Machine Learning

ShengYun Peng, Pin-Yu Chen, Jianfeng Chi et al. · gatech

This addresses safety vulnerabilities in LLM customization for users and developers, offering a novel mitigation approach that is not incremental but builds on a new paradigm.

Weizhi Zhang, Yangning Li, Yuanchen Bei et al. · pku

This addresses the problem of handling complex, multi-step information retrieval for users across diverse domains, representing a potential paradigm shift rather than an incremental improvement.

Silin Chen, Shaoxin Lin, Xiaodong Gu et al.

This addresses the inefficiency of redundant exploration in automated software engineering for developers, representing a new paradigm rather than an incremental improvement.

Botian Wang, Yawen Ouyang, Yaohui Li et al.

This addresses the challenge of handling diverse material tasks for researchers in materials science, representing a novel paradigm rather than an incremental improvement.

Ling Yang, Zhaochen Yu, Chenlin Meng et al.

This addresses a key limitation in text-to-image generation for applications requiring detailed and compositional visual content, representing a novel integration of reasoning and diffusion rather than an incremental improvement.

Yong Liu, Haoran Zhang, Chenyu Li et al.

This work addresses the need for scalable and generalizable models in time series analysis, moving beyond scenario-specific small models, though it is an early development in the field.

Marcos V. Conde, Gregor Geigle, Radu Timofte

This work addresses the challenge of flexible and high-quality image restoration for users by enabling natural language control, representing a novel benchmark in the field.

Anne Ouyang, Simon Guo, Simran Arora et al.

This work addresses the problem of efficient GPU kernel generation for machine learning architectures, which is significant for ML engineers and researchers.

Yingming Pu, Tao Lin, Hongyu Chen

This addresses the need for more efficient and systematic AI-driven scientific discovery across domains like nanomaterials and biomolecules, representing a novel paradigm shift rather than an incremental improvement.

Niklas Muennighoff, Hongjin Su, Liang Wang et al. · microsoft-research

This addresses the inefficiency of using separate models for retrieval and generation in applications like RAG, speeding it up by over 60% for long documents.

Cong Hua, Qianqian Xu, Shilong Bao et al.

This addresses the problem of dominant modalities overpowering weak ones in multi-modal learning, offering a novel approach for researchers in that domain.

Juyuan Wang, Rongchen Zhao, Wei Wei et al.

This addresses the challenge of stateful long narrative reasoning for applications like story analysis and comprehension, representing a novel paradigm rather than an incremental improvement.

Bidipta Sarkar, Warren Xia, C. Karen Liu et al.

This work addresses the problem of effective communication in complex social settings for autonomous agents and humans, providing an incremental improvement in multi-agent reinforcement learning.

Lianghui Zhu, Bencheng Liao, Qian Zhang et al.

This work addresses the problem of computation and memory constraints in high-resolution image processing for vision researchers and practitioners, offering a potential next-generation backbone with significant efficiency gains.

Nikola Jovanović, Robin Staab, Martin Vechev

This exposes a critical security flaw in AI-generated content detection, challenging the deployment readiness of current LLM watermarking methods.

Zhiqiang Que, Chang Sun, Sudarshan Paramesvaran et al.

This enables accurate, real-time GNN inference for particle physics trigger systems, advancing next-generation hardware triggers at CERN.

Yichuan Mo, Yuji Wang, Zeming Wei et al. · pku

This addresses security vulnerabilities in LLMs for users relying on safe AI interactions, representing a novel approach to intrinsic robustness through prompt optimization.

Miltiadis Kofinas, Boris Knyazev, Yan Zhang et al.

This addresses the challenge of processing neural network parameters with permutation symmetry for researchers and practitioners in machine learning, offering a novel approach that is not incremental.

Maksym Andriushchenko, Francesco Croce, Nicolas Flammarion

This exposes critical security flaws in widely used AI models, posing risks for deployment in sensitive applications.

Yue Wu, Zhiqing Sun, Huizhuo Yuan et al. · cmu

This addresses the challenge of accurately capturing human preferences for language model alignment, offering a novel approach that outperforms existing methods without relying on external supervision from stronger models.