Carnegie Mellon at ICLR 2026 – Machine Studying Weblog | ML@CMU

EditBench: Evaluating LLM Talents to Carry out Actual-World Instructed Code Edits

Authors: Wayne Chi (CMU), Valerie Chen (Carnegie Mellon College), Ryan Shar (Apple), Aditya Mittal (CMU, Carnegie Mellon College), Jenny Liang (College of Laptop Science, Carnegie Mellon College), Wei-Lin Chiang (UC Berkeley / LMSYS), Anastasios Angelopoulos (College of California Berkeley), Ion Stoica (), Graham Neubig (Carnegie Mellon College), Ameet Talwalkar (College of California-Los Angeles), Chris Donahue (CMU / Google DeepMind)

This work introduces EditBench, a brand new benchmark for testing how properly AI fashions can edit present code primarily based on consumer directions. In contrast to prior benchmarks, it makes use of real-world coding duties and contexts, together with issues like the encompassing code and cursor place. The benchmark contains 545 numerous issues, and outcomes present that the majority fashions battle—just a few obtain robust efficiency. The research additionally finds that having extra reasonable context considerably impacts how properly fashions carry out, highlighting the significance of evaluating code-editing in real-world settings.

UALM: Unified Audio Language Mannequin for Understanding, Technology and Reasoning

Authors: Jinchuan Tian (CMU, Carnegie Mellon College), Sang-gil Lee (NVIDIA), Zhifeng Kong (NVIDIA), Sreyan Ghosh (Nvidia), Arushi Goel (NVIDIA), Chao-Han Huck Yang (NVIDIA Analysis), Wenliang Dai (NVIDIA), Zihan Liu (Nvidia), Hanrong Ye (NVIDIA), Shinji Watanabe (Carnegie Mellon College), Mohammad Shoeybi (NVIDIA), Bryan Catanzaro (NVIDIA), Rafael Valle (NVIDIA), Wei Ping (Nvidia)

This paper introduces the Unified Audio Language Mannequin (UALM), a single mannequin designed to deal with audio understanding, text-to-audio era, and multimodal reasoning collectively. As a substitute of treating these as separate duties, UALM learns to each interpret and generate audio, reaching efficiency corresponding to specialised state-of-the-art fashions. The authors additionally present that combining textual content and audio throughout the mannequin’s reasoning course of improves its means to deal with advanced duties. Total, the work demonstrates a step towards extra common AI techniques that may motive throughout each language and sound.

Agent Knowledge Protocol: Unifying Datasets for Numerous, Efficient Wonderful-tuning of LLM Brokers

Authors: Yueqi Track (CMU), Ketan Ramaneti (Amazon), Zaid Sheikh (Carnegie Mellon College), Ziru Chen (Ohio State College, Columbus), Boyu Gou (Ohio State College, Columbus), Tianbao Xie (the College of Hong Kong, College of Hong Kong), Yiheng Xu (College of Hong Kong), Danyang Zhang (Shanghai Jiao Tong College), Apurva Gandhi (Carnegie Mellon College), Fan Yang (Fujitsu), Joseph Liu (College of Laptop Science, Carnegie Mellon College), Tianyue Ou (Carnegie Mellon College), Zhihao Yuan (Carnegie Mellon College), Frank F Xu (Carnegie Mellon College), Shuyan Zhou (Fb), Xingyao Wang (All Palms AI), Xiang Yue (Carnegie Mellon College), Tao Yu (College of Hong Kong), Huan Solar (Ohio State College), Yu Su (Ohio State College), Graham Neubig (Carnegie Mellon College)

This work introduces the Agent Knowledge Protocol (ADP), a standardized format for representing coaching information for AI brokers. The authors argue that the primary problem isn’t an absence of information, however that present datasets are fragmented throughout completely different codecs and instruments. ADP acts as a standard “interlingua,” making it simpler to mix numerous information sources—like coding, looking, and gear use—right into a single coaching pipeline. By changing 13 datasets into this unified format, the authors present that fashions skilled on the mixed information obtain improved efficiency.

MotionStream: Actual-Time Video Technology with Interactive Movement Controls

Authors: Joonghyuk Shin (Seoul Nationwide College), Zhengqi Li (Google), Richard Zhang (Adobe), Jun-Yan Zhu (Carnegie Mellon College), Jaesik Park (Seoul Nationwide College), Eli Shechtman (Adobe), Xun Huang (Adobe Analysis)

This paper introduces MotionStream, a system for producing movies in actual time primarily based on movement and textual content inputs. In contrast to prior strategies that take minutes to provide a video, MotionStream can stream outcomes at as much as 29 frames per second on a single GPU. The important thing concept is to coach a quick, causal mannequin that may generate video constantly, utilizing methods that stop high quality from degrading over lengthy sequences. Consequently, customers can interactively management movement—like drawing paths or transferring a digital camera—and see the video replace immediately.

OpenThoughts: Knowledge Recipes for Reasoning Fashions

Authors: Etash Guha (Stanford College, Anthropic), Ryan Marten (Harbor), Sedrick Keh (Toyota Analysis Institute), Negin Raoof (College of California, Berkeley), Georgios Smyrnis (College of Texas, Austin), Hritik Bansal (College of California, Los Angeles), Marianna Nezhurina (Juelich Supercomputing Heart, LAION, Tuebingen College), Jean Mercat (Toyota Analysis Institute (TRI)), Trung Vu (Google), Zayne Sprague (New York College), Ashima Suvarna (UCLA), Benjamin Feuer (Stanford College), Leon Liangyu Chen (Stanford College), Zaid Khan (College of North Carolina at Chapel Hill), Eric Frankel (Division of Laptop Science, College of Washington), Sachin Grover (Arizona State College), Caroline Choi (None), Niklas Muennighoff (Stanford College), Shiye Su (Stanford College), Wanjia Zhao (Stanford College), John Yang (Princeton College), Shreyas Pimpalgaonkar (New York College), Kartik sharma (Georgia Institute of Know-how), Charlie Ji (College of California, Berkeley), Yichuan Deng (Division of Laptop Science, College of Washington), Sarah Pratt (College of Washington), Vivek Ramanujan (Division of Laptop Science, College of Washington), Jon Saad-Falcon (Laptop Science Division, Stanford College), Stutee Acharya (College of South Florida), Jeffrey Li (Carnegie Mellon College), Achal Dave (Anthropic), Alon Albalak (SynthLabs), Kushal Arora (McGill College), Blake Wulfe (Toyota Analysis Institute), Chinmay Hegde (New York College), Greg Durrett (New York College), Sewoong Oh (College of Washington), Mohit Bansal (UNC Chapel Hill), Saadia Gabriel (College of Washington), Aditya Grover (UCLA), Kai-Wei Chang (College of Virginia Most important Campus), Vaishaal Shankar (Apple), Aaron Gokaslan (Cornell College), Mike Merrill (None), Tatsunori Hashimoto (Stanford College), Yejin Choi (Stanford College / NVIDIA), Jenia Jitsev (LAION; Juelich Supercomputing Heart, Analysis Heart Juelich), Reinhard Heckel (Technical College Munich), Maheswaran Sathiamoorthy (College of Southern California), Alex Dimakis (Electrical Engineering & Laptop Science Division, College of California, Berkeley), Ludwig Schmidt (College of Washington / Stanford / Anthropic)

This work introduces the OpenThoughts venture, which goals to create high-quality, open-source datasets for coaching reasoning-focused AI fashions. The authors present that fashions skilled on their public information can match or exceed the efficiency of robust present techniques that depend on non-public datasets. By fastidiously learning and bettering their information era course of, they construct bigger and higher datasets that considerably increase efficiency throughout math, coding, and science benchmarks. Total, the venture demonstrates that open information alone may be sufficient to coach extremely succesful reasoning fashions.

Mamba-3: Improved Sequence Modeling utilizing State Area Rules

Authors: Aakash Sunil Lahoti (CMU, Carnegie Mellon College), Kevin Li (Carnegie Mellon College), Berlin Chen (Princeton College), Caitlin Wang (Princeton College), Aviv Bick (Carnegie Mellon College), Zico Kolter (Carnegie Mellon College), Tri Dao (Princeton College), Albert Gu (Cartesia AI CMU)

This paper introduces Mamba-3, a brand new mannequin designed to make AI inference sooner and extra environment friendly with out sacrificing efficiency. Whereas many environment friendly options to Transformers scale back computation, they typically battle with duties like monitoring long-term data; Mamba-3 addresses this with improved state modeling and a extra expressive replace mechanism. The mannequin additionally makes use of a multi-input, multi-output design to spice up accuracy with out slowing down era. Total, Mamba-3 reveals that it’s potential to enhance each effectivity and functionality on the identical time, pushing ahead the tradeoff between velocity and efficiency.

Overcoming Joint Intractability with Lossless Hierarchical Speculative Decoding

Authors: Yuxuan Zhou (Impartial Researcher), Fei Huang (Alibaba Group), Heng Li (Carnegie Mellon College), Fengyi Wu (College of Washington), Tianyu Wang (College of Washington), Jianwei Zhang (Alibaba Group), Junyang Lin (Alibaba Group), Zhi-Qi Cheng (College of Washington)

This paper introduces Hierarchical Speculative Decoding (HSD), a brand new technique to hurry up giant language mannequin inference by bettering the verification step in speculative decoding whereas preserving actual output distributions. It addresses the problem of “joint intractability” in sequence-level verification by organizing resampling right into a hierarchy that redistributes chance mass throughout branches, enabling extra tokens to be accepted without delay. The strategy is theoretically confirmed to be lossless and empirically reveals constant velocity enhancements throughout fashions and benchmarks, outperforming prior tokenwise and blockwise verification strategies. Total, HSD presents a sensible and common method to speed up decoding with out sacrificing constancy, reaching state-of-the-art effectivity when built-in into present frameworks.

Distributional Equivalence in Linear Non-Gaussian Latent-Variable Cyclic Causal Fashions: Characterization and Studying

Authors: Haoyue Dai (Carnegie Mellon College), Immanuel Albrecht (FernUniversität in Hagen), Peter Spirtes (Carnegie Mellon College), Kun Zhang (Carnegie Mellon College & MBZUAI)

This paper research causal discovery in linear non-Gaussian fashions with latent variables and cycles, specializing in when completely different causal graphs are observationally indistinguishable. It offers the primary common characterization of distributional equivalence on this setting, introducing new instruments—particularly edge rank constraints—to explain when two fashions generate the identical noticed information. Constructing on this principle, the authors derive sensible graphical standards and transformations to enumerate all equal fashions and suggest an algorithm to recuperate your entire equivalence class from information. Total, the work removes the necessity for robust structural assumptions and presents a common, principled framework for latent-variable causal discovery.

Revela: Dense Retriever Studying through Language Modeling

Authors: Fengyu Cai (Technische Universität Darmstadt), Tong Chen (College of Washington), Xinran Zhao (Carnegie Mellon College), Sihao Chen (Microsoft), Hongming Zhang (Tencent AI Lab Seattle), Sherry Wu (Carnegie Mellon College), Iryna Gurevych (Technical College of Darmstadt / Mohamed bin Zayed College of Synthetic Intelligence), Heinz Koeppl (TU Darmstadt)

This paper introduces Revela, a self-supervised framework for coaching dense retrievers by leveraging language modeling goals as a substitute of counting on annotated query-document pairs. It augments next-token prediction with an in-batch consideration mechanism that enables paperwork to attend to one another, enabling the retriever to be taught cross-document relationships collectively with a language mannequin. Experiments throughout domain-specific, reasoning-intensive, and common benchmarks present that Revela matches or surpasses supervised and API-based retrievers whereas utilizing considerably much less information and compute. Total, the work demonstrates a scalable and environment friendly different for retriever studying instantly from uncooked textual content with robust generalization throughout domains.

Latent Particle World Fashions: Self-supervised Object-centric Stochastic Dynamics Modeling

Authors: Tal Daniel (Carnegie Mellon College), Carl Qi (College of Texas at Austin), Dan Haramati (Brown College), Amir Zadeh (Lambda), Chuan Li (Lambda Labs), Aviv Tamar (Technion), Deepak Pathak (Carnegie Mellon College), David Held (Carnegie Mellon College)

This paper introduces the Latent Particle World Mannequin (LPWM), a self-supervised, object-centric world mannequin that learns to decompose scenes into latent particles (e.g., keypoints, masks, and object attributes) instantly from uncooked video with out supervision. It proposes a novel per-particle latent motion mechanism that fashions stochastic dynamics, enabling the system to seize advanced multi-object interactions and generate numerous future predictions. The mannequin is skilled end-to-end and helps versatile conditioning on actions, language, and purpose photos, reaching state-of-the-art efficiency on each real-world and artificial video prediction duties. Past video modeling, LPWM additionally demonstrates robust potential for decision-making purposes similar to imitation studying by leveraging its realized latent dynamics.

LoongRL: Reinforcement Studying for Superior Reasoning over Lengthy Contexts

Authors: Siyuan Wang (Shanghai Jiao Tong College), Gaokai Zhang (Carnegie Mellon College), Li Lyna Zhang (Microsoft Analysis Asia), Ning Shang (Microsoft), Fan Yang (Microsoft Analysis), Dongyao Chen (Shanghai Jiaotong College), Mao Yang (Peking College)

The authors introduce LoongRL, a reinforcement studying framework designed to enhance long-context reasoning in giant language fashions by coaching them on difficult, synthesized duties. They suggest KeyChain, an information development technique that embeds hidden query chains inside lengthy paperwork, forcing fashions to carry out multi-step planning, retrieval, and reasoning reasonably than counting on shortcuts. By means of RL coaching, fashions develop an emergent “plan–retrieve–motive–recheck” reasoning sample that generalizes from shorter (16K) to for much longer (128K) contexts. Experiments present that LoongRL considerably boosts long-context reasoning efficiency whereas sustaining robust short-context skills, reaching outcomes corresponding to a lot bigger fashions.

Exchangeability of GNN Representations with Purposes to Graph Retrieval

Authors: Kartik Nair (Carnegie Mellon College), Indradyumna Roy (IIT Bombay, Aalto College), Soumen Chakrabarti (IIT Bombay), Anirban Dasgupta (IIT Gandhinagar), Abir De (Indian Institute of Know-how Bombay)

This paper introduces the idea of exchangeability in graph neural networks (GNNs), displaying that the size of realized node embeddings are statistically interchangeable because of random initialization and permutation-invariant coaching. This property implies that embedding parts share equivalent distributions, enabling simplifications in how graph similarities are computed. Leveraging this perception, the authors approximate advanced transportation-based graph distances utilizing easier Euclidean operations on sorted embedding values. They additional suggest GRAPHHASH, a locality-sensitive hashing framework that permits environment friendly and scalable graph retrieval, reaching robust efficiency in comparison with present strategies.

TusoAI: Agentic Optimization for Scientific Strategies

Authors: Alistair Turcan (College of Laptop Science, Carnegie Mellon College), Kexin Huang (Stanford College), Lei Li (College of Laptop Science, Carnegie Mellon College), Martin J. Zhang (Carnegie Mellon College)

Vlaser: Imaginative and prescient-Language-Motion Mannequin with Synergistic Embodied Reasoning

Authors: Ganlin Yang (College of Science and Know-how of China), Tianyi Zhang (Zhejiang College; Shanghai Synthetic Intelligence Laboratory), Haoran Hao (Carnegie Mellon College), Weiyun Wang (Fudan College), Yibin Liu (Northeastern College), Dehui Wang (Shanghai Jiaotong College), Guanzhou Chen (Shanghai AI Laboratory, Shanghai Jiaotong College), Zijian Cai (Shenzhen College), Junting Chen (nationwide college of singaore, Nationwide College of Singapore), Weijie Su (College of Science and Know-how of China), Wengang Zhou (College of Science and Know-how of China), Yu Qiao (Shanghai Aritifcal Intelligence Laboratory), Jifeng Dai (Tsinghua College, Tsinghua College), Jiangmiao Pang (Shanghai AI Laboratory), Gen Luo (Shanghai AI Laboratory), Wenhai Wang (Shanghai AI Laboratory), Yao Mu (Shanghai Jiao Tong College), Zhi Hou (Shanghai Synthetic Intelligence Laboratory)

MetaVLA: Unified Meta Co-Coaching for Environment friendly Embodied Adaptation

Authors: Chen Li (Carnegie Mellon College), Zhantao Yang (Carnegie Mellon College), Han Zhang (Carnegie Mellon College), Fangyi Chen (ByteDance Inc.), Chenchen Zhu (Meta AI), Anudeepsekhar Bolimera (Carnegie Mellon College), Marios Savvides (Carnegie Mellon College)

RobotArena $infty$: Scalable Robotic Benchmarking through Actual-to-Sim Translation

Authors: Yash Jangir (Carnegie Mellon College), Yidi Zhang (), Kashu Yamazaki (CMU, Carnegie Mellon College), Chenyu Zhang (Peking College), Kuan-Hsun Tu (Nationwide Taiwan College), Tsung-Wei Ke (Division of pc science and informational engineering, Nationwide Taiwan College), Lei Ke (Carnegie Mellon College), Yonatan Bisk (Carnegie Mellon College), Katerina Fragkiadaki (CMU)

Generalizable Finish-to-Finish Device-Use RL with Artificial CodeGym

Authors: Weihua Du (Tsinghua College), HaileiGong (Huawei Applied sciences Ltd.), Zhan Ling (UC San Diego), Kang Liu (ByteDance Inc.), Lingfeng Shen (Johns Hopkins College), Xuesong Yao (ByteDance Inc.), Yufei Xu (ByteDance Inc.), Dingyuan Shi (ByteDance Inc.), Yiming Yang (Carnegie Mellon College), Jiecao Chen (ByteDance Inc.)

WearVox: An Selfish Multichannel Voice Assistant Benchmark for Wearables

Authors: Zhaojiang Lin (Meta), YONG XU (Meta), Kai Solar (Meta), Jing Zheng (Ant Group), Yin Huang (Fb), Surya Appini (Meta), Krish Narang (Fb), Renjie Tao (Fb), Ishan Jain (Fb), Siddhant Arora (Carnegie Mellon College), Ruizhi Li (Fb), Yiteng Huang (Fb), Kaushik Patnaik (Apple), Wenfang Xu (Meta Platforms, Inc.), Suwon Shon (ASAPP), Yue Liu (Meta), Ahmed Aly (Fb), Anuj Kumar (Meta), Florian Metze (Carnegie Mellon College), Xin Dong (Fb)

Evaluating AI Brokers to Cybersecurity Professionals in Actual-World Penetration Testing

Authors: Justin Lin (Laptop Science Division, Stanford College), Eliot Jones (Grey Swan), Donovan Jasper (Stanford College), Ethan Ho (Stanford College), Anna Wu (Laptop Science Division, Stanford College), Arnold Yang (Stanford College), Neil Perry (Princeton College), Andy Zou (CMU, Carnegie Mellon College), Matt Fredrikson (College of Wisconsin, Madison), Zico Kolter (Carnegie Mellon College), Percy Liang (Stanford College), Dan Boneh (Stanford College), Daniel Ho (Stanford College)

Certain by semanticity: common legal guidelines governing the generalization-identification tradeoff

Authors: Marco Nurisso (Polytechnic College of Turin), Jesseba Fernando (Northeastern College), Raj Deshpande (Northeastern College London), Alan Perotti (Intesa Sanpaolo AI Analysis), Raja Marjieh (Princeton College), Steven Frankland (Dartmouth Faculty), Richard Lewis (Carnegie Mellon College), Taylor Webb (College of California, Los Angeles), Declan Campbell (Princeton College), Francesco Vaccarino (Politecnico di Torino), Jonathan Cohen (Princeton College), Giovanni Petri (Community Science Institute, Northeastern College London)

Zero-shot Forecasting by Simulation Alone

Authors: Boris Oreshkin (Amazon), Mayank Jauhari (Amazon), Ravi Kiran Selvam (Amazon), Malcolm Wolff (Amazon), Wenhao Pan (College of Washington), Shankar Ramasubramanian (Amazon), KIN GUTIERREZ (Carnegie Mellon College), Tatiana Konstantinova (Amazon), Andres Potapczynski (New York College), Mengfei Cao (Amazon.com), Dmitry Efimov (Amazon), Michael W Mahoney (College of California Berkeley), Andrew Gordon Wilson (New York College)

Self-Bettering Imaginative and prescient-Language-Motion Fashions with Knowledge Technology through Residual RL

Authors: Wenli Xiao (Carnegie Mellon College), Haotian Lin (CMU, Carnegie Mellon College), Andy Peng (College of California, Berkeley), Haoru Xue (College of California, Berkeley), Tairan He (NVIDIA), Zhengyi Luo (Carnegie Mellon College), Yuqi Xie (NVIDIA), Fengyuan Hu (NVIDIA), Jim Fan (NVIDIA), Guanya Shi (CMU, Carnegie Mellon College), Yuke Zhu (NVIDIA / UT-Austin)

Bettering Attributed Lengthy-form Query Answering with Intent Consciousness

Authors: Xinran Zhao (CMU, Carnegie Mellon College), Aakanksha Naik (Allen Institute for Synthetic Intelligence), Jay DeYoung (Allen Institute for Synthetic Intelligence), Joseph Chee Chang (Allen Institute for Synthetic Intelligence), Jena Hwang (Allen Institute for Synthetic Intelligence), Sherry Wu (Carnegie Mellon College), Varsha Kishore (Cornell College)

BFM-Zero: A Promptable Behavioral Basis Mannequin for Humanoid Management Utilizing Unsupervised Reinforcement Studying

Authors: Yitang Li (), Zhengyi Luo (Carnegie Mellon College), Tonghe Zhang (Carnegie Mellon College), Cunxi Dai (Carnegie Mellon College), Anssi Kanervisto (Microsoft Analysis), Andrea Tirinzoni (Meta, FAIR), Haoyang Weng (Tsinghua College, Tsinghua College), Kris Kitani (Carnegie Mellon College), Mateusz Guzek (Meta AI), Ahmed Touati (Meta AI Analysis), Alessandro Lazaric (Fb), Matteo Pirotta (Meta), Guanya Shi (CMU, Carnegie Mellon College)

Actual-Time Reasoning Brokers in Evolving Environments

Authors: Yuletide Wen (Tsinghua College, Tsinghua College), Yixin Ye (Shanghai Jiaotong College), Yanzhe Zhang (Georgia Institute of Know-how), Diyi Yang (Stanford College), Hao Zhu (Carnegie Mellon College)

ExpertLongBench: Benchmarking Language Fashions on Skilled-Stage Lengthy-Kind Technology Duties with Structured Checklists

Authors: Jie Ruan (College of Michigan – Ann Arbor), Inderjeet Nair (College of Michigan – Ann Arbor), Shuyang Cao (Bloomberg), Amy Liu (College of Michigan), Sheza Munir (College of Toronto), Micah Pollens-Dempsey (College of Michigan – Ann Arbor), Yune-Ting Chiang (College of Michigan – Ann Arbor), Lucy Kates (College of Michigan – Ann Arbor), Nicholas David (College of Michigan – Ann Arbor), Sihan Chen (Carnegie Mellon College), Ruxin Yang (College of Michigan – Ann Arbor), Yuqian Yang (College of Michigan – Ann Arbor), Jihyun Gump (College of Michigan – Ann Arbor), Tessa Bialek (College of Michigan Regulation College), Vivek Sankaran (College of Michigan – Ann Arbor), Margo Schlanger (College of Michigan – Ann Arbor), Lu Wang (College of Michigan)

ViPRA: Video Prediction for Robotic Actions

Authors: Sandeep Kumar Routray (Skild AI), Hengkai Pan (CMU, Carnegie Mellon College), Unnat Jain (Fb AI Analysis), Shikhar Bahl (Skild AI), Deepak Pathak (Carnegie Mellon College)

Contact-guided Real2Sim from Monocular Video with Planar Scene Primitives

Authors: Zihan Wang (Amazon), Jiashun Wang (College of Laptop Science, Carnegie Mellon College), Jeff Tan (Carnegie Mellon College), Yiwen Zhao (College of Laptop Science, Carnegie Mellon College), Jessica Hodgins (RAI Institute), Shubham Tulsiani (Carnegie Mellon College), Deva Ramanan (College of Laptop Science, Carnegie Mellon College)

The Device Decathlon: Benchmarking Language Brokers for Numerous, Real looking, and Lengthy-Horizon Job Execution

Authors: Junlong Li (The Hong Kong College of Science and Know-how), Wenshuo Zhao (Zhejiang College), Jian Zhao (Beijing College of Posts and Telecommunications), Weihao Zeng (Hong Kong College of Science and Know-how), Haoze Wu (Zhejiang College), Xiaochen Wang (None), Rui Ge (Shanghai Jiaotong College), Yuxuan Cao (HKUST), Yuzhen Huang (HKUST), Wei Liu (HKUST), Junteng LIU (HKUST), Zhaochen Su (The Hong Kong College of Science and Know-how), Yiyang Guo (Fudan College), FAN ZHOU (Shanghai Jiao Tong College), Lueyang Zhang (The Hong Kong College of Science and Know-how), Juan Michelini (Universidad de la República), Xingyao Wang (All Palms AI), Xiang Yue (Carnegie Mellon College), Shuyan Zhou (Fb), Graham Neubig (Carnegie Mellon College), Junxian He (HKUST)

SAC Stream: Pattern-Environment friendly Reinforcement Studying of Stream-Primarily based Insurance policies through Velocity-Reparameterized Sequential Modeling

Authors: Yixian Zhang (Tsinghua College, Tsinghua College), Shu-ang Yu (Tsinghua College), Tonghe Zhang (Carnegie Mellon College), Mo Guang (Li Auto Inc.), Haojia Hui (Li Auto Inc.), Kaiwen Lengthy (Li Auto Inc.), Yu Wang (Tsinghua Univ.), Chao Yu (Tsinghua College), Wenbo Ding (Tsinghua College, Tsinghua College)

Multi-Object System Identification from Movies

Authors: Chunjiang Liu (Carnegie Mellon College), Xiaoyuan Wang (Carnegie Mellon College), Qingran Lin (Georgia Institute of Know-how), Albert Xiao (Carnegie Mellon College), Haoyu Chen (Harvard College, Harvard College), Shizheng Wen (ETHZ – ETH Zurich), Hao Zhang (UIUC), Lu Qi (Insta360), Ming-Hsuan Yang (Google DeepMind), Laszlo A. Jeni (Carnegie Mellon College), Min Xu (Carnegie Mellon College), Yizhou Zhao (Snap Inc.)

Studying an Picture Enhancing Mannequin with out Picture Enhancing Pairs

Authors: Nupur Kumari (Carnegie Mellon College), Sheng-Yu Wang (CMU, Carnegie Mellon College), Cherry Zhao (Adobe Analysis), Yotam Nitzan (Adobe Analysis), Yuheng Li (Adobe Methods), Krishna Kumar Singh (Adobe Methods), Richard Zhang (Adobe), Eli Shechtman (Adobe), Jun-Yan Zhu (Carnegie Mellon College), Xun Huang (Adobe Analysis)

Controllable Video Technology with Provable Disentanglement

Authors: Yifan Shen (Mohamed bin Zayed College of Synthetic Intelligence), Peiyuan Zhu (Mohamed bin Zayed College of Synthetic Intelligence), Zijian Li (Mohamed bin Zayed College of Synthetic Intelligence), Shaoan Xie (Carnegie Mellon College), Namrata Deka (Carnegie Mellon College), Zongfang Liu (Zhejiang College), Zeyu Tang (Stanford College), Guangyi Chen (MBZUAI&CMU), Kun Zhang (Carnegie Mellon College & MBZUAI)

Digital Neighborhood: An Open World for People, Robots, and Society

Authors: Qinhong Zhou (College of Massachusetts at Amherst), Hongxin Zhang (UMass Amherst), Xiangye Lin (College of Massachusetts at Amherst), Zheyuan Zhang (Johns Hopkins College), Yutian Chen (Carnegie Mellon College), Wenjun Liu (College of Massachusetts at Amherst), Zunzhe Zhang (Tsinghua College), Sunli Chen (College of Massachusetts at Amherst), Lixing Fang (College of Massachusetts at Amherst), Qiushi Lyu (College of Illinois, Urbana-Champaign), Xinyu Solar (South China College of Know-how), Jincheng Yang (College of Maryland, Faculty Park), Zeyuan Wang (Tsinghua College, Tsinghua College), Bao Dang (College of Massachusetts at Amherst), Zhehuan Chen (Peking College), Daksha Ladia (College of Massachusetts Amherst), Quang Dang (College of Massachusetts at Amherst), Jiageng Liu (College of Massachusetts at Amherst), Chuang Gan (MIT-IBM Watson AI Lab)

Sooner Imaginative and prescient Transformers with Adaptive Patches

Authors: Rohan Choudhury (None), JungEun Kim (Normal Robotics), Jinhyung Park (Carnegie Mellon College), Eunho Yang (Korea Superior Institute of Science & Know-how), Laszlo A. Jeni (Carnegie Mellon College), Kris Kitani (Carnegie Mellon College)

VINCIE: Unlocking In-context Picture Enhancing from Video

Authors: Leigang Qu (Nationwide College of Singapore), Feng Cheng (ByteDance Seed), Ziyan Yang (ByteDance Inc.), Qi Zhao (ByteDance Inc.), Shanchuan Lin (ByteDance), Yichun Shi (None), Yicong Li (Nationwide College of Singapore), Wenjie Wang (College of Science and Know-how of China), Tat-Seng Chua (Nationwide College of Singapore), Lu Jiang (Carnegie Mellon College)

lmgame-Bench: How Good are LLMs at Enjoying Video games?

Authors: Lanxiang Hu (College of California, San Diego), Mingjia Huo (College of California, San Diego), Yuxuan Zhang (College of California, San Diego), Haoyang Yu (College of California San Diego), Eric P Xing (CMU), Ion Stoica (), Tajana Rosing (College of California, San Diego), Haojian Jin (None), Hao Zhang (College of California, San Diego)

SpineBench: A Clinically Salient, Stage-Conscious Benchmark Powered by the SpineMed-450k Corpus

Authors: Ming Zhao (Jilin College), Wenhui Dong (NanJing College), Yang Zhang (Chinese language Individuals’s Liberation Military Normal Hospital), wangyou (College of the Chinese language Academy of Sciences), Zhonghao Zhang (Ningxia College), Zian Zhou (Zhejiang College), YUNZHI GUAN (Fudan College), Liukun Xu (Nanjing Medical College), Wei Peng (Stanford College), Zhaoyang Gong (Fudan College), Zhicheng Zhang (Chinese language Individuals’s Liberation Military Normal Hospital), Dachuan li (Fudan College), Xiaosheng Ma (Fudan College), Yuli Ma (Peking College), Jianing Ni (Carnegie Mellon College), Changjiang Jiang (Ant Group), Lixia Tian (Beijing Jiaotong College), Chen Qixin (Zhejiang College), Xia Kaishun (Zhejiang College of Know-how), Pingping Liu (Jilin College), Tongshun Zhang (Jilin College), ZhiqiangLiu (Huazhong College of Science and Know-how), Zhongan Bi (Zhejiang Lab), Chenyang Si (Nanyang Technological College), Tiansheng Solar (Chinese language Individuals’s Liberation Military Normal Hospital), Caifeng Shan (Nanjing College)

SeedVR2: One-Step Video Restoration through Diffusion Adversarial Publish-Coaching

Authors: Jianyi Wang (Nanyang Technological College), Shanchuan Lin (ByteDance), Zhijie Lin (Zhejiang College), Yuxi Ren (ByteDance Inc.), Meng Wei (ByteDance Inc.), Zongsheng Yue (Xi’an Jiaotong College), Shangchen Zhou (Nanyang Technological College), Hao Chen (ByteDance Inc.), Yang Zhao (Bytedance Inc.), Ceyuan Yang (ByteDance), Xuefeng Xiao (ByteDance), Chen Change Loy (Nanyang Technological College), Lu Jiang (Carnegie Mellon College)

Combination of Contexts for Lengthy Video Technology

Authors: Shengqu Cai (Stanford College), Ceyuan Yang (ByteDance), Lvmin Zhang (Stanford College), Yuwei Guo (The Chinese language College of Hong Kong), Junfei Xiao (Johns Hopkins College), Ziyan Yang (ByteDance Inc.), Yinghao Xu (Stanford College), Zhenheng Yang (Tiktok), Alan Yuille (Johns Hopkins College), Leonidas Guibas (Stanford College), Maneesh Agrawala (Stanford College), Lu Jiang (Carnegie Mellon College), Gordon Wetzstein (Stanford College)

pySpatial: Producing 3D Visible Applications for Zero-Shot Spatial Reasoning

Authors: Zhanpeng Luo (College of Pittsburgh), Ce Zhang (Carnegie Mellon College), Silong Yong (Division of Automation, Tsinghua College, Tsinghua College), Cunxi Dai (Carnegie Mellon College), Qianwei Wang (College of Michigan – Ann Arbor), Haoxi Ran (Carnegie Mellon College), Guanya Shi (CMU, Carnegie Mellon College), Katia Sycara (Carnegie Mellon College), Yaqi Xie (Carnegie Mellon College)

Sharp Monocular View Synthesis in Much less Than a Second

Authors: Lars Mescheder (Apple), Wei Dong (Apple), Shiwei Li (Apple), Xuyang BAI (Apple), Marcel Santos (Apple), Peiyun Hu (Carnegie Mellon College), Bruno Lecouat (Telecom ParisTech), Mingmin Zhen (Apple), Amaël Delaunoy (Apple), Tian Fang (Hong Kong College of Science and Know-how), Yanghai Tsin (Apple), Stephan Richter (Apple), Vladlen Koltun (Apple)

S2GO: Streaming Sparse Gaussian Occupancy

Authors: Jinhyung Park (Carnegie Mellon College), Chensheng Peng (College of California, Berkeley), yihan hu (Utilized Instinct), Wenzhao Zheng (UC Berkeley), Kris Kitani (Carnegie Mellon College), Wei Zhan (College of California Berkeley)

Captain Cinema: In direction of Brief Film Technology

Authors: Junfei Xiao (Johns Hopkins College), Ceyuan Yang (ByteDance), Lvmin Zhang (Stanford College), Shengqu Cai (Stanford College), Yang Zhao (Bytedance Inc.), Yuwei Guo (The Chinese language College of Hong Kong), Gordon Wetzstein (Stanford College), Maneesh Agrawala (Stanford College), Alan Yuille (Johns Hopkins College), Lu Jiang (Carnegie Mellon College)

VisCoder2: Constructing Multi-Language Visualization Coding Brokers

Authors: Yuansheng Ni (College of Waterloo), Songcheng Cai (College of Waterloo), Xiangchao Chen (College of Waterloo), Jiarong Liang (College of Waterloo), Zhiheng LYU (College of Hong Kong), Jiaqi Deng (Korea Superior Institute of Science & Know-how), Kai Zou (NetMind.AI), PING NIE (Peking College), Fei Yuan (Shanghai Synthetic Clever Laboratory), Xiang Yue (Carnegie Mellon College), Wenhu Chen (College of Waterloo)

e3: Studying to Discover Allows Extrapolation of Check-Time Compute for LLMs

Authors: Amrith Setlur (Carnegie Mellon College), Matthew Yang (Carnegie Mellon College), Charlie Snell (College of California, Berkeley), Jeremiah Greer (Oumi AI PBC), Ian Wu (Carnegie Mellon College), Virginia Smith (Carnegie Mellon College), Max Simchowitz (Massachusetts Institute of Know-how), Aviral Kumar (College of California Berkeley)

Log-Linear Consideration

Authors: Guo (), Songlin Yang (ShanghaiTech College), Tarushii Goel (Massachusetts Institute of Know-how), Eric P Xing (CMU), Tri Dao (Princeton College), Yoon Kim (MIT)

Generalized Parallel Scaling with Interdependent Generations

Authors: Harry Dong (Carnegie Mellon College), David Brandfonbrener (NYU), Eryk Helenowski (Fb), Yun He (Meta), Mrinal Kumar (Fb), Han Fang (Meta GenAI), Yuejie Chi (Carnegie Mellon College), Karthik Abinav Sankararaman (Fb)

On Code-Induced Reasoning in LLMs

Authors: Abdul Waheed (Maharaja Agrasen Institute of Know-how, New Delhi), Zhen Wu (Carnegie Mellon College), Carolyn Rose (College of Laptop Science, Carnegie Mellon College), Daphne Ippolito (College of Engineering and Utilized Science, College of Pennsylvania)

A number of-Prediction-Powered Inference

Authors: Charlie Cowen-Breen (Massachusetts Institute of Know-how), Alekh Agarwal (Google), Stephen Bates (Massachusetts Institute of Know-how), William W. Cohen (Carnegie Mellon College), Jacob Eisenstein (Google), Amir Globerson (Google), Adam Fisch (Google DeepMind)

Command-V: Coaching-Free Illustration Finetuning Switch

Authors: Barry Wang (Carnegie Mellon College), Avi Schwarzschild (Carnegie Mellon College), Alexander Robey (CMU, Carnegie Mellon College), Ali Payani (Cisco Methods), Charles Fleming (Cisco), Mingjie Solar (College of Laptop Science, Carnegie Mellon College), Daphne Ippolito (College of Engineering and Utilized Science, College of Pennsylvania)

Immediate-MII: Meta-Studying Instruction Induction for LLMs

Authors: Emily Xiao (Carnegie Mellon College), Yixiao Zeng (XPeng Motors / Carnegie Mellon College), Ada Chen (CMU, Carnegie Mellon College), Chin-Jou Li (Language Applied sciences Institute, Carnegie Mellon College), Amanda Bertsch (Carnegie Mellon College), Graham Neubig (Carnegie Mellon College)

HARDTESTGEN: A Excessive-High quality RL Verifier Technology Pipeline for LLM Algorithimic Coding

Authors: Zhongmou He (Carnegie Mellon College), Yee Man Choi (College of Waterloo), Kexun Zhang (Carnegie Mellon College), Ivan Bercovich (UC Santa Barbara + ScOp VC), Jiabao Ji (College of California, Santa Barbara), Junting Zhou (Peking College), Dejia Xu (College of Texas at Austin), Aidan Zhang (Carnegie Mellon College), Yixiao Zeng (XPeng Motors / Carnegie Mellon College), Lei Li (College of Laptop Science, Carnegie Mellon College)

Reevaluating Coverage Gradient Strategies for Imperfect-Info Video games

Authors: Max Rudolph (College of Texas at Austin), Nathan Lichtlé (Electrical Engineering & Laptop Science Division, College of California, Berkeley), Sobhan Mohammadpour (MIT), Alexandre M Bayen (None), Zico Kolter (Carnegie Mellon College), Amy Zhang (UT Austin), Gabriele Farina (Massachusetts Institute of Know-how), Eugene Vinitsky (New York College), Samuel Sokota (Carnegie Mellon College)

GEM: A Gymnasium for Generalist LLMs

Authors: Zichen Liu (Sea AI Lab), Anya Sims (College of Oxford), Keyu Duan (nationwide college of singaore, Nationwide College of Singapore), Changyu Chen (Stanford College), Simon Yu (Northeastern College), Xiangxin Zhou (UCAS), Haotian Xu (Tsinghua College, Tsinghua College), Shaopan Xiong (Alibaba Group), Bo Liu (Nationwide College of Singapore), Chenmien Tan (College of Edinburgh), Weixun Wang (Tianjin College), Hao Zhu (Carnegie Mellon College), Weiyan Shi (Columbia College), Diyi Yang (Stanford College), Michael Qizhe Shieh (Nationwide College of Singapore), Yee Whye Teh (College of Oxford and Google DeepMind), Wee Solar Lee (Nationwide College of Singapore), Min Lin (Sea AI Lab)

BEAT: Visible Backdoor Assaults on VLM-based Embodied Brokers through Contrastive Set off Studying

Authors: Qiusi Zhan (College of Illinois Urbana-Champaign), Hyeonjeong Ha (College of Illinois Urbana-Champaign), Rui Yang (Hong Kong College of Science and Know-how), Sirui Xu (College of Illinois at Urbana-Champaign), Hanyang Chen (College of Illinois at Urbana-Champaign), Liang-Yan Gui (UIUC), Yu-Xiong Wang (UIUC), Huan Zhang (CMU), Heng Ji (College of Illinois at Urbana-Champaign), Daniel Kang (UIUC)

VLSU: Mapping the Limits of Joint Multimodal Understanding for AI Security

Authors: Shruti Palaskar (Apple), Leon Gatys (Apple), Mona Abdelrahman (Apple), Mar Jacobo (Apple), Laurence Lindsey (Apple), Rutika Moharir (Apple), Gunnar Lund (Grammarly), Yang Xu (Apple), Navid Shiee (Apple), Jeffrey Bigham (Carnegie Mellon College), Charles Maalouf (Apple), Joseph Cheng (Apple)

Generative Worth Conflicts Reveal LLM Priorities

Authors: Andy Liu (Carnegie Mellon College), Kshitish Ghate (College of Washington), Mona Diab (Carnegie Mellon College), Daniel Fried (Carnegie Mellon College), Atoosa Kasirzadeh (Alan Turing Institute), Max Kleiman-Weiner (Frequent Sense Machines)

PluriHarms: Benchmarking the Full Spectrum of Human Judgments on AI Hurt

Authors: Jing-Jing Li (College of California, Berkeley), Joel Mire (Carnegie Mellon College), Eve Fleisig (UC Berkeley), Valentina Pyatkin (Ai2, ETH AI Heart), Anne Collins (College of California, Berkeley), Maarten Sap (Carnegie Mellon College), Sydney Levine (NYU / Google Deepmind)

Spectrum Tuning: Publish-Coaching for Distributional Protection and In-Context Steerability

Authors: Taylor Sorensen (people&), Benjamin Newman (College of Washington), Jared Moore (Laptop Science Division, Stanford College), Chan Younger Park (College of Texas at Austin), Jillian Fisher (College of Washington), Niloofar Mireshghallah (Carnegie Mellon College), Liwei Jiang (None), Yejin Choi (Stanford College / NVIDIA)

Convergence of Remorse Matching in Potential Video games and Constrained Optimization

Authors: Ioannis Anagnostides (Carnegie Mellon College), Emanuel Tewolde (Carnegie Mellon College), Brian Zhang (MIT), Ioannis Panageas (Donald Bren College of Info and Laptop Sciences, College of California, Irvine), Vincent Conitzer (Carnegie Mellon College), Tuomas Sandholm (Carnegie Mellon College)

Pattern Complexity and Illustration Means of Check-time Scaling Paradigms

Authors: Baihe Huang (College of California, Berkeley), Shanda Li (Carnegie Mellon College), Tianhao Wu (College of California, Berkeley), Yiming Yang (Carnegie Mellon College), Ameet Talwalkar (College of California-Los Angeles), Kannan Ramchandran (), Michael Jordan (College of California, Berkeley), Jiantao Jiao (College of California Berkeley)

Polynomial Convergence of Riemannian Diffusion Fashions

Authors: Xingyu Xu (CMU, Carnegie Mellon College), Ziyi Zhang (CMU, Carnegie Mellon College), Yorie Nakahira (Researcher at NII LLM Heart Assistant Professor at CMU), Guannan Qu (Carnegie Mellon College), Yuejie Chi (Carnegie Mellon College)

Studying-Augmented Second Estimation on Time-Decay Fashions

Authors: Soham Nagawanshi (Texas A&M College – Faculty Station), Shalini Panthangi (CMU, Carnegie Mellon College), Chen Wang (Rice College and Texas A&M College), David Woodruff (Carnegie Mellon College), Samson Zhou (Texas A&M College)

RLP: Reinforcement as a Pretraining Goal

Authors: Ali Hatamizadeh (Nvidia), Syeda Nahida Akter (Carnegie Mellon College), Shrimai Prabhumoye (NVIDIA), Jan Kautz (NVIDIA), Mostofa Patwary (NVIDIA), Mohammad Shoeybi (NVIDIA), Bryan Catanzaro (NVIDIA), Yejin Choi (Stanford College / NVIDIA)

Assume Then Embed: Generative Context Improves Multimodal Embedding

Authors: Xuanming Cui (College of Central Florida), Jianpeng Cheng (Meta), Hong-You Chen (Ohio State College), Satya Narayan Shukla (Meta), Abhijeet Awasthi (Indian Institute of Know-how Bombay), Xichen Pan (New York College), Chaitanya Ahuja (Carnegie Mellon College), Shlok Mishra (Fb), Taipeng Tian (Meta), Qi Guo (Fb), Ser-Nam Lim (College of Central Florida), Aashu Singh (Fb), Xiangjun Fan (Meta)

MANZANO: A Easy and Scalable Unified Multimodal Mannequin with a Hybrid Imaginative and prescient Tokenizer

Authors: Yanghao Li (Apple), Rui Qian (Apple), Bowen Pan (Massachusetts Institute of Know-how), Haotian Zhang (NVIDIA), Haoshuo Huang (Apple), Bowen Zhang (Apple), Jialing Tong (Apple), Haoxuan You (Apple AI/ML), Xianzhi Du (Apple), Zhe Gan (Apple), Hyunjik Kim (DeepMind), Chao Jia (Google), Zhenbang Wang (Apple), Yinfei Yang (Apple), Mingfei Gao (Apple), Zi-Yi Dou (Carnegie Mellon College), Wenze Hu (UCLA, College of California, Los Angeles), Chang Gao (Waymo), Dongxu Li (SalesForce.com), Philipp Dufter (Apple), Zirui Wang (Apple AI/ML), Guoli Yin (Apple), Zhengdong Zhang (Google), Chen Chen (Apple), Yang Zhao (College of California, Berkeley), Ruoming Pang (None), Zhifeng Chen (Apple)

TrustGen: A Platform of Dynamic Benchmarking on the Trustworthiness of Generative Basis Fashions

Authors: Yue Huang (College of Notre Dame), Chujie Gao (Mohamed bin Zayed College of Synthetic Intelligence), Siyuan Wu (None), Haoran Wang (Emory College), Xiangqi Wang (College of Notre Dame), Jiayi Ye (Sichuan College), Yujun Zhou (College of Notre Dame), Yanbo Wang (Mohamed bin Zayed College of Synthetic Intelligence), Jiawen Shi (Huazhong College of Science and Know-how), Qihui Zhang (Sichuan College), Han Bao (College of Notre Dame), Zhaoyi Liu (College of Illinois at Urbana-Champaign), Yuan Li (College of Cambridge), Tianrui Guan (Division of Laptop Science, College of Maryland, Faculty Park), Peiran Wang (College of California, Los Angeles), Haomin Zhuang (College of Notre Dame), Dongping Chen (College of Washington), Kehan Guo (College of Notre Dame), Andy Zou (CMU, Carnegie Mellon College), Bryan Hooi (Nationwide College of Singapore), Caiming Xiong (Salesforce Analysis), Elias Stengel-Eskin (Division of Laptop Science, UT Austin), Hongyang Zhang (College of Waterloo), Hongzhi Yin (College of Queensland), Huan Zhang (CMU), Huaxiu Yao (UNC-Chapel Hill), Jieyu Zhang (Division of Laptop Science, College of Washington), Jaehong Yoon (NTU Singapore), Kai Shu (Emory College), Ranjay Krishna (Division of Laptop Science), Swabha Swayamdipta (College of Southern California), Weijia Shi (College of Washington, Seattle), Xiang Li (Massachusetts Normal Hospital), Yuexing Hao (Massachusetts Institute of Know-how), Zhihao Jia (College of Laptop Science, Carnegie Mellon College), Zhize Li (KAUST), Xiuying Chen (Mohamed bin Zayed College of Synthetic Intelligence), Zhengzhong Tu (Texas A&M College – Faculty Station), Xiyang Hu (Arizona State College), Tianyi Zhou (MBZUAI), Jieyu Zhao (College of Southern California), Lichao Solar (Lehigh College), Furong Huang (College of Maryland), Or Cohen-Sasson (College of Miami), Prasanna Sattigeri (IBM Analysis), Anka Reuel (Stanford College), Max Lamparth (Stanford College), Yue Zhao (College of Southern California), Nouha Dziri (Allen Institute for AI), Yu Su (Ohio State College), Huan Solar (Ohio State College), Heng Ji (College of Illinois at Urbana-Champaign), Chaowei Xiao (Johns Hopkins College/NVIDIA), Mohit Bansal (UNC Chapel Hill), Nitesh Chawla (College of Notre Dame), Jian Pei (Simon Fraser College), Jianfeng Gao (Microsoft Analysis), Michael Backes (CISPA Helmholtz Heart for Info Safety), Philip Yu (College of Illinois, Chicago), Neil Gong (), Pin-Yu Chen (IBM Analysis AI), Bo Li (College of Illinois, Urbana Champaign), Daybreak Track (Berkeley), Xiangliang Zhang (College of Notre Dame)

RefineBench: Evaluating Refinement Functionality of Language Fashions through Checklists

Authors: Younger-Jun Lee (KAIST), Seungone Kim (Carnegie Mellon College), Byung-Kwan Lee (NVIDIA), Minkyeong Moon (Yonsei College), Yechan Hwang (), Jong Myoung Kim (Korea Superior Institute of Science & Know-how), Graham Neubig (Carnegie Mellon College), Sean Welleck (Carnegie Mellon College), Ho-Jin Choi (Korea Superior Institute of Science & Know-how)

Scaling Group Inference for Numerous and Excessive-High quality Technology

Authors: Gaurav Parmar (Carnegie Mellon College), Or Patashnik (Tel Aviv College), Daniil Ostashev (Snap Inc.), Kuan-Chieh Wang (Snap Inc.), Kfir Aberman (Google), Srinivasa Narasimhan (Carnegie Mellon College), Jun-Yan Zhu (Carnegie Mellon College)

A lot Ado About Noising: Dispelling the Myths of Generative Robotic Management

Authors: Chaoyi Pan (Carnegie Mellon College), Giridharan Anantharaman (Fb), Nai-Chieh Huang (Carnegie Mellon College), Claire Jin (College of Laptop Science, Carnegie Mellon College), Daniel Pfrommer (None), Chenyang Yuan (Toyota Analysis Institute), Frank Permenter (Toyota Analysis Institute), Guannan Qu (Carnegie Mellon College), Nicholas Boffi (CMU, Carnegie Mellon College), Guanya Shi (CMU, Carnegie Mellon College), Max Simchowitz (Massachusetts Institute of Know-how)

Taming Imperfect Course of Verifiers: A Sampling Perspective on Backtracking

Authors: Dhruv Rohatgi (Massachusetts Institute of Know-how), Abhishek Shetty (College of California Berkeley), Donya Saless (College of California, Berkeley), Yuchen Li (Carnegie Mellon College), Ankur Moitra (Massachusetts Institute of Know-how), Andrej Risteski (Carnegie Mellon College), Dylan Foster (Microsoft Analysis NYC)

Ada-Diffuser: Latent-Conscious Adaptive Diffusion for Determination-Making

Authors: Fan Feng (College of California, San Diego), Selena Ge (College of California, San Diego), Minghao Fu (College of California, San Diego), Zijian Li (Mohamed bin Zayed College of Synthetic Intelligence), Yujia Zheng (Carnegie Mellon College), Zeyu Tang (Stanford College), Yingyao Hu (Johns Hopkins College), Biwei Huang (College of California, San Diego), Kun Zhang (Carnegie Mellon College & MBZUAI)

ZeroGR: A Generalizable and Scalable Framework for Zero-Shot Generative Retrieval

Authors: Weiwei Solar (Carnegie Mellon College), Keyi Kong (Shandong College), xinyu ma (Institute of Computing Know-how，Chinese language Academy of Science), Shuaiqiang Wang (Baidu Inc.), Dawei Yin (Baidu), Maarten de Rijke (College of Amsterdam), Zhaochun Ren (Leiden College), Yiming Yang (Carnegie Mellon College)

TokUR: Token-Stage Uncertainty Estimation for Massive Language Mannequin Reasoning

Authors: Tunyu Zhang (Rutgers College), Haizhou Shi (ML Lab@Rutgers), Yibin Wang (None), Hengyi Wang (Rutgers College), Xiaoxiao He (Fb), Zhuowei Li (Amazon), Haoxian Chen (Columbia College), Ligong Han (Rutgers College), Kai Xu (Amazon), Huan Zhang (CMU), Dimitris Metaxas (Rutgers College), Hao Wang (Rutgers College)

Terminal-Bench: Benchmarking Brokers on Arduous, Real looking Duties in Command Line Interfaces

Authors: Mike Merrill (None), Alexander Shaw (Brigham Younger College), Nicholas Carlini (Anthropic), Boxuan Li (Microsoft), Harsh Raj (Northeastern College), Ivan Bercovich (UC Santa Barbara + ScOp VC), Lin Shi (Cornell College), Jeong Shin (Snorkel AI), Thomas Walshe (Reflection AI), E. Kelly Buchanan (Columbia College), Junhong Shen (Carnegie Mellon College), Guanghao Ye (Massachusetts Institute of Know-how), Haowei Lin (Peking College), Jason Poulos (Impartial Researcher), Maoyu Wang (), Marianna Nezhurina (Juelich Supercomputing Heart, LAION, Tuebingen College), Di Lu (Tencent), Orfeas Menis Mastromichalakis (Nationwide Technical College of Athens), Zhiwei Xu (College of Michigan), Zizhao Chen (Division of Laptop Science, Cornell College), Yue Liu (NUS), Robert Zhang (College of Texas at Austin), Leon Liangyu Chen (Stanford College), Anurag Kashyap (Amazon), Jan-Lucas Uslu (Stanford College), Jeffrey Li (Carnegie Mellon College), Jianbo Wu (College of California, Merced), Minghao Yan (Division of Laptop Science, College of Wisconsin – Madison), Track Bian (College of Wisconsin-Madison), Vedang Sharma (Fremont Unified College District), Ke Solar (Amazon), Steven Dillmann (Stanford College), Akshay Anand (College of California, Berkeley), Andrew Lanpouthakoun (Stanford College), Bardia Koopah (College of California, Berkeley), Changran Hu (Sambanova Methods, Inc), Etash Guha (Stanford College, Anthropic), Gabriel Dreiman (Insitro), Jiacheng Zhu (Massachusetts Institute of Know-how), Karl Krauth (Stanford), Li Zhong (Anthropic), Niklas Muennighoff (Stanford College), Robert Amanfu (Impartial), Shangyin Tan (College of California, Berkeley), Shreyas Pimpalgaonkar (New York College), Tushar Aggarwal (Microsoft Analysis / Stanford), Xiangning Lin (CMU), Xin Lan (Michigan State College), Xuandong Zhao (UC Berkeley), Yiqing Liang (Brown College), Yuanli Wang (Boston College), Zilong (Ryan) Wang (UC San Diego), Changzhi Zhou (Tencent), David Heineman (Allen Institute for Synthetic Intelligence), Hange Liu (Microsoft), Harsh Trivedi (Allen Institute for Synthetic Intelligence), John Yang (Princeton College), Junhong Lin (Massachusetts Institute of Know-how), Manish Shetty (College of California, Berkeley), Michael Yang (College of California, Santa Barbara), Nabil Omi (Microsoft Analysis), Negin Raoof (College of California, Berkeley), Shanda Li (Carnegie Mellon College), Terry Yue Zhuo (Data61, CSIRO), Wuwei Lin (OpenAI), Yiwei Dai (Cornell College), Yuxin Wang (Dartmouth Faculty), Wenhao Chai (Princeton College), Shang Zhou (College of California, San Diego), Dariush Wahdany (CISPA Helmholtz Heart), Ziyu She (None), Jiaming Hu (Boston College), Zhikang Dong (State College of New York at Stony Brook), Yuxuan Zhu (College of Illinois Urbana-Champaign), Sasha Cui (Yale College), Ahson Saiyed (College of Virginia, Charlottesville), Arinbjörn Kolbeinsson (UVA & K01), Christopher Rytting (Brigham Younger College), Ryan Marten (Harbor), Yixin Wang (College of Michigan – Ann Arbor), Jenia Jitsev (LAION; Juelich Supercomputing Heart, Analysis Heart Juelich), Alex Dimakis (Electrical Engineering & Laptop Science Division, College of California, Berkeley), Andy Konwinski (College of California, Berkeley), Ludwig Schmidt (College of Washington / Stanford / Anthropic)

A Dense Subset Index for Collective Question Protection

Authors: Kartik Nair (Carnegie Mellon College), Pritish Chakraborty (Indian Institute of Know-how Bombay, Indian Institute of Know-how, Bombay), Atharva Tambat (Indian Institute of Know-how Bombay, Indian Institute of Know-how, Bombay), Indradyumna Roy (IIT Bombay, Aalto College), Soumen Chakrabarti (IIT Bombay), Anirban Dasgupta (IIT Gandhinagar), Abir De (Indian Institute of Know-how Bombay,)

MCP-Bench: Benchmarking Device-Utilizing LLM Brokers with Complicated Actual-World Duties through MCP Servers

Authors: Zhenting Wang (Rutgers College), Qi Chang (Accenture), Hemani Patel (College of California, Berkeley), Shashank Biju (College of California, Berkeley), Cheng-En Wu (Accenture), Quan Liu (Accenture), Aolin Ding (Accenture), Alireza Rezazadeh (Accenture), Ankit Parag Shah (Carnegie Mellon College), Yujia Bao (Accenture), Eugene Siow (Accenture)

STEM: SCALING TRANSFORMERS WITH EMBEDDING MODULES

Authors: Ranajoy Sadhukhan (Carnegie Mellon College), Sheng Cao (Meta), Harry Dong (Carnegie Mellon College), Changsheng Zhao (Meta Inc.), Attiano Purpura-Pontoniere (Meta – UCLA), Yuandong Tian (Meta AI Analysis), Zechun Liu (Meta), Beidi Chen (CMU, Carnegie Mellon College)

YuE: Scaling Open Basis Fashions for Lengthy-Kind Music Technology

Authors: Ruibin Yuan (Hong Kong College of Science and Know-how), Hanfeng Lin (Hong Kong College of Science and Know-how), Shuyue Guo (Beijing College of Posts and Telecommunications), Ge Zhang (College of Waterloo), Jiahao Pan (Hong Kong College of Science and Know-how), Yongyi Zang (Smule, Inc.), Haohe Liu (Ohio State College), Yiming Liang (College of the Chinese language Academy of Sciences), Wenye Ma (Mohamed bin Zayed College of Synthetic Intelligence), Xingjian Du (College of Rochester), Xeron Du (01.AI), Zhen Ye (The Hong Kong College of Science and Know-how), Tianyu Zheng (Beijing College of Posts and Telecommunications), Zhengxuan Jiang (Zhejiang College), Yinghao MA (Queen Mary College of London), Minghao Liu (2077AI), Zeyue Tian (Hong Kong College of Science and Know-how), Ziya Zhou (The Hong Kong College of Science and Know-how), Liumeng Xue (Hong Kong College of Science and Know-how), Xingwei Qu (College of Manchester), Yizhi Li (College of Manchester), Shangda Wu (Tencent), Tianhao Shen (Tianjin College), Ziyang Ma (Shanghai Jiao Tong College), Jun Zhan (Fudan College), Chunhui Wang (JD.com), Yatian Wang (The Hong Kong College of Science and Know-how), Xiaowei Chi (Hong Kong College of Science and Know-how), Xinyue Zhang (Nationwide College of Singapore), Zhenzhu Yang (China College of Geoscience Beijing), XiangzhouWang (Wuhan College of Engineering Science), Shansong Liu (Institute of Synthetic Intelligence (TeleAI), China Telecom), Lingrui Mei (College of the Chinese language Academy of Sciences), Peng Li (Hong Kong College of Science and Know-how), JUNJIE WANG (None), Jianwei Yu (Microsoft), Guojian Pang (ByteDance Inc.), Xu Li (Kuaishou- 快手科技), Zihao Wang (CMU, Carnegie Mellon College；ZJU，Zhejiang College), Xiaohuan Zhou (ByteDance Inc.), Lijun Yu (Google DeepMind), Emmanouil Benetos (Queen Mary College of London), Yong Chen (Geely Vehicle Analysis Institute (Ningbo) Co., Ltd), Chenghua Lin (College of Manchester ), Xie Chen (Shanghai Jiaotong College), Gus Xia (MBZUAI), Zhaoxiang Zhang (Institute of automation, Chinese language academy of science, Chinese language Academy of Sciences), Chao Zhang (Division of Digital Engineering, Tsinghua College), Wenhu Chen (College of Waterloo), Xinyu Zhou (Megvii Inc.), Xipeng Qiu (Fudan College), Roger Dannenberg (Carnegie Mellon College), JIAHENG LIU (Nanjing College), Jian Yang (Beihang College), Wenhao Huang (01.AI), Wei Xue (Hong Kong College of Science and Know-how), Xu Tan (Microsoft Analysis), Yike Guo (Imperial Faculty London)

PAT3D: Physics-Augmented Textual content-to-3D Scene Technology

Authors: Guying Lin (Carnegie Mellon College), Kemeng Huang (College of Hong Kong), Michael Liu (CMU, Carnegie Mellon College), Ruihan Gao (Carnegie Mellon College), Hanke Chen (Carnegie Mellon College), Lyuhao Chen (Carnegie Mellon College), Beijia Lu (Carnegie Mellon College), Taku Komura (the College of Hong Kong, College of Hong Kong), Yuan Liu (The College of Hong Kong), Jun-Yan Zhu (Carnegie Mellon College), Minchen Li (College of Engineering and Utilized Science, College of Pennsylvania)

Goedel-Prover-V2: Scaling Formal Theorem Proving with Scaffolded Knowledge Synthesis and Self-Correction

Authors: Yong Lin (Princeton College), Shange Tang (Princeton College), Bohan Lyu (Tsinghua College), Ziran Yang (Princeton College), Jui-Hui Chung (Princeton College), Haoyu Zhao (Princeton College), Lai Jiang (College of Illinois at Urbana-Champaign), Yihan Geng (Peking College), Jiawei Ge (Princeton College), Jingruo Solar (Stanford College), Jiayun Wu (Carnegie Mellon College), Jiri Gesi (Amazon Science), Ximing Lu (College of Washington), David Acuna (NVIDIA / Univ of Toronto), Kaiyu Yang (Meta), Hongzhou Lin (Amazon), Yejin Choi (Stanford College / NVIDIA), Danqi Chen (Princeton College), Sanjeev Arora (Princeton College), Chi Jin (Princeton College)

Numerous Dictionary Studying

Authors: Yujia Zheng (Carnegie Mellon College), Zijian Li (Mohamed bin Zayed College of Synthetic Intelligence), Shunxing Fan (Mohamed bin Zayed College of Synthetic Intelligence), Andrew Gordon Wilson (New York College), Kun Zhang (Carnegie Mellon College & MBZUAI)

Accelerating Eigenvalue Dataset Technology through Chebyshev Subspace Filter

Authors: Hong Wang (College of Science and Know-how of China), Jie Wang (College of Science and Know-how of China), Jian Luo (Stony Brook College), huanshuo dong (College of Science and Know-how of China), Yeqiu Chen (College of Science and Know-how of China), Runmin Jiang (Carnegie Mellon College), Zhen Huang (College of Science and Know-how of China)

Past Listening to: Studying Job-Agnostic ExG Representations from Earphones through Physiology-Knowledgeable Tokenization

Authors: Hyungjun Yoon (Korea Superior Institute of Science & Know-how), Seungjoo Lee (Carnegie Mellon College), Yu Wu (Dartmouth Faculty), XiaoMeng Chen (Shanghai Jiaotong College), Taiting Lu (Pennsylvania State College), Freddy Liu (College of Pennsylvania, College of Pennsylvania), Taeckyung Lee (KAIST), Hyeongheon Cha (Korea Superior Institute of Science & Know-how), Haochen Zhao (), Gaoteng Zhao (Northwest College), Dongyao Chen (Shanghai Jiaotong College), Cecilia Mascolo (College of Cambridge), Sung-Ju Lee (UCLA Laptop Science Division, College of California, Los Angeles), Lili Qiu (Microsoft)

The CoT Encyclopedia: Analyzing, Predicting, and Controlling how a Reasoning Mannequin will Assume

Authors: Seongyun Lee (KAIST AI), Seungone Kim (Carnegie Mellon College), Minju Web optimization (Korea Superior Institute of Science & Know-how), Yongrae Jo (KAIST), Dongyoung Go (Cornell College), Hyeonbin Hwang (Korea Superior Institute of Science & Know-how), Jinho Park (Korea Superior Institute of Science & Know-how), Xiang Yue (Carnegie Mellon College), Sean Welleck (Carnegie Mellon College), Graham Neubig (Carnegie Mellon College), Moontae Lee (College of Illinois, Chicago), Minjoon Web optimization (KAIST)

PersonaX: Multimodal Datasets with LLM-Inferred Conduct Traits

Authors: Loka Li (MBZUAI), Wong Kang (Mohamed bin Zayed College of Synthetic Intelligence), Minghao Fu (College of California, San Diego), Guangyi Chen (MBZUAI&CMU), Zhenhao Chen (MBZUAI), Gongxu Luo (Mohamed bin Zayed College of Synthetic Intelligence), Yuewen Solar (Mohamed bin Zayed College of Synthetic Intelligence), Salman Khan (Mohamed bin Zayed College of Synthetic Intelligence), Peter Spirtes (Carnegie Mellon College), Kun Zhang (Carnegie Mellon College & MBZUAI)