Research Blogs News Tags Projects

Research

Select a Tag

Published on
June 5, 2025
From Theory to Practice: Real-World Use Cases on Trustworthy LLM-Driven Process Modeling, Prediction and Automation
process-modeling generalization-and-specialization BPMN-design pharmacovigilance Business-Process-Management prediction sustainable-textile-design human-AI-collaboration human-agency-versus-automation transparency-and-efficiency automation knowledge-graph-augmented-LLMs Large-Language-Models explainable-Machine-Learning multi-agent-systems
• German Research Center for Artificial Intelligence (DFKI) and Saarland University, 66123 Saarbrücken, Saarland, Germany
Published on
June 5, 2025
ConText: Driving In-context Learning for Text Removal and Segmentation
visual-in-context-learning self-prompting-strategy optical-character-recognition task-chaining-compositor ConText-model text-segmentation reasoning-as-reconstruction context-aware-aggregation text-removal visual-heterogeneity
• University of Science and Technology• Tsinghua University• Peking University• Fudan University
Published on
June 5, 2025
CoLa: Chinese Character Decomposition with Compositional Latent Components
latent-variable-model Chinese-character-recognition compositionality cross-dataset-generalization zero-shot-problem compositional-components learning-to-learn cognitive-principles Chinese-character-decomposition
• Shanghai Key Laboratory of Intelligent Information Processing, School of Computer Science, Fudan University
Published on
June 5, 2025
Mark My Words: A Robust Multilingual Model for Punctuation in Text and Speech Transcripts
transcripts-of-spontaneous-speech large-language-models multilingual-models low-resource-NLP disfluencies punctuation-restoration downstream-tasks
• Nilekani Centre at AI4Bharat• Indian Institute of Technology, Madras
Published on
June 5, 2025
Attention-Only Transformers via Unrolled Subspace Denoising
self-attention vision-tasks signal-to-noise-ratio transformers denoising low-dimensional-subspaces representation-learning language-tasks
• Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor
Published on
June 5, 2025
Knockout LLM Assessment: Using Large Language Models for Evaluations through Iterative Pairwise Comparisons
evaluations Pearson-correlation university-level-exam-scoring scoring-accuracy Large-Language-Models pairwise-comparisons machine-translation knockout-tournament-system
• Karlsruhe Institute of Technology
Published on
June 5, 2025
When Does Closeness in Distribution Imply Representational Similarity? An Identifiability Perspective
pre-training-approaches wider-networks representational-similarity deep-neural-networks Kullback-Leibler-divergence empirical-observations identifiability-theory distributional-distance CIFAR-10 autoregressive-language-models
• Technical University of Denmark• University of Trento, Italy• Helmholtz AI, Munich• Technical University of Munich
Published on
June 5, 2025
Unifying Uniform and Binary-coding Quantization for Accurate Compression of Large Language Models
accurate-compression large-language-models computational-overhead optimization-techniques binary-coding-quantization flexible-mapping memory-overhead quantization uniform-quantization
• Seoul National University• NAVER Cloud
Published on
June 5, 2025
FedFACT: A Provable Framework for Controllable Group-Fairness Calibration in Federated Learning
Bayes-optimal-classifiers personalized-cost-sensitive-learning convergence-guarantees bi-level-optimization group-fairness model-fairness multi-class-classification Global-Fairness generalization-guarantees accuracy-fairness-trade-off Federated-Learning Local-Fairness
• Zhejiang University• Hangzhou Dianzi University
Published on
June 5, 2025
ClozeMath: Improving Mathematical Reasoning in Language Models by Learning to Fill Equations
large-language-models text-infilling-task Chain-of-Thought-decoding Beam-Search ablation-study MATH ClozeMath mathematical-reasoning GSM8K Chain-of-Thought-format masked-equations GSM-Symbolic
• Qualcomm AI Research• Nanyang Technological University, Singapore
Published on
June 5, 2025
AhaKV: Adaptive Holistic Attention-Driven KV Cache Eviction for Efficient Inference of Large Language Models
benchmark-tasks value-vectors KV-cache information-entropy inference bias-reduction Artificial-Intelligence attention-score Large-Language-Models self-attention-mechanism
• South China University of Technology• South China University of Technology; Pazhou Laboratory
Published on
June 5, 2025
Act-as-Pet: Benchmarking the Abilities of Large Language Models as E-Pets in Social Network Services
virtual-pet-companionship intelligent-scheduling memory-based-dialogues benchmarking self-evolution self-interaction human-interaction developmental-behaviors psychological-conversations Large-Language-Models
Published on
June 5, 2025
Understanding Physical Properties of Unseen Deformable Objects by Leveraging Large Language Models and Robot Actions
physical-properties task-planning bin-packing robot-actions unseen-objects deformability Large-Language-Models
Published on
June 5, 2025
Scaling CrossQ with Weight Normalization
model-free-reinforcement-learning training-dynamics update-to-data-ratio Q-bias-explosion sample-efficiency critic-network-weights weight-normalization DeepMind-control-benchmark CrossQ Reinforcement-learning
• Technical University of Darmstadt, hessian.AI• University of Freiburg• Technical University of Darmstadt, hessian.AI, German Research Center for AI (DFKI), Robotics Institute Germany (RIG)
Published on
June 5, 2025
PPO in the Fisher-Rao geometry
Trust-Region-Policy-Optimization Fisher-Rao-PPO sub-linear-convergence theoretical-guarantees monotonic-policy-improvement reinforcement-learning KL-divergence policy-gradient-method surrogate-loss Fisher-Rao-geometry Proximal-Policy-Optimization
• School of Mathematical and Computer Sciences, Heriot-Watt University, Edinburgh, UK, and Maxwell Institute for Mathematical Sciences, Edinburgh, UK• Not provided• School of Mathematics, University of Edinburgh, UK, and Simtopia, UK• School of Mathematics, University of Edinburgh, UK, The Alan Turing Institute, UK and Simtopia, UK
Published on
June 5, 2025
Misalignment or misuse? The AGI alignment tradeoff
AI-control-methods misaligned-AGI social-context misuse-risk governance AGI-alignment AI-safety alignment-techniques catastrophic-risks robustness
Published on
June 5, 2025
HUMOF: Human Motion Forecasting in Interactive Social Scenes
human-environment-interactions human-motion-forecasting motion-prediction-methods human-human-interactions hierarchical-interaction-feature-representation interactive-social-scenes coarse-to-fine-interaction-reasoning-module spatial-and-frequency-perspectives
• The University of Hong Kong• ShanghaiTech University• Sun Yat-sen University• Shanghai Jiao Tong University
Published on
June 5, 2025
Dropout-Robust Mechanisms for Differentially Private and Fully Decentralized Mean Estimation
pairwise-noise-cancellation communication-cost decentralized-mean-estimation data-intensive-processing differential-privacy low-variance-correlated-noise accuracy information-leakage cryptographic-solutions adversarial-models
• Insa-Lyon, Lyon, France
Published on
June 5, 2025
An Open-source Capping Machine Suitable for Confined Spaces
vision-system manual-capping capping-performance compact-spaces industrial-capping-machines open-source-technology automated-capping self-driving-laboratories
• University of XYZ• Institute of ABC• Department of DEF• School of GHI
Published on
June 5, 2025
PromptCanvas: Composable Prompting Workspaces Using Dynamic Widgets for Exploration and Iteration in Creative Writing
composable-prompting creativity-support collaborative-writing cognitive-load PromptCanvas AI-generated-content widget-based-experience dynamic-interfaces
• LMU Munich, Munich, Germany
Published on
June 5, 2025
SAAT: Synergistic Alternating Aggregation Transformer for Image Super-Resolution
feature-dependency self-attention Efficient-Channel-&-Window-Synergistic-Attention-Group spatial-attention image-super-resolution channel-attention Spatial-&-Window-Synergistic-Attention-Group spatial-structure-relationships Transformers
• Ocean University of China
Published on
June 5, 2025
ComRoPE: Scalable and Robust Rotary Position Embedding Parameterized by Trainable Commuting Angle Matrices
ImageNet-1K-dataset Rotary-Positional-Encoding scalability positional-robustness Transformer-architecture positional-information trainable-commuting-angle-matrices positional-encoding-research
• Tsinghua University• Tsinghua University, Shenzhen University
Published on
June 5, 2025
Generating Pedagogically Meaningful Visuals for Math Word Problems: A New Benchmark and Analysis of Text-to-Image Models
visual-generation multimodal-education Text-to-Image-models math-word-problems pedagogically-meaningful-visuals automated-methods educational-content
Published on
June 5, 2025
Sign-SGD is the Golden Gate between Multi-Node to Single-Node Learning: Significant Boost via Parameter-Free Optimization
distributed-learning large-language-models machine-learning-problems deterministic-Sign-SGD memory-efficient-approach gradient-compression Sign-SGD effective-stepsize multi-node-learning resource-intensive-training stochastic-learning momentum-methods single-node-training
Published on
June 5, 2025
Verbalized Confidence Triggers Self-Verification: Emergent Behavior Without Explicit Reasoning Supervision
large-language-models calibrated-uncertainty interpretability self-verification-behavior verbalized-confidence-score test-time-scaling chain-of-thought-reasoning uncertainty-calibration confidence-calibration supervised-fine-tuning
• Department of Physics, University of Science• Department of Chemistry, University of Science• Department of Biology, University of Science• Department of Mathematics, University of Science
Published on
June 5, 2025
MFLA: Monotonic Finite Look-ahead Attention for Streaming Speech Recognition
quasi-monotonic-alignment wait-k-decoding-strategy Monotonic-Finite-Look-ahead-Attention large-pre-trained-speech-models Continuous-Integrate-and-Fire-mechanism latency-and-quality-trade-off prefix-to-prefix-training-framework streaming-speech-recognition
• Department of Computer Science, University of XYZ• Department of Mathematics, University of ABC• Department of Physics, University of DEF• Department of Chemistry, University of GHI
Published on
June 5, 2025
On the Closed-Form of Flow Matching: Generalization Does Not Arise from Target Stochasticity
diffusion-techniques statistical-performance image-datasets high-dimensional-settings conditional-flow-matching-loss inductive-biases flow-matching-techniques deep-generative-models synthetic-samples
• Université Jean Monnet Saint-Etienne, CNRS, Institut d’Optique Graduate School, Inria, Laboratoire Hubert Curien UMR 5516, F-42023, SAINT-ETIENNE, France• Inria, ENS de Lyon, CNRS, Université Claude Bernard Lyon 1, LIP, UMR 5668, 69342, Lyon cedex 07, France• Unknown
Published on
June 5, 2025
FSHNet: Fully Sparse Hybrid Network for 3D Object Detection
feature-extraction dynamic-sparse-label-assignment Argoverse2-benchmark sparse-upsampling-module SlotFormer-block long-range-detection sparse-3D-detectors Waymo-benchmark 3D-object-detection nuScenes-benchmark sparse-voxels
• School of Computer Science and Engineering, Sun Yat-sen University• Key Laboratory of Machine Intelligence and Advanced Computing, Sun Yat-sen University• School of Artificial Intelligence, Sun Yat-sen University
Published on
June 5, 2025
Pl\"uckeRF: A Line-based 3D Representation for Few-view Reconstruction
Published on
June 5, 2025
OSGNet @ Ego4D Episodic Memory Challenge 2025
Moment-Queries Ego4D-Episodic-Memory-Challenge localization-accuracy egocentric-video-localization early-fusion-based-video-localization-model Goal-Step Natural-Language-Queries
• Harbin Institute of Technology (Shenzhen)• Shandong Jianzhu University