KeyRec: Bounded Visual Memory for Streaming and Long-Video Understanding

KeyRec creates bounded visual memory via recent caches and structured event banks to enable efficient long-video and streaming understanding with only 10% of visual tokens, outperforming compressed baselines.

Zihan Chen, Xuejian Rong, Xiaojuan Wang, Boqing Gong and 3 more

Published Sep 26, 2026 · 0 citations · ▲ 11 on Hugging Face

AdaST: Adaptive Coupling for Spatial-Temporal Forecasting

AdaST adaptively decomposes and recombines spatial-temporal data via heterogeneity-aware experts to match distinct coupling regimes, significantly outperforming state-of-the-art forecasting baselines.

Zhenyu Lei, Chenghao Liu, Yushun Dong, Qi R. Wang and 1 more

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published Sep 20, 2026 · 0 citations

Grounding Memory Summarization in Utility Intent

MemSuit improves memory summarization by self-distilling query-conditioned utility awareness into raw-conversation entries and decomposing blocks to prevent collateral erasure, boosting answer quality across query types.

Zhenyu Lei, Mingjia Shi, Xingbo Fu, Haoyu He and 2 more

Published Aug 20, 2026 · 0 citations

TeacherGRPO: Closing the Capacity Gap in Reasoning Distillation via Teacher Alignment

TeacherGRPO aligns teachers to student distributions via reinforcement learning to overcome reasoning distillation's Gap Curse and improves student performance.

Zhenyu Lei, Zihan Chen, Yaochen Zhu, Shangbin Feng and 4 more

Published Aug 20, 2026 · 0 citations

MoCo: A One-Stop Shop for Model Collaboration Research

MoCo unifies 26 model collaboration methods and 25 benchmarks to show collaboration outperforms single models in 61% of settings by up to 25.8%.

Shangbin Feng, Yuyang Bai, Ziyuan Yang, Yike Wang and 16 more

Published Jan 29, 2026 · 0 citations · Code ★ 63

Reforming the Mechanism: Editing Reasoning Patterns in LLMs with Circuit Reshaping

REdit reshapes LLM neural circuits before editing to reduce reasoning-pattern interference, improving generality and locality over broad training baselines.

Zhenyu Lei, Qiong Wu, Jianxiong Dong, Yinhan He and 3 more

Published Jan 25, 2026 · 0 citations

BrainTAP: Brain Disorder Prediction with Adaptive Distill and Selective Prior Integration

BrainTAP predicts brain disorders by adaptively distilling cross-modal connectivity features and selectively fusing expert anatomical priors to outperform baselines on the ABCD dataset.

Zhenyu Lei, Aiying Zhang, Song Wang, Han Fan and 1 more

Published Jan 15, 2026 · 0 citations

LOKA: Conflict-Aware LLM Knowledge Update with Adaptive Knowledge Memory

LOKA introduces conflict-aware LLM knowledge updates using adaptive multi-unit memory with learned routing, improving accuracy and flexibility over separate unlearning and learning methods.

Binchi Zhang, Zhengzhang Chen, Zaiyi Zheng, Jundong Li and 1 more

Published 2026 · 0 citations

MolEdit: Knowledge Editing for Multimodal Molecule Language Models

MolEdit enables targeted knowledge editing in multimodal molecule language models via multi-expert adapters and expertise-aware switching, improving reliability and locality.

Zhenyu Lei, Patrick Soga, Yaochen Zhu, Yinhan He and 2 more

Published Oct 20, 2025 · 0 citations

MAPLE: Many-Shot Adaptive Pseudo-Labeling for In-Context Learning

MAPLE uses influence-based pseudo-labeling to adaptively select many-shot ICL demonstrations, boosting LLM performance without extensive labeling costs.

Zihan Chen, Song Wang, Zhen Tan, Jundong Li and 1 more

Published May 22, 2025 · 0 citations

Virtual Nodes Can Help: Tackling Distribution Shifts in Federated Graph Learning

FedVN uses learnable virtual nodes and client-specific edge generators to eliminate graph distribution shifts across federated clients, improving GNN performance.

Xingbo Fu, Zihan Chen, Yinhan He, Song Wang and 3 more

Published Apr 11, 2025 · 3 citations

Out-of-Distribution Generalization on Graphs via Progressive Inference

GPro improves graph out-of-distribution generalization by decomposing causal invariant learning into progressive inference steps and augmenting training with counterfactual samples, outperforming state-of-the-art methods by 4.91%.

Yiming Xu, Bin Shi, Zhen Peng, Huixiang Liu and 2 more

Published Apr 11, 2025 · 3 citations

BrainMAP: Learning Multiple Activation Pathways in Brain Networks

BrainMAP learns multiple brain network activation pathways via sequential models and Mixture-of-Experts, improving fMRI analysis and interpretability.

Song Wang, Zhenyu Lei, Zhen Tan, Jiaqi Ding and 7 more

Published Apr 11, 2025 · 2 citations

Beyond the Permutation Symmetry of Transformers: The Role of Rotation for Model Fusion

Rotation symmetry generalizes permutation symmetry continuously for transformers, improving parameter matching and model fusion across language and vision tasks.

Binchi Zhang, Zaiyi Zheng, Zhengzhang Chen, Jundong Li

Published Feb 1, 2025 · 0 citations

Enhancing Distribution and Label Consistency for Graph Out-of-Distribution Generalization

Proposed framework improves graph OOD generalization by jointly enhancing environment augmentation and invariant subgraph extraction consistency.

Song Wang, Xiaodong Yang, Rashidul Islam, Huiyuan Chen and 3 more

Published Jan 7, 2025 · 0 citations

CoRAG: Enhancing Hybrid Retrieval-Augmented Generation through a Cooperative Retriever Architecture

CoRAG dynamically selects textual or graph retrieval and blends results for global hybrid knowledge access, outperforming local hybrid RAG on QA benchmarks.

Zaiyi Zheng, Song Wang, Zihan Chen, Yaochen Zhu and 4 more

Published 2025 · 0 citations

Graph Neural Networks Are More Than Filters: Revisiting and Benchmarking from A Spectral Perspective

A spectral benchmark reveals that graph neural networks capture diverse frequency components through non-linear layers, not just neighborhood aggregation filters.

Yushun Dong, Patrick Soga, Yinhan He, Song Wang and 1 more

Published Dec 10, 2024 · 1 citation

Enhancing Distribution and Label Consistency for Graph Out-of-Distribution Generalization

A modifier unifies graph augmentation and invariant subgraph extraction to improve distribution and label consistency for graph OOD generalization.

Song Wang, Xiaodong Yang, Rashidul Islam, Huiyuan Chen and 3 more

Published Dec 9, 2024 · 3 citations

Adversarial Attacks on Fairness of Graph Neural Networks

G-FairAttack is a framework that adversarially corrupts fairness in fairness-aware graph neural networks while preserving prediction utility.

Binchi Zhang, Yushun Dong, Chen Chen, Yada Zhu and 2 more

Published Oct 20, 2023 · 1 citation · ▲ 1 on Hugging Face · Code ★ 11

LMBot: Distilling Graph Knowledge into Language Model for Graph-less Deployment in Twitter Bot Detection

LMBot distills graph neural network knowledge into language models for efficient graph-less Twitter bot detection, achieving state-of-the-art results across four benchmarks.

Cai, Zijian, Zhaoxuan Tan, Zhenyu Lei, Zhu, Zifeng and 3 more

Published Jun 30, 2023 · 0 citations

BIC: Twitter Bot Detection with Text-Graph Interaction and Semantic Consistency

BIC detects Twitter bots via text-graph interaction and semantic consistency modeling, achieving high detection accuracy.

Zhenyu Lei, Herun Wan, Wenqian Zhang, Shangbin Feng and 4 more

Published 2023 · 35 citations

KCD: Knowledge Walks and Textual Cues Enhanced Political Perspective Detection in News Media

KCD integrates knowledge walks and textual cues to improve political perspective detection in news media.

Wenqian Zhang, Shangbin Feng, Zilong Chen, Zhenyu Lei and 2 more

Published 2022 · 18 citations

Mixture-of-Experts for Online Matrix Completion on a Drifting Union of Subspaces

Renpu Liu, Jing Yang

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Lost on Campus: Evaluating Embodied Spatial Reasoning of Vision-Language Models in the Wild

Zehan Zheng, Yanyuan Chen, Deming Li, Yutao Tang and 4 more

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026

AI Construct Lexis: An Ontology of the Hidden Assumptions in AI Evaluation

Olawale Salaudeen, Florian E. Dorner, Tom Sühr, Sang Truong and 12 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Embedding Security Properties into AI-Enabled Cyber-Physical Systems

Ziyan An, John Stankovic, Meiyi Ma

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026

Click3R: Interactive Stereo 3D Reconstruction with Sparse Correspondence Clicks

Hao Gu, Jin Yao, Zezhou Cheng

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Saliency-Aware Multi-Route Thinking: Grounding and Reasoning on Vision-Language Agents

Mingjia Shi, Yinhan He, Yaochen Zhu, Cassie Dong and 1 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

RnR: a meta-solver for causal discovery in undersampled time series data

Mohammadsajad Abavisani, Kseniya Solovyeva, David Danks, Vince D. Calhoun and 1 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

When Safety Becomes An Outlier: Understanding the Retention of LLM Safety Behaviors

Binchi Zhang, Hadi Abdullah, Yiwei Cai, Jundong Li

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

Reinforcing Multimodal Reasoning Against Visual Degradation

ROMA improves multimodal reasoning robustness to visual corruption via dual-pass RL optimization that avoids reward poisoning while preserving clean accuracy.

Rui Liu, Dian Yu, Haolin Liu, Yucheng Shi and 4 more

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 5 on Hugging Face

DD-Ranking: Rethinking the Evaluation of Dataset Distillation

DD-Ranking reveals dataset distillation gains come from extra evaluation techniques rather than image quality, proposing fair metrics to assess true synthetic dataset value.

Zekai Li, Xinhao Zhong, Samir Khaki, Zhiyuan Liang and 36 more

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

Search-Augmented Masked Diffusion Models for Constrained Generation

SearchDiff integrates informed search into masked diffusion denoising to enforce hard constraints and optimize non-differentiable properties at inference time without retraining, outperforming discrete diffusion and autoregressive baselines.

Huu Binh Ta, Michael Cardei, Alvaro Velasquez, Ferdinando Fioretto

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

Simulation-Informed Diffusion for Decentralized Multi-robot Motion Planning

Simulation-Informed Diffusion predicts neighboring robots' future trajectories to plan safer decentralized multi-robot motion with minimal communication, scaling to 108 robots amid 160 obstacles.

JINHAO LIANG, Sven Koenig, Ferdinando Fioretto

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

MedKIT: Evaluating Knowledge Integration and Generalization in Large Language Models

MedKIT evaluates medical LLM knowledge integration via clinical updates, revealing strong recall but limited relational, compositional, and operational generalization across 12 strategies.

Lukas Thede, Yash Kumar, David Chen, Danielle Bitterman and 3 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Predicting Plasticity in Deep Continual Learning: A Theoretical Perspective

Existing plasticity diagnostics fail to predict trainability, but optimization readiness, combining gradient strength and reliability, lower-bounds optimization gain and predicts plasticity more reliably.

Jiuqi Wang, Jayanth Srinivasa, Claire Chen, Shuze D Liu and 2 more

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026

MathlibPR: Pull Request Merge-Readiness Benchmark for Formal Mathematical Libraries

MathlibPR benchmarks LLMs on reviewing real Mathlib4 pull requests, finding they cannot distinguish merge-ready from build-passing but unmerged PRs.

Zixuan Xie, Xinyu Liu, Shangtong Zhang

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

PixelDense: Dense Prediction as Representation Alignment for Pixel Diffusion

PixelDense aligns pixel diffusion with frozen dense-prediction teachers via separate semantic and geometric projection streams and orthogonality penalties, improving GenEval to 0.8093 and training speed by 1.23x.

Lehan Yang, Daiqing Qi, Wenhao Zhang, Avery Li and 8 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 20 on Hugging Face · Code ★ 3

G-Zero: Self-Play for Open-Ended Generation from Zero Data

G-Zero uses intrinsic predictive-shift rewards in a verifier-free co-evolutionary framework that enables continuous LLM self-improvement across open-ended unverifiable domains without external judges.

Chengsong Huang, Haolin Liu, Tong Zheng, Runpeng(Leo) Dai and 6 more

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 17 on Hugging Face · Code ★ 30

Rethinking Visual Attribution for Chest X-ray Reasoning in Large Vision Language Models

A causal framework reveals standard visual attribution methods poorly explain chest X-ray reasoning in vision-language models, and MedFocus improves evidence localization via concept-based optimal transport.

Guangzhi Xiong, Qiao Jin, Sanchit Sinha, Zhiyong Lu and 1 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 1 on Hugging Face · Code ★ 6

On the Divergence of Differential Temporal Difference Learning without Local Clocks

Differential temporal difference learning converges with local clocks but can diverge with global clocks in average-reward reinforcement learning.

David Antrobius, Shangtong Zhang

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

Latent Q-Barrier Shielding for Safe In-Context Reinforcement Learning

Latent Q-Barrier shielding filters actions via predicted future costs to improve safe in-context reinforcement learning reward-safety tradeoffs under out-of-distribution deployment shifts.

Minjae Kwon, Amir Moeini, Shangtong Zhang, Lu Feng

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026

Constrained Code Generation with Discrete Diffusion

CDC integrates constraint satisfaction into discrete diffusion code generation via training-free neurosymbolic operators that locally steer denoising toward feasible programs, improving correctness, security, and syntax over baselines.

Lize Shao, Michael Cardei, Zichen Xie, Ferdinando Fioretto and 1 more

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026

CASL: Concept-Aligned Sparse Latents for Interpreting Diffusion Models

CASL aligns diffusion model sparse latents with semantic concepts via supervised linear mapping, enabling precise concept-specific editing and causal interpretability.

Zhenghao He, Guangzhi Xiong, Boyang Wang, Sanchit Sinha and 1 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

Drift Flow Matching

Drift Flow Matching bridges one-step drift models and iterative flow matching to enable adaptive sampling computation for flexible generation quality and efficiency.

Chenrui Ma, Xi Xiao, Lin Zhao, Tianyang Wang and 2 more

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

You Only Need Minimal RLVR Training: Extrapolating LLMs via Rank-1 Trajectories

RLVR weight updates are near rank-1 and predictable, so RELEX extrapolates them via linear regression to match full training with only 15% of steps.

Zhepei Wei, Xinyu Zhu, Wei-Lin Chen, Chengsong Huang and 2 more

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 50 on Hugging Face · Code ★ 16

Almost Sure Convergence Rates of Stochastic Approximation and Reinforcement Learning via a Poisson-Moreau Drift

A Poisson-Moreau drift yields near-optimal almost sure convergence rates for contractive stochastic approximation under Markovian noise.

Xinyu Liu, Zixuan Xie, Shangtong Zhang

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026

Achieving Optimal Static and Dynamic Regret Simultaneously in Bandits with Deterministic Losses

Optimal simultaneous static and dynamic regret is impossible against adaptive adversaries but achievable for deterministic losses against oblivious adversaries via negative static regret and Blackwell approachability.

Jian Qian, Chen-Yu Wei

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Aligning Language Model Benchmarks with Pairwise Preferences

Reweighting benchmark items aligns static language model benchmarks with downstream pairwise preferences to rank unseen models, using as few as 20 well-chosen models.

Marco Gutierrez, Xinyi Leng, Hannah Chen, Jonathan Richard Schwarz and 2 more

Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

VeriContest: A Competitive-Programming Benchmark for Verifiable Code Generation

VeriContest introduces 946 competitive programming problems with verified Rust specifications and proofs, showing state-of-the-art models reach only 5.29% on end-to-end verifiable generation.

Zichen Xie, Mrigank Pawagi, Yuxin Liu, Aaditi Rai and 4 more

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 1 on Hugging Face

Bridging Modalities, Spanning Time: Structured Memory for Ultra-Long Agentic Video Reasoning

MAGIC-Video unifies episodic, semantic, and visual content via a multimodal memory graph and narrative chain for agentic ultra-long video reasoning, outperforming prior agentic systems by up to 10.1 points.

Jiazheng Li, Chi-Hao Wu, Yunze Liu, Kaize Ding and 2 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026