AAAI 2025

3,735 papers

Top Research Directions

The 5,540 research directions with AAAI 2025 papers, ranked by paper count, next to 2024. Share is the fraction of the edition’s papers in the direction.

#Direction20252024Share 2025Share 20242024 rank
1Bypassing the Computational Bottleneck of Adversarial Training29140.8%0.5%6
2The Cross Modal Representation Gap27230.7%0.7%1
3Factual Inaccuracy and Hallucination in Large Language Models2580.7%0.3%51
4Efficient Sequence Modeling with Linear State Space Models2300.6%0.0%—
5Mitigating Error Propagation in Pseudo Labeling23120.6%0.4%12
6Spatial Attention Masking in Generative Diffusion Models21100.6%0.3%32
7The Sensitivity of Multi-Objective Loss Balancing21120.6%0.4%15
8Cross-Modal Attention Fusion for Heterogeneous Data Integration2090.5%0.3%35
9Aligning Global Semantics with Local Spatial Precision1990.5%0.3%34
10Clean Accuracy Degradation in Machine Learning Defenses19170.5%0.5%3
11Overcoming the Homophily Assumption in Graph Neural Networks19130.5%0.4%10
12Dynamic Preference Modeling in Interactive Systems18100.5%0.3%25
13The Computational Overhead of Generative Data Synthesis18130.5%0.4%11
14Over-Smoothing in Graph Neural Networks17140.5%0.5%7
15Efficient Training and Inference for Diffusion Models16110.4%0.4%18
16Hyperparameter Sensitivity in Complex Machine Learning Frameworks1630.4%0.1%608
17Overcoming the Iterative Inference Latency of Diffusion Models16160.4%0.5%5
18Dynamic Routing and Expert Specialization in Modular Architectures1420.4%0.1%1124
19Generalizing from Synthetic Data to the Real World14110.4%0.4%19
20Source Free Domain Adaptation1490.4%0.3%41
21The Mechanics of Backdoor Attacks in Neural Networks14170.4%0.5%4
22Compressing Large Foundation Models via Knowledge Distillation13130.3%0.4%8
23Energy Efficient Computing with Spiking Neural Networks13100.3%0.3%26
24Foundational Literacy for the Age of Artificial Intelligence1330.3%0.1%581
25Grounding Language Models with External Knowledge Retrieval1310.3%0.0%2973

Fastest growing

Largest gain in share since 2024, among directions with at least 11 papers in 2025.

Direction20242025Share change
Grounding Language Models with External Knowledge Retrieval113×10.8
Dynamic Routing and Expert Specialization in Modular Architectures214×5.8
Hyperparameter Sensitivity in Complex Machine Learning Frameworks316×4.4
Foundational Literacy for the Age of Artificial Intelligence313×3.6
Exploiting Temporal Redundancy in Video Neural Networks311×3.0
Reducing the Memory and Latency of Language Model Inference311×3.0
Factual Inaccuracy and Hallucination in Large Language Models825×2.6
Conformal Prediction for Rigorous Uncertainty Quantification412×2.5
Preserving Semantic Identity in Graph Contrastive Learning412×2.5
Evaluating Generative AI with Large Language Models411×2.3

New this year

Directions with no 2024 paper and at least 6 in 2025.

Direction2025
Efficient Sequence Modeling with Linear State Space Models23
Preserving Temporal Identity Consistency in Video Generation8
Scaling 3D Vision with Linear State Space Models8
Visual State Space Models for Linear Complexity Global Context8
Sub-Quadratic Sequence Modeling for Long Contexts7
The Mechanics of Qualitative and Quantitative Data Integration7
The Structural Vulnerability of Language Models to Prompt Injection7
The Synthetic-to-Real Gap in Atmospheric Restoration7
Aligning Confidence and Localization Quality in Object Detectors6
Cross Domain Generalization in Recommender Systems6

Lost share

Largest loss in share since 2024, among directions with at least 8 papers in 2024.

Direction20242025Share change
Resolving Mesh Interpenetration in 3D Reconstruction and Generation81×0.1
Denoising Web Scraped Datasets for Vision Language Models92×0.2
Scaling Spectral Graph Neural Networks82×0.2
Mitigating Non-Stationary Distribution Shift in Time-Series Forecasting113×0.2
Causal Inference with Unobserved Confounding103×0.2
Continual Learning Without Data Rehearsal83×0.3
Mitigating Long Range Sparsity in 3D Object Detection83×0.3
Parameter Synthesis for Dynamic Model Adaptation83×0.3
Anomaly Detection via Generative Manifold Projection104×0.3
Reconstructing Private Data from Model Gradients115×0.4

Download all directions as CSV

Institutions

Institutions credited by AAAI 2025 papers: a paper credits every institution named in its affiliations, once. 1,397 institutions share 3,663 credited papers.

#InstitutionPapersShare20242024 rank
1Zhejiang University 1734.7%1322
2Tsinghua University 1534.2%1441
3University of Chinese Academy of Sciences 1484.0%1025
4Peking University 1343.7%1123
5Chinese Academy of Sciences 1293.5%1084
6University of Science and Technology of China 1133.1%868
7Shanghai Jiao Tong University 1093.0%887
8Beihang University 992.7%5914
9National University of Singapore 892.4%799
10Nanyang Technological University 842.3%7510
11Harbin Institute of Technology 832.3%6411
12Sun Yat-sen University 752.0%4822
13Nanjing University 742.0%5317
14Tencent (China) 732.0%886
15Fudan University 711.9%6212
16Chinese University of Hong Kong 661.8%4723
17Hong Kong Polytechnic University 651.8%3838
18Hong Kong University of Science and Technology 631.7%4034
19University of Electronic Science and Technology of China 631.7%5616
20City University of Hong Kong 611.7%3050
21Southeast University 611.7%3247
22Carnegie Mellon University 591.6%5815
23Harbin Institute of Technology Shenzhen Graduate School 581.6%3936
24Huawei Technologies (China) 581.6%5218
25Wuhan University 571.6%4428
26Beijing University of Posts and Telecommunications 561.5%4035
27Peng Cheng Laboratory 561.5%6113
28Shenzhen University 551.5%3344
29Xi'an Jiaotong University 541.5%5119
30Huazhong University of Science and Technology 531.4%4230

Download all institutions as CSV

About this data

Papers. AAAI 2025 lists 3,735 papers on Lacuna, duplicate versions counted once; rejected and withdrawn submissions are left out.

Directions. Directions are Lacuna’s clustering of the concept elements extracted from each paper, so a paper can belong to several and direction counts add up to more than the paper count. A direction’s share is its papers divided by the edition’s papers. Comparisons with 2024 use shares, since the edition went from 3,097 to 3,735 papers.

Institutions. A paper credits each institution named in its affiliations, resolved to a ROR record, once. 3,663 of 3,735 papers (98.1%) credit at least one institution and are the base for institution shares; the rest have no resolved affiliation yet.

Data. Counts come from Lacuna’s data build of 2026-11-18 and change as papers are added. Tables as CSV: directions · institutions. Cite as: Lacuna, “AAAI 2025”, https://lacuna.tiptreesystems.com/venue/aaai/2025.

Top Authors

AuthorPapers
Dongxiao He9
Gengyu Lyu8
Shuicheng YAN8
Chaozhuo Li7
Chaozhuo Li7
Cheng Wang7
Di Jin7
Mingli Song7
Tianlong Chen7
Xueyang Fu7
Yuxuan Liang7
Hao Fei6
Sarit Kraus6
Shirui Pan6
Tao Dai6
Tian Ye6
Wangmeng Zuo6
Wenqi Ren6
Xiangyu Zhao6
Yongjian Deng6

Impactful Papers

Papers from AAAI 2025 that shaped research clusters.

Foundational

Influential

Papers

Title
10 Open Challenges Steering the Future of Vision-Language-Action Models
(1+1) Genetic Programming with Functionally Complete Instruction Sets Can Evolve Boolean Conjunctions and Disjunctions with Arbitrarily Small Error
2-in-1 Phishing Detection via Large LM Distillation and Small LM Perturbation (Student Abstract)
3CAD: A Large-Scale Real-World 3C Product Dataset for Unsupervised Anomaly Detection
3D²-Actor: Learning Pose-Conditioned 3D-Aware Denoiser for Realistic Gaussian Avatar Modeling
3D Annotation-Free Learning by Distilling 2D Open-Vocabulary Segmentation Models for Autonomous Driving
3D-aware Select, Expand, and Squeeze Token for Aerial Action Recognition
3D Denoisers Are Good 2D Teachers: Molecular Pretraining via Denoising and Cross-Modal Distillation
3DHumanEdit: Multi-modal Body Part-aware Conditioning Information Integration for 3D Human Manipulation
3DMambaIPF: A State Space Model for Iterative Point Cloud Filtering via Differentiable Rendering
3D Measurement of Complex Textured Objects Based on Bidirectional Fringe Projection
3DPGS: 3D Probabilistic Graph Search for Archaeological Piece Grouping
3D-RPE: Enhancing Long-Context Modeling Through 3D Rotary Position Encoding
3M-Game: Multi-Modal Multi-Task Multi-Teacher Learning for Game Event Detection (Student Abstract)
3SAT: A Simple Self-Supervised Adversarial Training Framework
4D Diffusion for Dynamic Protein Structure Prediction with Reference and Motion Guidance
A (1+?)-Approximation for Ultrametric Embedding in Subquadratic Time
A²RNet: Adversarial Attack Resilient Network for Robust Infrared and Visible Image Fusion
AAAI-25, Sponsored by the Association for the Advancement of Artificial Intelligence, February 25 - March 4, 2025, Philadelphia, PA, USA
AAKR: Adversarial Attack-based Knowledge Retention for Continual Semantic Segmentation
ABC3: Active Bayesian Causal Inference with Cohn Criteria in Randomized Experiments
A Black-Box Evaluation Framework for Semantic Robustness in Bird's Eye View Detection
ABQ-LLM: Arbitrary-Bit Quantized Inference Acceleration for Large Language Models
Accelerated Diffusion via High-Low Frequency Decomposition for Pan-Sharpening
Accelerated Methods with Compressed Communications for Distributed Optimization Problems Under Data Similarity
Accessible, At-Home Detection of Parkinson's Disease via Multi-Task Video Analysis
Accessible Hardware Implementation for Multi-Agent Collective Construction
ACCon: Angle-Compensated Contrastive Regularizer for Deep Regression
Accounting for Spatial Variability with the Histogram of Oriented Gradients Based Masking Improves Performance of Masked Autoencoder over Hyperspectral Satellite Imagery (Student Abstract)
Accurate and Regret-Aware Numerical Problem Solver for Tabular Question Answering
Accurate Estimation of Feature Importance Faithfulness for Tree Models
Accurate Link Prediction for Edge-Incomplete Graphs via PU Learning
Accurate Nucleic Acid-Binding Residue Identification Based Domain-Adaptive Protein Language Model and Explainable Geometric Deep Learning
Achieving Balanced Representation in School Choice with Diversity Goals
Achieving Ensemble-Like Performance in a Single Model: A Feature Diversification Framework for Image-Text Matching
Achieving Lightweight Super-Resolution for Real-Time Computer Graphics
Achieving Maximin Share and EFX/EF1 Guarantees Simultaneously
Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement
A Compact Implicit Neural Representation for Efficient Storage of Massive 4D Functional Magnetic Resonance Imaging
A Compact Model for Mathematics Problem Representations Distilled from BERT
A Complete Algorithm for Optimization Modulo Nonlinear Real Arithmetic
A Comprehensive Evaluation on Event Reasoning of Large Language Models
A Comprehensive Overhaul of Multimodal Assistant with Small Language Models
A Computationally Grounded Framework for Cognitive Attitudes
A Continuous-time Tractable Model for Present-biased Agents
Acoustic-to-Hyper-Spectral: Hyper-Spectral Image Construction from Frequency Spectrums Through Simulated Annealing (Student Abstract)
ACPBench: Reasoning About Action, Change, and Planning
Acting Beyond Learning: Imagination-Assisted Decision-Making in the Visual-based Multi-Agent Cooperative Scenarios
Action-Agnostic Point-Level Supervision for Temporal Action Detection
Active Fourier Auditor for Estimating Distributional Properties of ML Models

Page 1/75 next »