Logic 讲者单位：安徽大学讲者：杨震
23	题目：When Can We Trust Fairness Audits? Identifying Reliability Boundaries of Third-party Audit Conclusions 讲者单位：中国科学院计算技术研究所讲者：刘元浩
24	题目：MemCast: Memory-Driven Time Series Forecasting withExperience-Conditioned Reasoning 讲者单位：中国科学技术大学讲者：陶小玉
25	题目：A Graph Foundation Model for Unified Clustering 讲者单位：天津大学讲者：韩仁达
26	题目：PTCG-Bench: Can LLM Agents Master Pokémon Trading Card Game? 讲者单位：浙江大学讲者：华东东
27	题目：ScaleCUA: Scaling Computer Use Agents with Verifiable Task Synthesis and Efficient Online RL 讲者单位：清华大学讲者：吕博文
28	题目：MemBubble: Benchmarking How Agent Memory Shapes LLM Response Diversity 讲者单位：中国科学院计算技术研究所讲者：谭家俊
29	题目：MathSight: A Benchmark Exploring Have Vision-Language Models Really Seen in University-Level Mathematical Reasoning? 讲者单位：首都师范大学信息工程学院讲者：崔耀
30	题目：PaperScout: An Autonomous Agent for Academic Paper Search with Process-Aware Sequence-Level Policy Optimization (EMNLP2026) 讲者单位：中国科学技术大学讲者：程明月
31	题目：EVA: A Memory-Augmented Harness For Long-Horizon Embodied Tasks 讲者单位：浙江大学讲者：丁思韬
32	题目：Opinion Convergence and Bimodal Polarization in Networks of LLM Agents 讲者单位：北京师范大学新闻传播学院讲者：彭晨曦
33	题目：A Unified Framework for Context-Aware and Relation-Aware Graph Retrieval-Augmented Generation 讲者单位：浙江大学讲者：钟昊阳
34	题目：Self-Correcting Agentic RAG via Memory-Grounded Failure Localization 讲者单位：浙江大学讲者：钟昊阳
35	题目：Token-level Adaptive Weighting via Implicit Rewards for Inference- time Multi-objective Alignment (EMNLP 2026 已发表) 讲者单位：中国传媒大学讲者：胡斗
36	题目：On-Policy Representation Alignment in Preference Optimization 讲者单位：中国传媒大学讲者：胡斗
37	题目：Paper Influence Prediction for Newborns Based on Dual Intelligent Agents 讲者单位：中国人民大学信息学院讲者：彭子坚
38	题目：Seeing Further and Wider: Joint Spatio-Temporal Enlargement for Micro-Video Popularity Prediction 讲者单位：华中科技大学讲者：王达立
39	题目：DyBenchEval: A Comprehensive Evaluation Suite for Measuring LLM-as-Dynamic-Benchmarker 讲者单位：电子科技大学讲者：王璟烨
40	题目：Predicting multi‑technology convergence based on higher‑order interactions in temporal hypergraphs 讲者单位：对外经济贸易大学讲者：王瑞庭
41	题目：ACWM-RM: A Reference-Free Reward Model from Complementary Mismatch Preferences 讲者单位：复旦大学讲者：洪诗博
42	题目：Ex-ante Identification of Disruptive Technologies from Patent Text via LLM-based Semantic Evidence Aggregation 讲者单位：对外经济贸易大学讲者：丁方园

注：1、大会将设立“亮点海报” 8 项，由与会嘉宾与学者共同投票选出，获奖者会获得证书奖励。

2、为确保海报展示与现场交流效果，每篇录用海报至少需有1名作者注册参会。

四、报告主题

题目：AI Building AI: Progress Toward Recursive Self-Improvement

摘要：Recent advances in artificial intelligence are transforming not only what AI systems can accomplish, but also how AI itself is developed. As AI becomes increasingly capable of contributing to research and engineering, a compelling possibility emerges: AI systems that can help build increasingly powerful AI systems, potentially paving the way toward recursive self-improvement. In this talk, I will review the evolution of AI-driven AI research, from early approaches such as self-play and automated optimization to recent advances in AI research agents. I will then highlight recent progress at Recursive Superintelligence, showcasing our publicly released research results and our AI model, Flint, to illustrate how AI-driven research can accelerate model development and push the boundaries of AI capabilities and efficiency. Finally, I will discuss the broader vision of AI building AI, its potential to transform the pace of technological progress, and its implications for scientific discovery across disciplines.

五、会议注册

注册网址：https://www.cipsc.org.cn/Learn/index.aspx?itemid=5442

会员查询：https://www.cipsc.org.cn/Memberinquiry/index.aspx

会员注册：https://www.cipsc.org.cn/Register.aspx

注册图-新.jpg

六、住宿酒店推荐

一、北京辽宁大厦（会议酒店）

具体酒店信息请参考往期推文：通知｜SMP 2026 住宿及交通信息

二、周边协议酒店

预订方式：统一电话联系酒店经理，报 SMP 2026 享受大会协议价格预订。

田经理：17718973646

陈经理：15010501105

	纸质海报
1	题目：Surfacing the Unsaid: CUE-Bench for Affective Stance in Chinese Discourse 讲者单位：华中科技大学讲者：郑贞言
2	题目：Preserving General Capability in Personality-Aligned LLMs via Orthogonal Gradient Projection 讲者单位：首都师范大学讲者：魏琳悦
3	题目：Deep-EMO: Dual-Graph Event-to-Cognition Evolutionary Modelingand Cognitive-Traction Optimization for LLM-based Empathetic Response Generation（IEEE Transactions on Affective Computing 投稿中）讲者单位：南京信息工程大学讲者：段彦俭
4	题目：Logical Phase Transitions: Understanding Collapse in LLM Logical Reasoning 讲者单位：华中科技大学讲者：张云耀
5	题目：Semantic-Aware Logical Reasoning via a Semiotic Framework 讲者单位：华中科技大学讲者：张云耀
6	题目：MePe: Rethinking Multimodal Chinese Idiom Reading Comprehensionfrom a Metaphorical Perspective 讲者单位：华中农业大学讲者：王同官
7	题目：MemeScope: Probing and Causally Localizing Pragmatic Inferencein VLM Meme Understanding 讲者单位：华中农业大学讲者：王同官
8	题目：面向分布式推荐场景的个性化模型聚合技术研究讲者单位：北京交通大学讲者：张洪磊
9	题目：StabCF: A Stabilized Training Method for Collaborative Filtering 讲者单位：燕山大学讲者：刘思奇
10	题目：Beyond Words: Enhancing Desire, Emotion, and Sentiment Recognition with Non-Verbal Cues 讲者单位：华中农业大学讲者：陈维
11	题目：Semantic-Aligned Structural Abstraction for Multimodal Sentiment Analysis 讲者单位：华中农业大学讲者：陈维
12	题目：Euphemism Identification via Feature Fusion and Individualization 讲者单位：湖北工业大学讲者：胡玉雪
13	题目：Keyword-Oriented Multimodal Modeling for Euphemism Identification 讲者单位：湖北工业大学讲者：胡玉雪
14	题目：Strong yet Controllable LLM Abstention without Learning to Abstain 讲者单位：中国科学院大学讲者：张钧玮
15	题目：TrustDABench: Benchmarking Reliability and Robustness of LLMs for Structured Data Analysis 讲者单位：中国移动数智事业部九天研究院讲者：史铂深
16	题目：FairTCD: Dual-Teacher Temporal Contrastive Distillation for Twofold Fair Dynamic Graph Embedding 讲者单位：南京航空航天大学讲者：李逸聪
17	题目：MIDSim: Simulating Multi-Channel Information Diffusion in Social Media with LLM-Powered Multi-Agent System 讲者单位：中国科学院计算技术研究所讲者：刘乐熙
18	题目：HieRuProL: Hierarchical Rumor Propagation Control by Coupling Adversarial Team-based MARL with Local-to-Global Network Dynamics 讲者单位：南京信息工程大学讲者：孙杰
19	题目：Debiased Multimodal Personality Understandingthrough Dual Causal Intervention 讲者单位：首都师范大学讲者：冯航
20	题目：Diff-VLN: Brain-Inspired Efficient Vision-Language Navigation by Differentiable Arbitration on Goal-Directed and Habitual Policy 讲者单位：南京信息工程大学讲者：李知非
21	题目：CompTab: A Comprehensive Benchmark for Real-World TableQA with Complex Reasoning and Irregular Tables 讲者单位：安徽大学讲者：杨震
22	题目：SIP: Aligning LLM-based TableQA with Table Semantics via SQL Logic 讲者单位：安徽大学讲者：杨震
23	题目：When Can We Trust Fairness Audits? Identifying Reliability Boundaries of Third-party Audit Conclusions 讲者单位：中国科学院计算技术研究所讲者：刘元浩
24	题目：MemCast: Memory-Driven Time Series Forecasting withExperience-Conditioned Reasoning 讲者单位：中国科学技术大学讲者：陶小玉
25	题目：A Graph Foundation Model for Unified Clustering 讲者单位：天津大学讲者：韩仁达
26	题目：PTCG-Bench: Can LLM Agents Master Pokémon Trading Card Game? 讲者单位：浙江大学讲者：华东东
27	题目：ScaleCUA: Scaling Computer Use Agents with Verifiable Task Synthesis and Efficient Online RL 讲者单位：清华大学讲者：吕博文
28	题目：MemBubble: Benchmarking How Agent Memory Shapes LLM Response Diversity 讲者单位：中国科学院计算技术研究所讲者：谭家俊
29	题目：MathSight: A Benchmark Exploring Have Vision-Language Models Really Seen in University-Level Mathematical Reasoning? 讲者单位：首都师范大学信息工程学院讲者：崔耀
30	题目：PaperScout: An Autonomous Agent for Academic Paper Search with Process-Aware Sequence-Level Policy Optimization (EMNLP2026) 讲者单位：中国科学技术大学讲者：程明月
31	题目：EVA: A Memory-Augmented Harness For Long-Horizon Embodied Tasks 讲者单位：浙江大学讲者：丁思韬
32	题目：Opinion Convergence and Bimodal Polarization in Networks of LLM Agents 讲者单位：北京师范大学新闻传播学院讲者：彭晨曦
33	题目：A Unified Framework for Context-Aware and Relation-Aware Graph Retrieval-Augmented Generation 讲者单位：浙江大学讲者：钟昊阳
34	题目：Self-Correcting Agentic RAG via Memory-Grounded Failure Localization 讲者单位：浙江大学讲者：钟昊阳
35	题目：Token-level Adaptive Weighting via Implicit Rewards for Inference- time Multi-objective Alignment (EMNLP 2026 已发表) 讲者单位：中国传媒大学讲者：胡斗
36	题目：On-Policy Representation Alignment in Preference Optimization 讲者单位：中国传媒大学讲者：胡斗
37	题目：Paper Influence Prediction for Newborns Based on Dual Intelligent Agents 讲者单位：中国人民大学信息学院讲者：彭子坚
38	题目：Seeing Further and Wider: Joint Spatio-Temporal Enlargement for Micro-Video Popularity Prediction 讲者单位：华中科技大学讲者：王达立
39	题目：DyBenchEval: A Comprehensive Evaluation Suite for Measuring LLM-as-Dynamic-Benchmarker 讲者单位：电子科技大学讲者：王璟烨
40	题目：Predicting multi‑technology convergence based on higher‑order interactions in temporal hypergraphs 讲者单位：对外经济贸易大学讲者：王瑞庭
41	题目：ACWM-RM: A Reference-Free Reward Model from Complementary Mismatch Preferences 讲者单位：复旦大学讲者：洪诗博
42	题目：Ex-ante Identification of Disruptive Technologies from Patent Text via LLM-based Semantic Evidence Aggregation 讲者单位：对外经济贸易大学讲者：丁方园