arxivId stringlengths 11 18 | arxivBaseId stringlengths 9 16 | version int64 1 19 | title stringlengths 8 207 | abstract stringlengths 87 2.34k | abstractWordCount int64 12 315 | authors listlengths 1 3.44k | authorCount int64 1 3.44k | firstAuthor stringlengths 3 55 | authorAffiliations listlengths 0 184 | primaryCategory stringclasses 137
values | categories listlengths 1 7 | isCrossListed bool 2
classes | publishedAt timestamp[s]date 1995-10-01 00:00:00 2026-07-30 17:59:56 | publishedDate timestamp[s]date 1995-10-01 00:00:00 2026-07-30 00:00:00 | updatedAt timestamp[s]date 1995-10-01 00:00:00 2026-07-30 17:59:56 | doi stringclasses 655
values | doiUrl stringclasses 655
values | journalRef stringclasses 541
values | comment stringlengths 1 463 ⌀ | absUrl stringlengths 33 40 | pdfUrl stringlengths 33 40 | sourceQuery stringclasses 51
values | scrapedAt timestamp[s]date 2026-08-02 22:16:30 2026-08-02 22:40:37 | _run_id stringclasses 90
values | _dataset_id stringclasses 90
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2607.25628v1 | 2607.25628 | 1 | Kernel-Checked Exclusions for the Erdős-Selfridge Odd Covering Problem: Any Odd Covering of $\mathbb{Z}$ Has lcm Exceeding 10000 | The Erdős-Selfridge odd covering problem (Erdős problem #7) asks whether a covering system of $\mathbb{Z}$ exists whose moduli are all odd, distinct, and greater than 1. The problem is open. We present a Lean 4 formalization, checked end to end by the proof kernel, of the exclusion: any covering of $\mathbb{Z}$ by fini... | 254 | [
"Ibrahim Mian",
"Shayaan Siddique"
] | 2 | Ibrahim Mian | [] | cs.LO | [
"cs.LO",
"math.NT"
] | true | 2026-07-28T12:10:26 | 2026-07-28T00:00:00 | 2026-07-28T12:10:26 | null | null | null | 11 pages. Lean 4 sources, certificates, and CI at https://github.com/ibrahimmian36/centurion | https://arxiv.org/abs/2607.25628v1 | https://arxiv.org/pdf/2607.25628v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2607.19719v1 | 2607.19719 | 1 | Koopman Dreamer: Spectrally Constrained Latent Dynamics for Stable World-Model Imagination | Latent world models improve sample efficiency in continuous control by optimizing policies over imagined latent trajectories, but common neural transitions offer limited direct control over modal persistence and error accumulation in long rollouts. We propose Koopman Dreamer, a Dreamer-style world model with a spectral... | 188 | [
"Jiaqi Li",
"Xinglong Zhang",
"Haibin Xie",
"Yixing Lan",
"Wei Pan",
"Xin Xu"
] | 6 | Jiaqi Li | [] | cs.LG | [
"cs.LG",
"cs.RO"
] | true | 2026-07-22T03:38:15 | 2026-07-22T00:00:00 | 2026-07-22T03:38:15 | null | null | null | 20 pages, 13 figures, 11 tables | https://arxiv.org/abs/2607.19719v1 | https://arxiv.org/pdf/2607.19719v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2607.10203v3 | 2607.10203 | 3 | Adaptive Compute in Latent World Models: When Depth Helps, Hurts, or Doesn't Matter | Adaptive compute for world models -- early-exit or mixture-of-depths predictors that spend variable depth per rollout step -- presumes that extra depth buys better predictions. In autoregressive rollouts, where planning actually happens, that premise requires depth's per-step precision to survive composition. We test i... | 262 | [
"Achyuthan Sivasankar"
] | 1 | Achyuthan Sivasankar | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-07-11T08:27:33 | 2026-07-11T00:00:00 | 2026-07-18T16:02:20 | null | null | null | 15 pages, 9 figures, 2 tables | https://arxiv.org/abs/2607.10203v3 | https://arxiv.org/pdf/2607.10203v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2607.02025v1 | 2607.02025 | 1 | Evaluating Vision-Language Models as a Zero-Shot Learning Alternative to You Only Look Once and Optical Character Recognition for Nigerian License Plate Recognition | License Plate Recognition (LPR) systems are critical tools in traffic monitoring, security enforcement, and urban mobility management. Traditional LPR systems often rely on a multi-stage pipeline involving object detection using You Only Look Once (YOLO) and Optical Character Recognition (OCR), which suffer from limita... | 167 | [
"Ismail Ismail Tijjani",
"Ahmad Abubakar Mustapaha",
"Sunusi Ibrahim Muhammad",
"Muhammad Bashir Aliyu"
] | 4 | Ismail Ismail Tijjani | [] | cs.CV | [
"cs.CV"
] | false | 2026-07-02T10:55:24 | 2026-07-02T00:00:00 | 2026-07-02T10:55:24 | null | null | null | null | https://arxiv.org/abs/2607.02025v1 | https://arxiv.org/pdf/2607.02025v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2607.00796v1 | 2607.00796 | 1 | Task-Relevant Representation Decoupling for Visual Reinforcement Learning Generalization | Visual Reinforcement Learning (VRL) has achieved considerable success in solving control tasks. However, generalizing learned policies to new environments remains a major challenge, as agents often overfit to task-irrelevant features in the training environment. To solve this problem, we introduce the concept of decoup... | 134 | [
"Jinwen Wang",
"Youfang Lin",
"Xiaobo Hu",
"Qian Xu",
"Shuo Wang",
"Zhuo Chen",
"Kai Lv"
] | 7 | Jinwen Wang | [] | cs.LG | [
"cs.LG"
] | false | 2026-07-01T11:26:25 | 2026-07-01T00:00:00 | 2026-07-01T11:26:25 | 10.1145/3820367 | https://doi.org/10.1145/3820367 | ACM Transactions on Multimedia Computing, Communications and Applications (TOMM), 2026 | 23 pages, 13 figures | https://arxiv.org/abs/2607.00796v1 | https://arxiv.org/pdf/2607.00796v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2606.29820v1 | 2606.29820 | 1 | Dual-Flow Reinforcement Learning with State-Aware Exploration | In complex continuous-control reinforcement learning tasks, multimodal optimal actions often coincide with uncertain, multimodal return distributions, making reliable value estimation and multimodal exploration challenging. Existing value estimation methods using unimodal Gaussians restrict expressiveness and yield bia... | 144 | [
"Qijun Li",
"Zheng Fu",
"Qi Song",
"Yifei He",
"Weitao Zhou",
"Kun Jiang",
"Diange Yang"
] | 7 | Qijun Li | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-06-29T05:57:33 | 2026-06-29T00:00:00 | 2026-06-29T05:57:33 | null | null | null | 12 pages, 6 figures, 1 table. This work has been submitted to the IEEE for possible publication | https://arxiv.org/abs/2606.29820v1 | https://arxiv.org/pdf/2606.29820v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2606.18943v1 | 2606.18943 | 1 | Physics-IQ Verified | Video generative models ( VGMs) have become a new frontier that can be used not just for video generation but for a multitude of downstream tasks, including world modeling. To advance these tasks, a good video model must understand the physical reality of the world. Evaluating this understanding is an emerging field an... | 191 | [
"Tim Rädsch",
"Yuki M Asano",
"Hilde Kuehne",
"Stefan Bauer",
"Priyank Jaini",
"Robert Geirhos",
"Carsten T. Lüth"
] | 7 | Tim Rädsch | [] | cs.CV | [
"cs.CV"
] | false | 2026-06-17T11:23:52 | 2026-06-17T00:00:00 | 2026-06-17T11:23:52 | null | null | null | null | https://arxiv.org/abs/2606.18943v1 | https://arxiv.org/pdf/2606.18943v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2606.16605v1 | 2606.16605 | 1 | ARB4WM: An Adversarial Robustness Benchmark for World Models in Continuous Control | World models are widely used in robotic and agentic engineering control systems due to their ability to learn latent dynamics for planning and decision-making. As these systems are increasingly deployed in safety-critical settings, understanding their robustness under adversarial conditions has become essential. Howeve... | 217 | [
"Junjian Zhang",
"Hao Tan",
"Ruonan Li",
"Dong Zhu",
"Aiping Li",
"Zhaoquan Gu"
] | 6 | Junjian Zhang | [] | cs.AI | [
"cs.AI"
] | false | 2026-06-15T11:51:19 | 2026-06-15T00:00:00 | 2026-06-15T11:51:19 | null | null | null | 24 pages, 10 figures, 5 tables. Source code available at https://github.com/zaoanguai/ARB4WM | https://arxiv.org/abs/2606.16605v1 | https://arxiv.org/pdf/2606.16605v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2606.15891v1 | 2606.15891 | 1 | On the number of symmetric chain decompositions of the minuscule lattices $L(m,n)$ and $M(n)$ | We study the problem of enumerating symmetric chain decompositions (SCDs) of the minuscule lattices $L(m,n)$ of partitions in an $m$ by $n$ box and $M(n)$ of partitions into distinct parts at most $n$. We shift the focus from constructing a single SCD to analyzing the global structure of the set of SCDs. Let $\#SCD(P)$... | 214 | [
"Robert Dorward"
] | 1 | Robert Dorward | [] | math.CO | [
"math.CO",
"math.RT"
] | true | 2026-06-14T16:24:08 | 2026-06-14T00:00:00 | 2026-06-14T16:24:08 | null | null | null | 27 pages, 6 figures. In preparation for submission. Comments welcome | https://arxiv.org/abs/2606.15891v1 | https://arxiv.org/pdf/2606.15891v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.29151v2 | 2605.29151 | 2 | Real-rootedness of the Poincaré polynomials of $\overline{\mathcal M}_{0,n}$: an AI-assisted proof | We prove real-rootedness for the Poincaré polynomial \[ P_n(t)=\sum_{i=0}^{n-3} \dim H^{2i}(\overline{\mathcal M}_{0,n};\mathbb{Q})t^i \] of the Deligne--Mumford moduli space $\overline{\mathcal M}_{0,n}$ of stable $n$-pointed rational curves, proving a conjecture of Aluffi--Chen--Marcolli. The proof starts from the Ke... | 227 | [
"Gergely Bérczi",
"Young-Hoon Kiem"
] | 2 | Gergely Bérczi | [] | math.AG | [
"math.AG",
"cs.AI",
"cs.NE"
] | true | 2026-05-27T22:26:05 | 2026-05-27T00:00:00 | 2026-06-11T14:48:54 | null | null | null | 16 pages | https://arxiv.org/abs/2605.29151v2 | https://arxiv.org/pdf/2605.29151v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.23415v2 | 2605.23415 | 2 | Reflex: Reinforcement Learning with Reflection Symmetry Exploitation in State-Based Continuous Control | Reinforcement learning has long struggled with poor sample efficiency. One promising approach to mitigate this problem is leveraging group-invariant Markov Decision Processes ($G$-invariant MDPs). Existing works in this direction have primarily focused on image-based RL and rotational symmetry such as $\mathrm{SO(2)}$,... | 153 | [
"Shuai Zhen",
"Yifan Zhang",
"Yuling Wang",
"Yanhua Yu"
] | 4 | Shuai Zhen | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-05-22T09:24:44 | 2026-05-22T00:00:00 | 2026-06-04T07:02:27 | null | null | null | Some of the data in the paper contain errors and need to be confirmed for modification | https://arxiv.org/abs/2605.23415v2 | https://arxiv.org/pdf/2605.23415v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.23220v1 | 2605.23220 | 1 | WMAttack: Automated Attack Search for Adversarial Evaluation of World-Model Agents | Despite the growing use of world models as decision-making agents, their adversarial robustness remains underexplored due to the lack of dedicated automated evaluation methods. A key obstacle is that attack evaluation must be both accurate and efficient: weak manually tuned attacks can overestimate robustness, while ex... | 215 | [
"Zhixiang Guo",
"Siyuan Liang",
"Shi Fu",
"Cheng Guo",
"Andras Balogh",
"Mark Jelasity",
"Dacheng Tao"
] | 7 | Zhixiang Guo | [] | cs.LG | [
"cs.LG"
] | false | 2026-05-22T04:31:09 | 2026-05-22T00:00:00 | 2026-05-22T04:31:09 | null | null | null | null | https://arxiv.org/abs/2605.23220v1 | https://arxiv.org/pdf/2605.23220v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.23089v1 | 2605.23089 | 1 | Dreaming Smoothly and Sample Efficiently with Gradient Penalized Latent Dynamics | Model-based reinforcement learning improves sample efficiency by learning a world model. However, existing latent world models such as DreamerV3 do not explicitly enforce local smoothness in their learned transition dynamics, leaving a useful inductive bias for transition dynamics learning unexploited. We propose GPLD,... | 163 | [
"Romil V. Sonigra",
"P. R. Kumar"
] | 2 | Romil V. Sonigra | [
"Texas A&M University"
] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-05-21T22:40:40 | 2026-05-21T00:00:00 | 2026-05-21T22:40:40 | null | null | null | 17 pages and 9 figures | https://arxiv.org/abs/2605.23089v1 | https://arxiv.org/pdf/2605.23089v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.23007v1 | 2605.23007 | 1 | MadEvolve: Evolutionary Optimization of Trading Systems with Large Language Models | We explore the application of LLM-driven algorithm optimization to several common tasks in quantitative finance. MadEvolve, a general-purpose algorithm optimization framework inspired by DeepMind's Alpha-Evolve, was recently developed to optimize algorithms in computational cosmology. Here we demonstrate the utility of... | 136 | [
"Yurii Kvasiuk",
"Tianyi Li",
"Owen Colegrove",
"Moritz Münchmeyer"
] | 4 | Yurii Kvasiuk | [] | q-fin.TR | [
"q-fin.TR",
"cs.AI",
"cs.LG",
"q-fin.PM"
] | true | 2026-05-21T20:28:57 | 2026-05-21T00:00:00 | 2026-05-21T20:28:57 | null | null | null | null | https://arxiv.org/abs/2605.23007v1 | https://arxiv.org/pdf/2605.23007v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.21804v1 | 2605.21804 | 1 | Mapping Tomato Cropping Systems in California Using AlphaEarth Geospatial Embeddings and Deep Learning Analysis | Field-scale crop maps support supply-chain forecasting and policy, yet statewide crop identification still often depends on retrospective surveys or remote-sensing workflows built around hand-engineered spectral features. Those pipelines can be accurate, but they require repeated preprocessing and often lose robustness... | 220 | [
"Mohammadreza Narimani",
"Alireza Pourreza",
"Parastoo Farajpoor"
] | 3 | Mohammadreza Narimani | [] | eess.IV | [
"eess.IV",
"cs.CV",
"cs.LG"
] | true | 2026-05-20T23:00:54 | 2026-05-20T00:00:00 | 2026-05-20T23:00:54 | null | null | null | 5 pages, 3 figures, 1 table. Preprint submitted to ASABE 2026 AIM | https://arxiv.org/abs/2605.21804v1 | https://arxiv.org/pdf/2605.21804v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.19592v1 | 2605.19592 | 1 | Implicit Action Chunking for Smooth Continuous Control | Reinforcement learning often produces high-frequency oscillatory control signals that undermine the safety and stability required for physical deployment. Explicit action chunking addresses this by predicting fixed-horizon trajectories but scales the policy output dimension proportionally with the horizon length, leadi... | 184 | [
"Bosun Liang",
"Shuo Pei",
"Zirui Chen",
"Chuanzhi Fan",
"Chen Sun",
"Yuankai Wu",
"Huachun Tan",
"Yong Wang"
] | 8 | Bosun Liang | [] | cs.RO | [
"cs.RO",
"cs.AI"
] | true | 2026-05-19T09:35:47 | 2026-05-19T00:00:00 | 2026-05-19T09:35:47 | null | null | null | null | https://arxiv.org/abs/2605.19592v1 | https://arxiv.org/pdf/2605.19592v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.16030v2 | 2605.16030 | 2 | Mind Dreamer: Untethering Imagination via Active Causal Intervention on Latent Manifolds | Model-Based Reinforcement Learning yields sample efficiency via latent imagination, yet remains constrained by Historical Tethering: imagination is typically initialized from observed states. This creates a learning asymmetry, where the world model's manifold discovery outpaces the policy's sparse-reward optimization. ... | 192 | [
"Shaojun Xu",
"Xiaoling Zhou",
"Yihan Lin",
"Yapeng Meng",
"Xinglong Ji",
"Luping Shi",
"Rong Zhao"
] | 7 | Shaojun Xu | [] | cs.LG | [
"cs.LG",
"cs.RO"
] | true | 2026-05-15T15:05:58 | 2026-05-15T00:00:00 | 2026-05-27T13:07:05 | null | null | null | 34 pages, 7 figures, ICML 2026 accepted | https://arxiv.org/abs/2605.16030v2 | https://arxiv.org/pdf/2605.16030v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.15342v1 | 2605.15342 | 1 | Minerva-Ego: Spatiotemporal Hints for Egocentric Video Understanding | Video reasoning models are a core component of egocentric and embodied agents. However, standard benchmarks for assessing models provide only evaluation of the output (e.g. the answer to a question), without evaluation of intermediate reasoning steps, and most provide answers only in the text domain. We introduce Miner... | 152 | [
"Arsha Nagrani",
"Jasper Uijilings",
"Shyamal Buch",
"Tobias Weyand",
"Sudheendra Vijayanarasimhan",
"Bo Hu",
"Ramin Mehran",
"David A Ross",
"Cordelia Schmid"
] | 9 | Arsha Nagrani | [] | cs.CV | [
"cs.CV",
"cs.LG"
] | true | 2026-05-14T19:12:20 | 2026-05-14T00:00:00 | 2026-05-14T19:12:20 | null | null | null | null | https://arxiv.org/abs/2605.15342v1 | https://arxiv.org/pdf/2605.15342v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.14277v1 | 2605.14277 | 1 | Parallelizing Counterfactual Regret Minimization | Parallelization has played an instrumental role in the field of artificial intelligence (AI), drastically reducing the time taken to train and evaluate large AI models. In contrast to its impact in the broader field of AI, applying parallelization to computational game solving is relatively unexplored, despite its grea... | 161 | [
"Juho Kim",
"Tuomas Sandholm"
] | 2 | Juho Kim | [] | cs.AI | [
"cs.AI",
"cs.GT"
] | true | 2026-05-14T02:22:27 | 2026-05-14T00:00:00 | 2026-05-14T02:22:27 | null | null | null | This paper contains and extends ideas that were originally in arxiv:2408.14778 | https://arxiv.org/abs/2605.14277v1 | https://arxiv.org/pdf/2605.14277v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.10402v1 | 2605.10402 | 1 | Every finite group admits a just finite presentation | A finite presentation < X | R > of a finite group is called `just finite' if removing any relation from R results in a presentation for an infinite group. It has been an open question (Kourovka Notebook, Problem 21.10) whether every finite group admits such a presentation. We resolve this conjecture in the affirmative. | 55 | [
"Marc Lackenby"
] | 1 | Marc Lackenby | [] | math.GR | [
"math.GR",
"cs.AI"
] | true | 2026-05-11T11:42:08 | 2026-05-11T00:00:00 | 2026-05-11T11:42:08 | null | null | null | 5 pages. Significant assistance was provided by the AI co-mathematician tool developed by Google DeepMind | https://arxiv.org/abs/2605.10402v1 | https://arxiv.org/pdf/2605.10402v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.09157v1 | 2605.09157 | 1 | Revisiting Mixture Policies in Entropy-Regularized Actor-Critic | Mixture policies theoretically offer greater flexibility than unimodal policies in continuous action reinforcement learning, but the practical benefits of this complexity remain elusive. Mixture policies are notably absent from most state-of-the-art algorithms, raising a fundamental question: Is the added representatio... | 170 | [
"Jiamin He",
"Samuel Neumann",
"Jincheng Mei",
"Adam White",
"Martha White"
] | 5 | Jiamin He | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-05-09T20:37:52 | 2026-05-09T00:00:00 | 2026-05-09T20:37:52 | null | null | null | null | https://arxiv.org/abs/2605.09157v1 | https://arxiv.org/pdf/2605.09157v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.05951v1 | 2605.05951 | 1 | HaM-World: Soft-Hamiltonian World Models with Selective Memory for Planning | World models enable model-based planning through learned latent dynamics, but imagined rollouts become unstable as the planning horizon grows or the dynamics distribution shifts. We argue that this instability reflects two missing structures in planner-facing latents: history-conditioned memory for approximate Markov c... | 227 | [
"Haoyun Tang",
"Haodong Cui",
"Keyao Xu",
"Kun Wang",
"Zhandong Mei"
] | 5 | Haoyun Tang | [] | cs.AI | [
"cs.AI"
] | false | 2026-05-07T09:58:18 | 2026-05-07T00:00:00 | 2026-05-07T09:58:18 | null | null | null | 22 pages, 5 figures. Code: https://github.com/HaoyunT/HaM_World | https://arxiv.org/abs/2605.05951v1 | https://arxiv.org/pdf/2605.05951v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2605.04709v1 | 2605.04709 | 1 | ELVIS: Ensemble-Calibrated Latent Imagination for Long-Horizon Visual MPC | A central challenge of visual control with model-based reinforcement learning (RL) is reliable long-horizon planning: long rollouts with learned latent dynamics exhibit branching futures and multi-modal action-value distributions. In addition, compounding model errors amplified by visual occlusions make deep imaginatio... | 196 | [
"Yurui Du",
"Pinhao Song",
"Yutong Hu",
"Renaud Detry"
] | 4 | Yurui Du | [] | cs.LG | [
"cs.LG",
"cs.RO",
"eess.SY"
] | true | 2026-05-06T09:59:58 | 2026-05-06T00:00:00 | 2026-05-06T09:59:58 | null | null | null | null | https://arxiv.org/abs/2605.04709v1 | https://arxiv.org/pdf/2605.04709v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2604.26275v1 | 2604.26275 | 1 | Agentic AI in the Software Development Lifecycle: Architecture, Empirical Evidence, and the Reshaping of Software Engineering | The arrival of large language models (LLMs) capable of multi-step reasoning, tool use, and long-horizon planning has produced a qualitative shift in software engineering. Where earlier code-completion tools such as GitHub Copilot operated at the granularity of a line or function, modern agentic systems -- Claude Code, ... | 218 | [
"Happy Bhati"
] | 1 | Happy Bhati | [] | cs.SE | [
"cs.SE"
] | false | 2026-04-29T04:06:47 | 2026-04-29T00:00:00 | 2026-04-29T04:06:47 | null | null | null | 9 pages, 5 figures, 2 tables, survey paper | https://arxiv.org/abs/2604.26275v1 | https://arxiv.org/pdf/2604.26275v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2604.23099v2 | 2604.23099 | 2 | ProEval: Proactive Failure Discovery and Efficient Performance Estimation for Generative AI Evaluation | Evaluating generative AI models is increasingly resource-intensive due to slow inference, expensive raters, and a rapidly growing landscape of models and benchmarks. We propose ProEval, a proactive evaluation framework that leverages transfer learning to efficiently estimate performance and identify failure cases. ProE... | 157 | [
"Yizheng Huang",
"Wenjun Zeng",
"Aditi Kumaresan",
"Zi Wang"
] | 4 | Yizheng Huang | [] | cs.LG | [
"cs.LG",
"cs.AI",
"stat.ML"
] | true | 2026-04-25T01:33:57 | 2026-04-25T00:00:00 | 2026-06-01T21:43:03 | null | null | International Conference on Machine Learning, 2026 | Our open-sourced code and data can be found at https://github.com/google-deepmind/proeval | https://arxiv.org/abs/2604.23099v2 | https://arxiv.org/pdf/2604.23099v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2604.21964v1 | 2604.21964 | 1 | Lessons from External Review of DeepMind's Scheming Inability Safety Case | Safety cases for frontier AI systems should provide a convincing argument, supported by evidence, that the risk of harm is within an acceptable bound. When developers author their own safety cases, confirmation bias and conflicted incentives can affect the quality of argument. External review can help to address this. ... | 115 | [
"Stephen Barrett",
"Francisco Javier Campos Zabala",
"Sean P. Fillingham",
"Umair Siddique",
"James Walpole",
"Robin Bloomfield",
"Henry Papadatos"
] | 7 | Stephen Barrett | [] | cs.CY | [
"cs.CY"
] | false | 2026-04-23T17:45:42 | 2026-04-23T00:00:00 | 2026-04-23T17:45:42 | null | null | null | null | https://arxiv.org/abs/2604.21964v1 | https://arxiv.org/pdf/2604.21964v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2604.14698v2 | 2604.14698 | 2 | Mean Flow Policy Optimization | Diffusion models have recently emerged as expressive policy representations for online reinforcement learning (RL). However, their iterative generative processes introduce substantial training and inference overhead. To overcome this limitation, we propose to represent policies using MeanFlow models, a class of few-ste... | 127 | [
"Xiaoyi Dong",
"Xi Sheryl Zhang",
"Jian Cheng"
] | 3 | Xiaoyi Dong | [] | cs.LG | [
"cs.LG"
] | false | 2026-04-16T06:59:52 | 2026-04-16T00:00:00 | 2026-05-30T05:53:13 | null | null | null | ICML 2026 | https://arxiv.org/abs/2604.14698v2 | https://arxiv.org/pdf/2604.14698v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2604.07426v1 | 2604.07426 | 1 | GIRL: Generative Imagination Reinforcement Learning via Information-Theoretic Hallucination Control | Model-based reinforcement learning (MBRL) improves sample efficiency by optimizing policies inside imagined rollouts, but long-horizon planning degrades when model errors compound and imagined trajectories drift off the training manifold. We introduce GIRL (Generative Imagination Reinforcement Learning), a latent world... | 218 | [
"Prakul Sunil Hiremath"
] | 1 | Prakul Sunil Hiremath | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-04-08T17:14:21 | 2026-04-08T00:00:00 | 2026-04-08T17:14:21 | null | null | null | 20 pages, 2 figures, 7 tables; reinforcement learning, world models | https://arxiv.org/abs/2604.07426v1 | https://arxiv.org/pdf/2604.07426v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2604.03782v1 | 2604.03782 | 1 | An Improved Last-Iterate Convergence Rate for Anchored Gradient Descent Ascent | We analyze the last-iterate convergence of the Anchored Gradient Descent Ascent algorithm for smooth convex-concave min-max problems. While previous work established a last-iterate rate of $\mathcal{O}(1/t^{2-2p})$ for the squared gradient norm, where $p \in (1/2, 1)$, it remained an open problem whether the improved e... | 83 | [
"Anja Surina",
"Arun Suggala",
"George Tsoukalas",
"Anton Kovsharov",
"Sergey Shirobokov",
"Francisco J. R. Ruiz",
"Pushmeet Kohli",
"Swarat Chaudhuri"
] | 8 | Anja Surina | [] | math.OC | [
"math.OC",
"cs.AI"
] | true | 2026-04-04T16:08:55 | 2026-04-04T00:00:00 | 2026-04-04T16:08:55 | null | null | null | null | https://arxiv.org/abs/2604.03782v1 | https://arxiv.org/pdf/2604.03782v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2604.01913v1 | 2604.01913 | 1 | The Rank and Gradient Lost in Non-stationarity: Sample Weight Decay for Mitigating Plasticity Loss in Reinforcement Learning | Deep reinforcement learning (RL) suffers from plasticity loss severely due to the nature of non-stationarity, which impairs the ability to adapt to new data and learn continually. Unfortunately, our understanding of how plasticity loss arises, dissipates, and can be dissolved remains limited to empirical findings, leav... | 261 | [
"Zihao Wu",
"Hongyao Tang",
"Yi Ma",
"Jiashun Liu",
"Yan Zheng",
"Jianye Hao"
] | 6 | Zihao Wu | [] | cs.LG | [
"cs.LG"
] | false | 2026-04-02T11:29:46 | 2026-04-02T00:00:00 | 2026-04-02T11:29:46 | null | null | null | ICLR | https://arxiv.org/abs/2604.01913v1 | https://arxiv.org/pdf/2604.01913v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.27751v1 | 2603.27751 | 1 | SkyNet: Belief-Aware Planning for Partially-Observable Stochastic Games | In 2019, Google DeepMind released MuZero, a model-based reinforcement learning method that achieves strong results in perfect-information games by combining learned dynamics models with Monte Carlo Tree Search (MCTS). However, comparatively little work has extended MuZero to partially observable, stochastic, multi-play... | 236 | [
"Adam Haile"
] | 1 | Adam Haile | [] | cs.AI | [
"cs.AI"
] | false | 2026-03-29T16:09:50 | 2026-03-29T00:00:00 | 2026-03-29T16:09:50 | null | null | null | null | https://arxiv.org/abs/2603.27751v1 | https://arxiv.org/pdf/2603.27751v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.27450v2 | 2603.27450 | 2 | FlowRL: A Taxonomy and Modular Framework for Reinforcement Learning with Diffusion Policies | Thanks to their remarkable flexibility, diffusion models and flow models have emerged as promising candidates for policy representation. However, efficient reinforcement learning (RL) upon these policies remains a challenge due to the lack of explicit log-probabilities for vanilla policy gradient estimators. While nume... | 174 | [
"Chenxiao Gao",
"Edward Chen",
"Tianyi Chen",
"Bo Dai"
] | 4 | Chenxiao Gao | [] | cs.LG | [
"cs.LG"
] | false | 2026-03-29T00:02:50 | 2026-03-29T00:00:00 | 2026-06-15T06:10:57 | null | null | null | accepted by RLC 2026 | https://arxiv.org/abs/2603.27450v2 | https://arxiv.org/pdf/2603.27450v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.22748v1 | 2603.22748 | 1 | Experimental investigation of magnetic properties of MnFeCo$_{4}$Si$_{2}$ discovered by GNoME | AI-driven inorganic materials research has garnered significant attention due to its ability to reduce the time, labor, and cost associated with experiments. An AI model known as GNoME, recently developed by Google DeepMind, is particularly fascinating because it is integrated with the Materials Project open database. ... | 115 | [
"Shuhei Naganuma",
"Jiro Kitagawa"
] | 2 | Shuhei Naganuma | [] | cond-mat.mtrl-sci | [
"cond-mat.mtrl-sci"
] | false | 2026-03-24T03:23:16 | 2026-03-24T00:00:00 | 2026-03-24T03:23:16 | null | null | null | to appear in J. Magn. Magn, Mater | https://arxiv.org/abs/2603.22748v1 | https://arxiv.org/pdf/2603.22748v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.18202v2 | 2603.18202 | 2 | R2-Dreamer: Redundancy-Reduced World Models without Decoders or Augmentation | A central challenge in image-based Model-Based Reinforcement Learning (MBRL) is to learn representations that distill essential information from irrelevant visual details. While promising, reconstruction-based methods often waste capacity on large task-irrelevant regions. Decoder-free methods instead learn robust repre... | 148 | [
"Naoki Morihira",
"Amal Nahar",
"Kartik Bharadwaj",
"Yasuhiro Kato",
"Akinobu Hayashi",
"Tatsuya Harada"
] | 6 | Naoki Morihira | [] | cs.LG | [
"cs.LG",
"cs.AI",
"cs.RO"
] | true | 2026-03-18T18:54:07 | 2026-03-18T00:00:00 | 2026-03-20T02:05:12 | null | null | Published as a conference paper at ICLR 2026 | 20 pages, 12 figures, 2 tables | https://arxiv.org/abs/2603.18202v2 | https://arxiv.org/pdf/2603.18202v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.11110v1 | 2603.11110 | 1 | ResWM: Residual-Action World Model for Visual RL | Learning predictive world models from raw visual observations is a central challenge in reinforcement learning (RL), especially for robotics and continuous control. Conventional model-based RL frameworks directly condition future predictions on absolute actions, which makes optimization unstable: the optimal action dis... | 248 | [
"Jseen Zhang",
"Gabriel Adineera",
"Jinzhou Tan",
"Jinoh Kim"
] | 4 | Jseen Zhang | [] | cs.RO | [
"cs.RO",
"cs.AI"
] | true | 2026-03-11T11:27:08 | 2026-03-11T00:00:00 | 2026-03-11T11:27:08 | null | null | null | Submit KDD2026 | https://arxiv.org/abs/2603.11110v1 | https://arxiv.org/pdf/2603.11110v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.09619v2 | 2603.09619 | 2 | Context Engineering: From Prompts to Corporate Multi-Agent Architecture | As artificial intelligence (AI) systems evolve from stateless chatbots to autonomous multi-step agents, prompt engineering (PE), the discipline of crafting individual queries, proves necessary but insufficient. This paper introduces context engineering (CE) as a standalone discipline concerned with designing, structuri... | 229 | [
"Vera V. Vishnyakova"
] | 1 | Vera V. Vishnyakova | [] | cs.AI | [
"cs.AI",
"cs.MA"
] | true | 2026-03-10T12:58:31 | 2026-03-10T00:00:00 | 2026-03-13T03:59:39 | null | null | null | 25 pages, 1 figure | https://arxiv.org/abs/2603.09619v2 | https://arxiv.org/pdf/2603.09619v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.05113v1 | 2603.05113 | 1 | Decoupling Task and Behavior: A Two-Stage Reward Curriculum in Reinforcement Learning for Robotics | Deep Reinforcement Learning is a promising tool for robotic control, yet practical application is often hindered by the difficulty of designing effective reward functions. Real-world tasks typically require optimizing multiple objectives simultaneously, necessitating precise tuning of their weights to learn a policy wi... | 159 | [
"Kilian Freitag",
"Knut Åkesson",
"Morteza Haghir Chehreghani"
] | 3 | Kilian Freitag | [] | cs.LG | [
"cs.LG",
"cs.RO"
] | true | 2026-03-05T12:34:27 | 2026-03-05T00:00:00 | 2026-03-05T12:34:27 | null | null | null | null | https://arxiv.org/abs/2603.05113v1 | https://arxiv.org/pdf/2603.05113v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.02935v1 | 2603.02935 | 1 | Contextual Latent World Models for Offline Meta Reinforcement Learning | Offline meta-reinforcement learning seeks to learn policies that generalize across related tasks from fixed datasets. Context-based methods infer a task representation from transition histories, but learning effective task representations without supervision remains a challenge. In parallel, latent world models have de... | 110 | [
"Mohammadreza Nakheai",
"Aidan Scannell",
"Kevin Luck",
"Joni Pajarinen"
] | 4 | Mohammadreza Nakheai | [] | cs.LG | [
"cs.LG"
] | false | 2026-03-03T12:45:20 | 2026-03-03T00:00:00 | 2026-03-03T12:45:20 | null | null | null | null | https://arxiv.org/abs/2603.02935v1 | https://arxiv.org/pdf/2603.02935v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.02765v1 | 2603.02765 | 1 | Next Embedding Prediction Makes World Models Stronger | Capturing temporal dependencies is critical for model-based reinforcement learning (MBRL) in partially observable, high-dimensional domains. We introduce NE-Dreamer, a decoder-free MBRL agent that leverages a temporal transformer to predict next-step encoder embeddings from latent state sequences, directly optimizing t... | 113 | [
"George Bredis",
"Nikita Balagansky",
"Daniil Gavrilov",
"Ruslan Rakhimov"
] | 4 | George Bredis | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-03-03T09:04:28 | 2026-03-03T00:00:00 | 2026-03-03T09:04:28 | null | null | null | null | https://arxiv.org/abs/2603.02765v1 | https://arxiv.org/pdf/2603.02765v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.02613v1 | 2603.02613 | 1 | Real-Time Generative Policy via Langevin-Guided Flow Matching for Autonomous Driving | Reinforcement learning (RL) is a fundamental methodology in autonomous driving systems, where generative policies exhibit considerable potential by leveraging their ability to model complex distributions to enhance exploration. However, their inherent high inference latency severely impedes their deployment in real-tim... | 189 | [
"Tianze Zhu",
"Yinuo Wang",
"Wenjun Zou",
"Tianyi Zhang",
"Likun Wang",
"Letian Tao",
"Feihong Zhang",
"Yao Lyu",
"Shengbo Eben Li"
] | 9 | Tianze Zhu | [] | cs.LG | [
"cs.LG",
"cs.RO"
] | true | 2026-03-03T05:35:53 | 2026-03-03T00:00:00 | 2026-03-03T05:35:53 | null | null | null | null | https://arxiv.org/abs/2603.02613v1 | https://arxiv.org/pdf/2603.02613v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2602.24181v2 | 2602.24181 | 2 | A Mixed Diet Makes DINO An Omnivorous Vision Encoder | Pre-trained vision encoders like DINOv2 have demonstrated exceptional performance on unimodal tasks. However, we observe that their features are poorly aligned across different visual modalities. For instance, the feature embedding for an RGB image and its corresponding depth map of the same scene exhibit a cosine simi... | 159 | [
"Rishabh Kabra",
"Maks Ovsjanikov",
"Drew A. Hudson",
"Ye Xia",
"Skanda Koppula",
"Andre Araujo",
"Joao Carreira",
"Niloy J. Mitra"
] | 8 | Rishabh Kabra | [] | cs.CV | [
"cs.CV",
"cs.AI"
] | true | 2026-02-27T17:03:45 | 2026-02-27T00:00:00 | 2026-06-05T18:04:44 | null | null | null | CVPR 2026 Highlight | https://arxiv.org/abs/2602.24181v2 | https://arxiv.org/pdf/2602.24181v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2602.21201v3 | 2602.21201 | 3 | Aletheia tackles FirstProof autonomously | We report the performance of Aletheia (Feng et al., 2026b), a mathematics research agent powered by Gemini 3 Deep Think, on the inaugural FirstProof challenge. Within the allowed timeframe of the challenge, Aletheia autonomously solved 6 problems (2, 5, 7, 8, 9, 10) out of 10 according to majority expert assessments; w... | 90 | [
"Tony Feng",
"Junehyuk Jung",
"Sang-hyun Kim",
"Carlo Pagano",
"Sergei Gukov",
"Chiang-Chiang Tsai",
"David Woodruff",
"Adel Javanmard",
"Aryan Mokhtari",
"Dawsen Hwang",
"Yuri Chervonyi",
"Jonathan N. Lee",
"Garrett Bingham",
"Trieu H. Trinh",
"Vahab Mirrokni",
"Quoc V. Le",
"Thang ... | 17 | Tony Feng | [] | cs.AI | [
"cs.AI",
"cs.CL",
"cs.LG"
] | true | 2026-02-24T18:56:10 | 2026-02-24T00:00:00 | 2026-03-15T15:46:15 | null | null | null | 41 pages. Project page: https://github.com/google-deepmind/superhuman/tree/main/aletheia | https://arxiv.org/abs/2602.21201v3 | https://arxiv.org/pdf/2602.21201v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2602.19837v3 | 2602.19837 | 3 | Meta-Learning and Meta-Reinforcement Learning -- Tracing the Path towards DeepMind's Adaptive Agent | Humans are highly effective at utilizing prior knowledge to adapt to novel tasks, a capability that standard machine learning models struggle to replicate due to their reliance on task-specific training. Meta-learning overcomes this limitation by allowing models to acquire transferable knowledge from various tasks, ena... | 96 | [
"Björn Hoppmann",
"Christoph Scholz"
] | 2 | Björn Hoppmann | [] | cs.AI | [
"cs.AI",
"cs.LG"
] | true | 2026-02-23T13:39:58 | 2026-02-23T00:00:00 | 2026-05-06T07:57:45 | null | null | null | null | https://arxiv.org/abs/2602.19837v3 | https://arxiv.org/pdf/2602.19837v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2602.19582v1 | 2602.19582 | 1 | Advantage-based Temporal Attack in Reinforcement Learning | Extensive research demonstrates that Deep Reinforcement Learning (DRL) models are susceptible to adversarially constructed inputs (i.e., adversarial examples), which can mislead the agent to take suboptimal or unsafe actions. Recent methods improve attack effectiveness by leveraging future rewards to guide adversarial ... | 200 | [
"Shenghong He"
] | 1 | Shenghong He | [] | cs.LG | [
"cs.LG"
] | false | 2026-02-23T08:08:23 | 2026-02-23T00:00:00 | 2026-02-23T08:08:23 | null | null | null | null | https://arxiv.org/abs/2602.19582v1 | https://arxiv.org/pdf/2602.19582v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2602.14351v2 | 2602.14351 | 2 | WIMLE: Uncertainty-Aware World Models with IMLE for Sample-Efficient Continuous Control | Model-based reinforcement learning promises strong sample efficiency but often underperforms in practice due to compounding model error, unimodal world models that average over multi-modal dynamics, and overconfident predictions that bias learning. We introduce WIMLE, a model-based method that extends Implicit Maximum ... | 170 | [
"Mehran Aghabozorgi",
"Alireza Moazeni",
"Yanshu Zhang",
"Ke Li"
] | 4 | Mehran Aghabozorgi | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-02-15T23:53:16 | 2026-02-15T00:00:00 | 2026-04-03T18:41:06 | null | null | In Proceedings of the Fourteenth International Conference on Learning Representations (ICLR 2026), 2026 | Accepted at ICLR 2026. Website: https://mehranagh20.github.io/wimle/ Code: https://github.com/mehranagh20/wimle | https://arxiv.org/abs/2602.14351v2 | https://arxiv.org/pdf/2602.14351v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2603.03338v2 | 2603.03338 | 2 | AI Researchers' Views on Automating AI R&D and Intelligence Explosions | Many leading AI researchers expect AI development to exceed the transformative impact of all previous technological revolutions. This belief is based on the idea that AI will be able to automate the process of AI research itself, leading to a positive feedback loop. In August and September of 2025, we interviewed 25 le... | 236 | [
"Severin Field",
"Raymond Douglas",
"David Krueger"
] | 3 | Severin Field | [] | cs.CY | [
"cs.CY"
] | false | 2026-02-13T23:24:05 | 2026-02-13T00:00:00 | 2026-03-05T21:46:37 | null | null | null | 26 pages, 4 figures | https://arxiv.org/abs/2603.03338v2 | https://arxiv.org/pdf/2603.03338v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2602.12643v2 | 2602.12643 | 2 | Unifying Model-Free Efficiency and Model-Based Representations via Latent Dynamics | We present Unified Latent Dynamics (ULD), a novel reinforcement learning algorithm that unifies the efficiency of model-free methods with the representational strengths of model-based approaches, without incurring planning overhead. By embedding state-action pairs into a latent space in which the true value function is... | 198 | [
"Jashaswimalya Acharjee",
"Balaraman Ravindran"
] | 2 | Jashaswimalya Acharjee | [] | cs.LG | [
"cs.LG",
"cs.AI",
"stat.ML"
] | true | 2026-02-13T06:06:56 | 2026-02-13T00:00:00 | 2026-05-29T10:14:59 | null | null | null | Similarities found with a prior work. Hence, requesting for withdrawal until further notice | https://arxiv.org/abs/2602.12643v2 | https://arxiv.org/pdf/2602.12643v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2602.10177v3 | 2602.10177 | 3 | Towards Autonomous Mathematics Research | Recent advances in foundational models have yielded reasoning systems capable of achieving a gold-medal standard at the International Mathematical Olympiad. The transition from competition-level problem-solving to professional research, however, requires navigating vast literature and constructing long-horizon proofs. ... | 245 | [
"Tony Feng",
"Trieu H. Trinh",
"Garrett Bingham",
"Dawsen Hwang",
"Yuri Chervonyi",
"Junehyuk Jung",
"Joonkyung Lee",
"Carlo Pagano",
"Sang-hyun Kim",
"Federico Pasqualotto",
"Sergei Gukov",
"Jonathan N. Lee",
"Junsu Kim",
"Kaiying Hou",
"Golnaz Ghiasi",
"Yi Tay",
"YaGuang Li",
"Ch... | 28 | Tony Feng | [] | cs.LG | [
"cs.LG",
"cs.AI",
"cs.CL",
"cs.CY"
] | true | 2026-02-10T18:50:15 | 2026-02-10T00:00:00 | 2026-03-06T15:41:44 | null | null | null | 42 pages, updated with summary of FirstProof results. Accompanied blog post https://deepmind.google/blog/accelerating-mathematical-and-scientific-discovery-with-gemini-deep-think/ | https://arxiv.org/abs/2602.10177v3 | https://arxiv.org/pdf/2602.10177v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2602.02582v1 | 2602.02582 | 1 | Uncertainty and Fairness Awareness in LLM-Based Recommendation Systems | Large language models (LLMs) enable powerful zero-shot recommendations by leveraging broad contextual knowledge, yet predictive uncertainty and embedded biases threaten reliability and fairness. This paper studies how uncertainty and fairness evaluations affect the accuracy, consistency, and trustworthiness of LLM-gene... | 192 | [
"Chandan Kumar Sah",
"Xiaoli Lian",
"Li Zhang",
"Tony Xu",
"Syed Shazaib Shah"
] | 5 | Chandan Kumar Sah | [] | cs.AI | [
"cs.AI",
"cs.CL",
"cs.CY",
"cs.IR",
"cs.LG",
"cs.SE"
] | true | 2026-01-31T17:18:13 | 2026-01-31T00:00:00 | 2026-01-31T17:18:13 | null | null | null | Accepted at the Second Conference of the International Association for Safe and Ethical Artificial Intelligence, IASEAI26, 14 pages | https://arxiv.org/abs/2602.02582v1 | https://arxiv.org/pdf/2602.02582v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2601.21725v2 | 2601.21725 | 2 | Procedural Pretraining: Warming Up Language Models with Abstract Data | Pretraining language models directly on web-scale corpora is the de facto paradigm. We study an alternative where the model is initially exposed to abstract structured data to ease the subsequent acquisition of rich semantic knowledge, much like humans learning simple logic and mathematics before higher reasoning. We f... | 251 | [
"Liangze Jiang",
"Zachary Shinnick",
"Anton van den Hengel",
"Hemanth Saratchandran",
"Damien Teney"
] | 5 | Liangze Jiang | [] | cs.CL | [
"cs.CL",
"cs.LG"
] | true | 2026-01-29T13:48:43 | 2026-01-29T00:00:00 | 2026-05-27T19:52:53 | null | null | null | ICML 2026. Project page: https://zlshinnick.github.io/procedural-pretraining-page/ | https://arxiv.org/abs/2601.21725v2 | https://arxiv.org/pdf/2601.21725v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2601.19336v1 | 2601.19336 | 1 | From Observations to Events: Event-Aware World Model for Reinforcement Learning | While model-based reinforcement learning (MBRL) improves sample efficiency by learning world models from raw observations, existing methods struggle to generalize across structurally similar scenes and remain vulnerable to spurious variations such as textures or color shifts. From a cognitive science perspective, human... | 187 | [
"Zhao-Han Peng",
"Shaohui Li",
"Zhi Li",
"Shulan Ruan",
"Yu Liu",
"You He"
] | 6 | Zhao-Han Peng | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-01-27T08:20:44 | 2026-01-27T00:00:00 | 2026-01-27T08:20:44 | null | null | null | 43 pages, accepted by ICLR 2026 | https://arxiv.org/abs/2601.19336v1 | https://arxiv.org/pdf/2601.19336v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2601.12204v1 | 2601.12204 | 1 | Explicit and Implicit Finite Difference Solvers Implemented in JAX for Shock Wave Physics | Shock dynamics and nonlinear wave propagation are fundamental to computational fluid dynamics (CFD) and high-speed flow modeling. In this study, we developed explicit and implicit finite-difference solvers for the one-dimensional Burgers viscous equation to model shock formation, propagation, and dissipation. The gover... | 211 | [
"Avinash Potluri",
"Arturo Rodriguez",
"Taylor N. Garcia",
"Chelsea M. Caballero",
"Katrina I. Sanchez",
"Payal Helambe",
"Vineeth V. Kumar",
"Francisco O. Aguirre Ortega"
] | 8 | Avinash Potluri | [] | physics.flu-dyn | [
"physics.flu-dyn",
"physics.comp-ph"
] | true | 2026-01-18T00:52:55 | 2026-01-18T00:00:00 | 2026-01-18T00:52:55 | null | null | null | 11 pages | https://arxiv.org/abs/2601.12204v1 | https://arxiv.org/pdf/2601.12204v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2601.10904v1 | 2601.10904 | 1 | ARC Prize 2025: Technical Report | The ARC-AGI benchmark series serves as a critical measure of few-shot generalization on novel tasks, a core aspect of intelligence. The ARC Prize 2025 global competition targeted the newly released ARC-AGI-2 dataset, which features greater task complexity compared to its predecessor. The Kaggle competition attracted 1,... | 247 | [
"François Chollet",
"Mike Knoop",
"Gregory Kamradt",
"Bryan Landers"
] | 4 | François Chollet | [] | cs.AI | [
"cs.AI"
] | false | 2026-01-15T23:23:56 | 2026-01-15T00:00:00 | 2026-01-15T23:23:56 | null | null | null | null | https://arxiv.org/abs/2601.10904v1 | https://arxiv.org/pdf/2601.10904v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2601.10649v2 | 2601.10649 | 2 | MINERVA-Cultural: A Benchmark for Cultural and Multilingual Long Video Reasoning | Recent advancements in video models have shown tremendous progress, particularly in long video understanding. However, current benchmarks predominantly feature western-centric data and English as the dominant language, introducing significant biases in evaluation. To address this, we introduce MINERVA-Cultural, a chall... | 153 | [
"Darshan Singh",
"Arsha Nagrani",
"Kawshik Manikantan",
"Harman Singh",
"Dinesh Tewari",
"Tobias Weyand",
"Cordelia Schmid",
"Anelia Angelova",
"Shachi Dave"
] | 9 | Darshan Singh | [] | cs.CV | [
"cs.CV"
] | false | 2026-01-15T18:15:06 | 2026-01-15T00:00:00 | 2026-04-07T10:36:27 | null | null | null | Accepted to CVPR 2026 | https://arxiv.org/abs/2601.10649v2 | https://arxiv.org/pdf/2601.10649v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2601.09236v3 | 2601.09236 | 3 | Reward Learning through Ranking Mean Squared Error | Reward design remains a significant bottleneck in applying reinforcement learning (RL) to real-world problems. A popular alternative is reward learning, where reward functions are inferred from human feedback rather than manually specified. Recent work has proposed learning reward functions from human ratings rather th... | 159 | [
"Chaitanya Kharyal",
"Calarina Muslimani",
"Matthew E. Taylor"
] | 3 | Chaitanya Kharyal | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2026-01-14T07:18:12 | 2026-01-14T00:00:00 | 2026-06-03T19:47:43 | null | null | null | null | https://arxiv.org/abs/2601.09236v3 | https://arxiv.org/pdf/2601.09236v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2601.05943v2 | 2601.05943 | 2 | Global Optimization for Combinatorial Geometry Problems Revisited in the Era of LLMs | Recent progress in LLM-driven algorithm discovery, exemplified by DeepMind's AlphaEvolve, has produced new best-known solutions for a range of hard geometric and combinatorial problems. This raises a natural question: to what extent can modern off-the-shelf global optimization solvers match such results when the proble... | 154 | [
"Timo Berthold",
"Dominik Kamp",
"Gioni Mexi",
"Sebastian Pokutta",
"Imre Pólik"
] | 5 | Timo Berthold | [] | math.OC | [
"math.OC"
] | false | 2026-01-09T16:59:52 | 2026-01-09T00:00:00 | 2026-01-15T07:41:55 | null | null | null | null | https://arxiv.org/abs/2601.05943v2 | https://arxiv.org/pdf/2601.05943v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2601.01235v1 | 2601.01235 | 1 | Bruhat intervals that are large hypercubes | We study the question of finding big Bruhat intervals that are poset hypercubes in the symmetric group $S_n$. Using permutations suggested by AlphaEvolve (an evolutionary coding agent developed by Google DeepMind), we were led to an unusual situation in which the agent produced a pattern which performed well for the $n... | 205 | [
"Jordan Ellenberg",
"Nicolas Libedinsky",
"David Plaza",
"José Simental",
"Geordie Williamson"
] | 5 | Jordan Ellenberg | [] | math.CO | [
"math.CO",
"math.RT"
] | true | 2026-01-03T16:48:08 | 2026-01-03T00:00:00 | 2026-01-03T16:48:08 | null | null | null | 24 pages, comments welcome! | https://arxiv.org/abs/2601.01235v1 | https://arxiv.org/pdf/2601.01235v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2512.24440v1 | 2512.24440 | 1 | Towards mechanistic understanding in a data-driven weather model: internal activations reveal interpretable physical features | Large data-driven physics models like DeepMind's weather model GraphCast have empirically succeeded in parameterizing time operators for complex dynamical systems with an accuracy reaching or in some cases exceeding that of traditional physics-based solvers. Unfortunately, how these data-driven models perform computati... | 202 | [
"Theodore MacMillan",
"Nicholas T. Ouellette"
] | 2 | Theodore MacMillan | [] | physics.ao-ph | [
"physics.ao-ph",
"cs.LG",
"physics.comp-ph"
] | true | 2025-12-30T19:50:30 | 2025-12-30T00:00:00 | 2025-12-30T19:50:30 | null | null | null | 18 pages, 13 figures | https://arxiv.org/abs/2512.24440v1 | https://arxiv.org/pdf/2512.24440v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2601.00857v1 | 2601.00857 | 1 | Harvesting AlphaEarth: Benchmarking the Geospatial Foundation Model for Agricultural Downstream Tasks | Geospatial foundation models (GFMs) have emerged as a promising approach to overcoming the limitations in existing featurization methods. More recently, Google DeepMind has introduced AlphaEarth Foundation (AEF), a GFM pre-trained using multi-source EOs across continuous time. An annual and global embedding dataset is ... | 258 | [
"Yuchi Ma",
"Yawen Shen",
"Anu Swatantran",
"David B. Lobell"
] | 4 | Yuchi Ma | [] | cs.LG | [
"cs.LG"
] | false | 2025-12-30T01:04:22 | 2025-12-30T00:00:00 | 2025-12-30T01:04:22 | 10.1016/j.jag.2026.105258 | https://doi.org/10.1016/j.jag.2026.105258 | null | null | https://arxiv.org/abs/2601.00857v1 | https://arxiv.org/pdf/2601.00857v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2512.15036v2 | 2512.15036 | 2 | Spectral Representation-based Reinforcement Learning | In real-world applications with large state and action spaces, reinforcement learning (RL) typically employs function approximations to represent core components like the policies, value functions, and dynamics models. Although powerful approximations such as neural networks offer great expressiveness, they often prese... | 189 | [
"Chenxiao Gao",
"Haotian Sun",
"Na Li",
"Dale Schuurmans",
"Bo Dai"
] | 5 | Chenxiao Gao | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2025-12-17T02:54:42 | 2025-12-17T00:00:00 | 2026-01-28T00:45:02 | null | null | null | null | https://arxiv.org/abs/2512.15036v2 | https://arxiv.org/pdf/2512.15036v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2512.12812v2 | 2512.12812 | 2 | Does Tone Change the Answer? Evaluating Prompt Politeness Effects on Modern LLMs: GPT, Gemini, and LLaMA | Prompt engineering has emerged as a critical factor influencing large language model (LLM) performance, yet the impact of pragmatic elements such as linguistic tone and politeness remains underexplored, particularly across different model families. In this work, we propose a systematic evaluation framework to examine h... | 232 | [
"Hanyu Cai",
"Binqi Shen",
"Lier Jin",
"Lan Hu",
"Xiaojing Fan"
] | 5 | Hanyu Cai | [] | cs.CL | [
"cs.CL",
"cs.AI"
] | true | 2025-12-14T19:25:20 | 2025-12-14T00:00:00 | 2026-03-27T20:35:20 | null | null | null | null | https://arxiv.org/abs/2512.12812v2 | https://arxiv.org/pdf/2512.12812v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2512.10016v1 | 2512.10016 | 1 | Latent Action World Models for Control with Unlabeled Trajectories | Inspired by how humans combine direct interaction with action-free experience (e.g., videos), we study world models that learn from heterogeneous data. Standard world models typically rely on action-conditioned trajectories, which limits effectiveness when action labels are scarce. We introduce a family of latent-actio... | 177 | [
"Marvin Alles",
"Xingyuan Zhang",
"Patrick van der Smagt",
"Philip Becker-Ehmck"
] | 4 | Marvin Alles | [] | cs.LG | [
"cs.LG"
] | false | 2025-12-10T19:09:45 | 2025-12-10T00:00:00 | 2025-12-10T19:09:45 | null | null | null | null | https://arxiv.org/abs/2512.10016v1 | https://arxiv.org/pdf/2512.10016v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2512.07437v1 | 2512.07437 | 1 | KAN-Dreamer: Benchmarking Kolmogorov-Arnold Networks as Function Approximators in World Models | DreamerV3 is a state-of-the-art online model-based reinforcement learning (MBRL) algorithm known for remarkable sample efficiency. Concurrently, Kolmogorov-Arnold Networks (KANs) have emerged as a promising alternative to Multi-Layer Perceptrons (MLPs), offering superior parameter efficiency and interpretability. To mi... | 178 | [
"Chenwei Shi",
"Xueyu Luan"
] | 2 | Chenwei Shi | [] | cs.LG | [
"cs.LG",
"cs.AI",
"cs.CV",
"cs.NE",
"cs.RO"
] | true | 2025-12-08T11:13:15 | 2025-12-08T00:00:00 | 2025-12-08T11:13:15 | null | null | null | 23 pages, 8 figures, 3 tables | https://arxiv.org/abs/2512.07437v1 | https://arxiv.org/pdf/2512.07437v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2512.13706v1 | 2512.13706 | 1 | Mitigating Catastrophic Forgetting in Mathematical Reasoning Finetuning through Mixed Training | When finetuning large language models for specialized tasks such as mathematical reasoning, models exhibit catastrophic forgetting, losing previously learned capabilities. We investigate this by finetuning Flan-T5-Base (250M parameters) on the DeepMind Mathematics dataset and measuring forgetting on MultiNLI. Math-only... | 158 | [
"John Graham Reynolds"
] | 1 | John Graham Reynolds | [] | cs.LG | [
"cs.LG",
"cs.CL"
] | true | 2025-12-05T17:18:23 | 2025-12-05T00:00:00 | 2025-12-05T17:18:23 | null | null | null | 11 pages, 2 figures. Code available at https://github.com/johngrahamreynolds/mathematical_catastrophe_mitigation. Models available at https://huggingface.co/collections/MarioBarbeque/catastrophic-forgetting-in-mathematical-reasoning | https://arxiv.org/abs/2512.13706v1 | https://arxiv.org/pdf/2512.13706v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2512.01034v3 | 2512.01034 | 3 | AltNet: Addressing the Plasticity-Stability Dilemma in Reinforcement Learning | Artificial neural networks have shown remarkable success in supervised learning when trained on a single task using a fixed dataset. However, when neural networks are trained on a reinforcement learning task, their ability to continue learning from new experiences declines over time. This decline in learning ability is... | 227 | [
"Mansi Maheshwari",
"John C. Raisbeck",
"Bruno Castro da Silva"
] | 3 | Mansi Maheshwari | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2025-11-30T19:02:20 | 2025-11-30T00:00:00 | 2026-03-09T14:52:08 | null | null | null | null | https://arxiv.org/abs/2512.01034v3 | https://arxiv.org/pdf/2512.01034v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2511.20019v2 | 2511.20019 | 2 | How to Use Deep Learning to Identify Sufficient Conditions: A Case Study on Stanley's $e$-Positivity | In a study, published in Nature, researchers from DeepMind and mathematicians demonstrated a general framework using machine learning to make conjectures in pure mathematics. Here, we build upon this framework to develop a method for identifying sufficient conditions that imply a given mathematical statement. As a demo... | 153 | [
"Farid Aliniaeifard",
"Shu Xiao Li"
] | 2 | Farid Aliniaeifard | [] | math.CO | [
"math.CO"
] | false | 2025-11-25T07:38:56 | 2025-11-25T00:00:00 | 2026-05-05T07:11:00 | null | null | null | null | https://arxiv.org/abs/2511.20019v2 | https://arxiv.org/pdf/2511.20019v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2511.07904v3 | 2511.07904 | 3 | Test-driven Reinforcement Learning in Continuous Control | Reinforcement learning (RL) has been recognized as a powerful tool for robot control tasks. RL typically employs reward functions to define task objectives and guide agent learning. However, since the reward function serves the dual purpose of defining the optimal goal and guiding learning, it is challenging to design ... | 267 | [
"Zhao Yu",
"Xiuping Wu",
"Liangjun Ke"
] | 3 | Zhao Yu | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2025-11-11T06:58:52 | 2025-11-11T00:00:00 | 2025-12-09T07:05:57 | null | null | null | AAAI 2026 oral | https://arxiv.org/abs/2511.07904v3 | https://arxiv.org/pdf/2511.07904v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2511.04860v2 | 2511.04860 | 2 | GPT-5 at CTFs: Case Studies From Top-Tier Cybersecurity Events | OpenAI and DeepMind's AIs recently got gold at the IMO math olympiad and ICPC programming competition. We show frontier AI is similarly good at hacking by letting GPT-5 compete in elite CTF cybersecurity competitions. In one of this year's hardest events, it outperformed 93% of humans finishing 25th: between the world'... | 82 | [
"Reworr",
"Artem Petrov",
"Dmitrii Volkov"
] | 3 | Reworr | [] | cs.CR | [
"cs.CR"
] | false | 2025-11-06T22:48:10 | 2025-11-06T00:00:00 | 2025-11-14T18:57:12 | null | null | null | Proof reading edits | https://arxiv.org/abs/2511.04860v2 | https://arxiv.org/pdf/2511.04860v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2511.00423v3 | 2511.00423 | 3 | Bootstrap Off-policy with World Model | Online planning has proven effective in reinforcement learning (RL) for improving sample efficiency and final performance. However, using planning for environment interaction inevitably introduces a divergence between the collected data and the policy's actual behaviors, degrading both model learning and policy improve... | 178 | [
"Guojian Zhan",
"Likun Wang",
"Xiangteng Zhang",
"Jiaxin Gao",
"Masayoshi Tomizuka",
"Shengbo Eben Li"
] | 6 | Guojian Zhan | [] | cs.LG | [
"cs.LG",
"cs.AI",
"cs.RO"
] | true | 2025-11-01T06:33:04 | 2025-11-01T00:00:00 | 2026-01-15T13:38:44 | null | null | null | NeurIPS 2025 | https://arxiv.org/abs/2511.00423v3 | https://arxiv.org/pdf/2511.00423v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2510.27632v1 | 2510.27632 | 1 | Sketch-to-Layout: Sketch-Guided Multimodal Layout Generation | Graphic layout generation is a growing research area focusing on generating aesthetically pleasing layouts ranging from poster designs to documents. While recent research has explored ways to incorporate user constraints to guide the layout generation, these constraints often require complex specifications which reduce... | 185 | [
"Riccardo Brioschi",
"Aleksandr Alekseev",
"Emanuele Nevali",
"Berkay Döner",
"Omar El Malki",
"Blagoj Mitrevski",
"Leandro Kieliger",
"Mark Collier",
"Andrii Maksai",
"Jesse Berent",
"Claudiu Musat",
"Efi Kokiopoulou"
] | 12 | Riccardo Brioschi | [] | cs.CV | [
"cs.CV",
"cs.AI"
] | true | 2025-10-31T17:05:10 | 2025-10-31T00:00:00 | 2025-10-31T17:05:10 | null | null | null | 15 pages, 18 figures, GitHub link: https://github.com/google-deepmind/sketch_to_layout, accept at ICCV 2025 Workshop (HiGen) | https://arxiv.org/abs/2510.27632v1 | https://arxiv.org/pdf/2510.27632v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2511.11597v1 | 2511.11597 | 1 | CLINB: A Climate Intelligence Benchmark for Foundational Models | Evaluating how Large Language Models (LLMs) handle complex, specialized knowledge remains a critical challenge. We address this through the lens of climate change by introducing CLINB, a benchmark that assesses models on open-ended, grounded, multimodal question answering tasks with clear requirements for knowledge qua... | 169 | [
"Michelle Chen Huebscher",
"Katharine Mach",
"Aleksandar Stanić",
"Markus Leippold",
"Ben Gaiarin",
"Zeke Hausfather",
"Elisa Rawat",
"Erich Fischer",
"Massimiliano Ciaramita",
"Joeri Rogelj",
"Christian Buck",
"Lierni Sestorain Saralegui",
"Reto Knutti"
] | 13 | Michelle Chen Huebscher | [] | cs.AI | [
"cs.AI",
"cs.CL"
] | true | 2025-10-29T16:15:42 | 2025-10-29T00:00:00 | 2025-10-29T16:15:42 | null | null | null | Questions, system prompt and model judge prompts available here: https://www.kaggle.com/datasets/deepmind/clinb-questions | https://arxiv.org/abs/2511.11597v1 | https://arxiv.org/pdf/2511.11597v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2510.25529v1 | 2510.25529 | 1 | Off-policy Reinforcement Learning with Model-based Exploration Augmentation | Exploration is fundamental to reinforcement learning (RL), as it determines how effectively an agent discovers and exploits the underlying structure of its environment to achieve optimal performance. Existing exploration methods generally fall into two categories: active exploration and passive exploration. The former ... | 211 | [
"Likun Wang",
"Xiangteng Zhang",
"Yinuo Wang",
"Guojian Zhan",
"Wenxuan Wang",
"Haoyu Gao",
"Jingliang Duan",
"Shengbo Eben Li"
] | 8 | Likun Wang | [] | cs.AI | [
"cs.AI"
] | false | 2025-10-29T13:53:52 | 2025-10-29T00:00:00 | 2025-10-29T13:53:52 | null | null | null | null | https://arxiv.org/abs/2510.25529v1 | https://arxiv.org/pdf/2510.25529v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2510.24373v3 | 2510.24373 | 3 | AI for a Planet Under Pressure | Artificial intelligence (AI) is already driving scientific breakthroughs in a variety of research fields, ranging from the life sciences to mathematics. This raises a critical question: can AI be applied both responsibly and effectively to address complex and interconnected sustainability challenges? This report is the... | 130 | [
"Victor Galaz",
"Maria Schewenius",
"Jonathan F. Donges",
"Ingo Fetzer",
"Erik Zhivkoplias",
"Wolfram Barfuss",
"Louis Delannoy",
"Lan Wang-Erlandsson",
"Maximilian Gelbrecht",
"Jobst Heitzig",
"Jonas Hentati-Sundberg",
"Christopher Kennedy",
"Nielja Knecht",
"Romi Lotcheris",
"Miguel Ma... | 29 | Victor Galaz | [] | cs.CY | [
"cs.CY",
"cs.HC"
] | true | 2025-10-28T12:50:45 | 2025-10-28T00:00:00 | 2025-11-04T15:29:19 | null | null | null | 88 pages, 8 figures, 1 table | https://arxiv.org/abs/2510.24373v3 | https://arxiv.org/pdf/2510.24373v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2510.21418v1 | 2510.21418 | 1 | DreamerV3-XP: Optimizing exploration through uncertainty estimation | We introduce DreamerV3-XP, an extension of DreamerV3 that improves exploration and learning efficiency. This includes (i) a prioritized replay buffer, scoring trajectories by return, reconstruction loss, and value error and (ii) an intrinsic reward based on disagreement over predicted environment rewards from an ensemb... | 84 | [
"Lukas Bierling",
"Davide Pasero",
"Jan-Henrik Bertrand",
"Kiki Van Gerwen"
] | 4 | Lukas Bierling | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2025-10-24T12:58:27 | 2025-10-24T00:00:00 | 2025-10-24T12:58:27 | null | null | null | null | https://arxiv.org/abs/2510.21418v1 | https://arxiv.org/pdf/2510.21418v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2510.20927v2 | 2510.20927 | 2 | What do model reports say about their ChemBio benchmark evaluations? Comparing recent releases to the STREAM framework | Most frontier AI developers publicly document their safety evaluations of new AI models in model reports, including testing for chemical and biological (ChemBio) misuse risks. This practice provides a window into the methodology of these evaluations, helping to build public trust in AI systems, and enabling third party... | 204 | [
"Tom Reed",
"Tegan McCaslin",
"Luca Righetti"
] | 3 | Tom Reed | [] | cs.CY | [
"cs.CY"
] | false | 2025-10-23T18:35:15 | 2025-10-23T00:00:00 | 2025-10-28T17:49:18 | null | null | null | 12 pages, 6 figures. Includes appendices Added supplementary materials | https://arxiv.org/abs/2510.20927v2 | https://arxiv.org/pdf/2510.20927v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2510.06649v2 | 2510.06649 | 2 | Local Reinforcement Learning with Action-Conditioned Root Mean Squared Q-Functions | The Forward-Forward (FF) Algorithm is a recently proposed learning procedure for neural networks that employs two forward passes instead of the traditional forward and backward passes used in backpropagation. However, FF remains largely confined to supervised settings, leaving a gap at domains where learning signals ca... | 136 | [
"Frank Wu",
"Mengye Ren"
] | 2 | Frank Wu | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2025-10-08T05:06:09 | 2025-10-08T00:00:00 | 2026-04-03T17:37:35 | null | null | null | 18 pages, 11 figures | https://arxiv.org/abs/2510.06649v2 | https://arxiv.org/pdf/2510.06649v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2510.01457v4 | 2510.01457 | 4 | A Forensic Analysis of Synthetic Data in RL: Diagnosing and Solving Algorithmic Failures in Model-Based Policy Optimization | Synthetic data is central to data-efficient Dyna-style model-based reinforcement learning, but it can also degrade performance. We study this failure in Model-Based Policy Optimization (MBPO), which performs actor-critic updates using model-generated synthetic state transitions. Although MBPO reports strong sample-effi... | 200 | [
"Brett Barkley",
"David Fridovich-Keil"
] | 2 | Brett Barkley | [] | cs.LG | [
"cs.LG"
] | false | 2025-10-01T20:54:51 | 2025-10-01T00:00:00 | 2026-05-07T03:46:04 | null | null | null | null | https://arxiv.org/abs/2510.01457v4 | https://arxiv.org/pdf/2510.01457v4 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2509.24827v3 | 2509.24827 | 3 | Putnam-like dataset summary: LLMs as mathematical competition contestants | In this paper we summarize the results of the Putnam-like benchmark published by Google DeepMind. This dataset consists of 96 original problems in the spirit of the Putnam Competition and 576 solutions generated by LLMs. We analyze the performance of models on this set of problems to verify their ability to solve probl... | 104 | [
"Bartosz Bieganowski",
"Daniel Strzelecki",
"Robert Skiba",
"Mateusz Topolewski"
] | 4 | Bartosz Bieganowski | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2025-09-29T14:13:10 | 2025-09-29T00:00:00 | 2026-02-01T12:37:09 | null | null | null | 16 pages, 11 figures | https://arxiv.org/abs/2509.24827v3 | https://arxiv.org/pdf/2509.24827v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2509.24804v1 | 2509.24804 | 1 | DyMoDreamer: World Modeling with Dynamic Modulation | A critical bottleneck in deep reinforcement learning (DRL) is sample inefficiency, as training high-performance agents often demands extensive environmental interactions. Model-based reinforcement learning (MBRL) mitigates this by building world models that simulate environmental dynamics and generate synthetic experie... | 197 | [
"Boxuan Zhang",
"Runqing Wang",
"Wei Xiao",
"Weipu Zhang",
"Jian Sun",
"Gao Huang",
"Jie Chen",
"Gang Wang"
] | 8 | Boxuan Zhang | [] | cs.LG | [
"cs.LG"
] | false | 2025-09-29T13:54:42 | 2025-09-29T00:00:00 | 2025-09-29T13:54:42 | null | null | null | null | https://arxiv.org/abs/2509.24804v1 | https://arxiv.org/pdf/2509.24804v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2509.22964v4 | 2509.22964 | 4 | Functional Critics Are Essential for Actor-Critic: From Off-Policy Stability to Efficient Exploration | The actor-critic (AC) framework has achieved strong empirical success in off-policy reinforcement learning but suffers from the "moving target" problem, where the evaluated policy changes continually. Functional critics, or policy-conditioned value functions, address this by explicitly including a representation of the... | 243 | [
"Qinxun Bai",
"Yuxuan Han",
"Wei Xu",
"Zhengyuan Zhou"
] | 4 | Qinxun Bai | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2025-09-26T21:55:26 | 2025-09-26T00:00:00 | 2026-02-08T05:31:22 | null | null | null | null | https://arxiv.org/abs/2509.22964v4 | https://arxiv.org/pdf/2509.22964v4 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2509.18552v2 | 2509.18552 | 2 | Global Minimizers of Sigmoid Contrastive Loss | The meta-task of obtaining and aligning representations through contrastive pretraining is steadily gaining importance since its introduction in CLIP and ALIGN. In this paper we theoretically explain the advantages of synchronizing with trainable inverse temperature and bias under the sigmoid loss, as implemented in th... | 153 | [
"Kiril Bangachev",
"Guy Bresler",
"Iliyas Noman",
"Yury Polyanskiy"
] | 4 | Kiril Bangachev | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2025-09-23T02:24:23 | 2025-09-23T00:00:00 | 2026-03-11T01:13:04 | null | null | null | Author names listed in alphabetical order. NeurIPS 2025. New version includes some results on the geometry of CLIP in addition to geometry of SigLIP | https://arxiv.org/abs/2509.18552v2 | https://arxiv.org/pdf/2509.18552v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2509.07968v2 | 2509.07968 | 2 | SimpleQA Verified: A Reliable Factuality Benchmark to Measure Parametric Knowledge | We introduce SimpleQA Verified, a 1,000-prompt benchmark for evaluating Large Language Model (LLM) short-form factuality based on OpenAI's SimpleQA. It addresses critical limitations in OpenAI's benchmark, including noisy and incorrect labels, topical biases, and question redundancy. SimpleQA Verified was created throu... | 120 | [
"Lukas Haas",
"Gal Yona",
"Giovanni D'Antonio",
"Sasha Goldshtein",
"Dipanjan Das"
] | 5 | Lukas Haas | [] | cs.CL | [
"cs.CL"
] | false | 2025-09-09T17:53:58 | 2025-09-09T00:00:00 | 2026-03-10T12:28:26 | null | null | null | null | https://arxiv.org/abs/2509.07968v2 | https://arxiv.org/pdf/2509.07968v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2508.21809v1 | 2508.21809 | 1 | VoCap: Video Object Captioning and Segmentation from Any Prompt | Understanding objects in videos in terms of fine-grained localization masks and detailed semantic properties is a fundamental task in video understanding. In this paper, we propose VoCap, a flexible video model that consumes a video and a prompt of various modalities (text, box or mask), and produces a spatio-temporal ... | 196 | [
"Jasper Uijlings",
"Xingyi Zhou",
"Xiuye Gu",
"Arsha Nagrani",
"Anurag Arnab",
"Alireza Fathi",
"David Ross",
"Cordelia Schmid"
] | 8 | Jasper Uijlings | [] | cs.CV | [
"cs.CV"
] | false | 2025-08-29T17:43:58 | 2025-08-29T00:00:00 | 2025-08-29T17:43:58 | null | null | null | null | https://arxiv.org/abs/2508.21809v1 | https://arxiv.org/pdf/2508.21809v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2508.20228v2 | 2508.20228 | 2 | Robustness Assessment and Enhancement of Text Watermarking for Google's SynthID | Recent advances in LLM watermarking methods such as SynthID-Text by Google DeepMind offer promising solutions for tracing the provenance of AI-generated text. However, our robustness assessment reveals that SynthID-Text is vulnerable to meaning-preserving attacks, such as paraphrasing, copy-paste modifications, and bac... | 141 | [
"Xia Han",
"Qi Li",
"Jianbing Ni",
"Mohammad Zulkernine"
] | 4 | Xia Han | [] | cs.CR | [
"cs.CR",
"cs.CL"
] | true | 2025-08-27T19:17:09 | 2025-08-27T00:00:00 | 2025-10-21T19:31:09 | null | null | null | Accepted by TrustCom2025 | https://arxiv.org/abs/2508.20228v2 | https://arxiv.org/pdf/2508.20228v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2508.13922v1 | 2508.13922 | 1 | Categorical Policies: Multimodal Policy Learning and Exploration in Continuous Control | A policy in deep reinforcement learning (RL), either deterministic or stochastic, is commonly parameterized as a Gaussian distribution alone, limiting the learned behavior to be unimodal. However, the nature of many practical decision-making problems favors a multimodal policy that facilitates robust exploration of the... | 220 | [
"SM Mazharul Islam",
"Manfred Huber"
] | 2 | SM Mazharul Islam | [] | cs.LG | [
"cs.LG",
"cs.AI"
] | true | 2025-08-19T15:18:01 | 2025-08-19T00:00:00 | 2025-08-19T15:18:01 | null | null | null | 6 pages, 4 figures; Has been submitted and accepted at IEEE SMC, 2025 | https://arxiv.org/abs/2508.13922v1 | https://arxiv.org/pdf/2508.13922v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2508.11739v1 | 2508.11739 | 1 | Scalable Geospatial Data Generation Using AlphaEarth Foundations Model | High-quality labeled geospatial datasets are essential for extracting insights and understanding our planet. Unfortunately, these datasets often do not span the entire globe and are limited to certain geographic regions where data was collected. Google DeepMind's recently released AlphaEarth Foundations (AEF) provides ... | 160 | [
"Luc Houriez",
"Sebastian Pilarski",
"Behzad Vahedi",
"Ali Ahmadalipour",
"Teo Honda Scully",
"Nicholas Aflitto",
"David Andre",
"Caroline Jaffe",
"Martha Wedner",
"Rich Mazzola",
"Josh Jeffery",
"Ben Messinger",
"Sage McGinley-Smith",
"Sarah Russell"
] | 14 | Luc Houriez | [
"X the Moonshot Factory - Bellwether",
"Stanford University"
] | cs.LG | [
"cs.LG",
"cs.CV"
] | true | 2025-08-15T17:09:48 | 2025-08-15T00:00:00 | 2025-08-15T17:09:48 | null | null | null | 15 pages, 10 figures, 5 tables | https://arxiv.org/abs/2508.11739v1 | https://arxiv.org/pdf/2508.11739v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2508.09036v1 | 2508.09036 | 1 | Can We Trust AI to Govern AI? Benchmarking LLM Performance on Privacy and AI Governance Exams | The rapid emergence of large language models (LLMs) has raised urgent questions across the modern workforce about this new technology's strengths, weaknesses, and capabilities. For privacy professionals, the question is whether these AI systems can provide reliable support on regulatory compliance, privacy program mana... | 203 | [
"Zane Witherspoon",
"Thet Mon Aye",
"YingYing Hao"
] | 3 | Zane Witherspoon | [] | cs.CY | [
"cs.CY",
"cs.AI"
] | true | 2025-08-12T15:57:22 | 2025-08-12T00:00:00 | 2025-08-12T15:57:22 | null | null | null | null | https://arxiv.org/abs/2508.09036v1 | https://arxiv.org/pdf/2508.09036v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2508.01604v1 | 2508.01604 | 1 | Enhancing Math Reasoning in Small-sized LLMs via Preview Difficulty-Aware Intervention | Reinforcement learning scaling enhances the reasoning capabilities of large language models, with reinforcement learning serving as the key technique to draw out complex reasoning. However, key technical details of state-of-the-art reasoning LLMs, such as those in the OpenAI O series, Claude 3 series, DeepMind's Gemini... | 127 | [
"Xinhan Di",
"JoyJiaoW"
] | 2 | Xinhan Di | [] | cs.LG | [
"cs.LG"
] | false | 2025-08-03T05:41:36 | 2025-08-03T00:00:00 | 2025-08-03T05:41:36 | null | null | null | 7 pages, 1 table, accepted by SIM ICML@2025 Workshop | https://arxiv.org/abs/2508.01604v1 | https://arxiv.org/pdf/2508.01604v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2507.23675v2 | 2507.23675 | 2 | One-Step Flow Policy Mirror Descent | Diffusion policies have achieved great success in online reinforcement learning (RL) due to their strong expressive capacity. However, the inference of diffusion policy models relies on a slow iterative sampling process, which limits their responsiveness. To overcome this limitation, we propose Flow Policy Mirror Desce... | 136 | [
"Tianyi Chen",
"Haitong Ma",
"Na Li",
"Kai Wang",
"Bo Dai"
] | 5 | Tianyi Chen | [] | cs.LG | [
"cs.LG"
] | false | 2025-07-31T15:51:10 | 2025-07-31T00:00:00 | 2025-10-16T00:28:10 | null | null | null | null | https://arxiv.org/abs/2507.23675v2 | https://arxiv.org/pdf/2507.23675v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2507.23181v2 | 2507.23181 | 2 | Will Compute Bottlenecks Prevent an Intelligence Explosion? | The possibility of a rapid, "software-only" intelligence explosion brought on by AI's recursive self-improvement (RSI) is a subject of intense debate within the AI community. This paper presents an economic model and an empirical estimation of the elasticity of substitution between research compute and cognitive labor ... | 141 | [
"Parker Whitfill",
"Cheryl Wu"
] | 2 | Parker Whitfill | [] | econ.GN | [
"econ.GN"
] | false | 2025-07-31T01:27:49 | 2025-07-31T00:00:00 | 2025-08-16T19:36:26 | null | null | null | null | https://arxiv.org/abs/2507.23181v2 | https://arxiv.org/pdf/2507.23181v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2507.18519v2 | 2507.18519 | 2 | Revisiting Bisimulation Metric for Robust Representations in Reinforcement Learning | Bisimulation metric has long been regarded as an effective control-related representation learning technique in various reinforcement learning tasks. However, in this paper, we identify two main issues with the conventional bisimulation metric: 1) an inability to represent certain distinctive scenarios, and 2) a relian... | 169 | [
"Leiji Zhang",
"Zeyu Wang",
"Xin Li",
"Yao-Hui Li"
] | 4 | Leiji Zhang | [] | cs.LG | [
"cs.LG"
] | false | 2025-07-24T15:42:22 | 2025-07-24T00:00:00 | 2025-09-22T06:52:27 | null | null | null | null | https://arxiv.org/abs/2507.18519v2 | https://arxiv.org/pdf/2507.18519v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2507.14748v1 | 2507.14748 | 1 | Skill Learning via Policy Diversity Yields Identifiable Representations for Reinforcement Learning | Self-supervised feature learning and pretraining methods in reinforcement learning (RL) often rely on information-theoretic principles, termed mutual information skill learning (MISL). These methods aim to learn a representation of the environment while also incentivizing exploration thereof. However, the role of the r... | 155 | [
"Patrik Reizinger",
"Bálint Mucsányi",
"Siyuan Guo",
"Benjamin Eysenbach",
"Bernhard Schölkopf",
"Wieland Brendel"
] | 6 | Patrik Reizinger | [] | cs.LG | [
"cs.LG",
"cs.AI",
"stat.ML"
] | true | 2025-07-19T20:48:46 | 2025-07-19T00:00:00 | 2025-07-19T20:48:46 | null | null | null | 16 pages, 7 figures | https://arxiv.org/abs/2507.14748v1 | https://arxiv.org/pdf/2507.14748v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2507.14736v1 | 2507.14736 | 1 | Balancing Expressivity and Robustness: Constrained Rational Activations for Reinforcement Learning | Trainable activation functions, whose parameters are optimized alongside network weights, offer increased expressivity compared to fixed activation functions. Specifically, trainable activation functions defined as ratios of polynomials (rational functions) have been proposed to enhance plasticity in reinforcement lear... | 207 | [
"Rafał Surdej",
"Michał Bortkiewicz",
"Alex Lewandowski",
"Mateusz Ostaszewski",
"Clare Lyle"
] | 5 | Rafał Surdej | [] | cs.LG | [
"cs.LG"
] | false | 2025-07-19T19:53:08 | 2025-07-19T00:00:00 | 2025-07-19T19:53:08 | null | null | null | Accepted for oral presentation at CoLLAs 2025 | https://arxiv.org/abs/2507.14736v1 | https://arxiv.org/pdf/2507.14736v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2507.03578v1 | 2507.03578 | 1 | SciVid: Cross-Domain Evaluation of Video Models in Scientific Applications | In recent years, there has been a proliferation of spatiotemporal foundation models in different scientific disciplines. While promising, these models are often domain-specific and are only assessed within the particular applications for which they are designed. Given that many tasks can be represented as video modelin... | 196 | [
"Yana Hasson",
"Pauline Luc",
"Liliane Momeni",
"Maks Ovsjanikov",
"Guillaume Le Moing",
"Alina Kuznetsova",
"Ira Ktena",
"Jennifer J. Sun",
"Skanda Koppula",
"Dilara Gokay",
"Joseph Heyward",
"Etienne Pot",
"Andrew Zisserman"
] | 13 | Yana Hasson | [] | cs.CV | [
"cs.CV",
"cs.AI",
"cs.LG"
] | true | 2025-07-04T13:48:12 | 2025-07-04T00:00:00 | 2025-07-04T13:48:12 | null | null | null | ICCV 2025, GitHub repo: https://github.com/google-deepmind/scivid | https://arxiv.org/abs/2507.03578v1 | https://arxiv.org/pdf/2507.03578v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2507.03050v1 | 2507.03050 | 1 | From Turing to Tomorrow: The UK's Approach to AI Regulation | The UK has pursued a distinctive path in AI regulation: less cautious than the EU but more willing to address risks than the US, and has emerged as a global leader in coordinating AI safety efforts. Impressive developments from companies like London-based DeepMind began to spark concerns in the UK about catastrophic ri... | 294 | [
"Oliver Ritchie",
"Markus Anderljung",
"Tom Rachman"
] | 3 | Oliver Ritchie | [] | cs.CY | [
"cs.CY",
"cs.AI"
] | true | 2025-07-03T10:54:43 | 2025-07-03T00:00:00 | 2025-07-03T10:54:43 | null | null | null | This is a chapter intended for publication in a forthcoming edited volume. It is the version of the author's manuscript prior to acceptance for publication and has not undergone editorial and/or peer review on behalf of the Publisher | https://arxiv.org/abs/2507.03050v1 | https://arxiv.org/pdf/2507.03050v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2506.24068v3 | 2506.24068 | 3 | STACK: Adversarial Attacks on LLM Safeguard Pipelines | Frontier AI developers are relying on layers of safeguards to protect against catastrophic misuse of AI systems. Anthropic and OpenAI guard their latest Opus 4 model and GPT-5 models using such defense pipelines, and other frontier developers including Google DeepMind pledge to soon deploy similar defenses. However, th... | 178 | [
"Ian R. McKenzie",
"Oskar J. Hollinsworth",
"Tom Tseng",
"Xander Davies",
"Stephen Casper",
"Aaron D. Tucker",
"Robert Kirk",
"Adam Gleave"
] | 8 | Ian R. McKenzie | [] | cs.CL | [
"cs.CL",
"cs.AI"
] | true | 2025-06-30T17:21:08 | 2025-06-30T00:00:00 | 2026-02-05T01:02:44 | null | null | null | Add results on other models and datasets | https://arxiv.org/abs/2506.24068v3 | https://arxiv.org/pdf/2506.24068v3 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2506.21718v1 | 2506.21718 | 1 | Performance Prediction for Large Systems via Text-to-Text Regression | In many industries, predicting metric outcomes of large systems is a fundamental problem, driven largely by traditional tabular regression. However, such methods struggle on complex systems data in the wild such as configuration files or system logs, where feature engineering is often infeasible. We propose text-to-tex... | 143 | [
"Yash Akhauri",
"Bryan Lewandowski",
"Cheng-Hsi Lin",
"Adrian N. Reyes",
"Grant C. Forbes",
"Arissa Wongpanich",
"Bangding Yang",
"Mohamed S. Abdelfattah",
"Sagi Perel",
"Xingyou Song"
] | 10 | Yash Akhauri | [] | cs.LG | [
"cs.LG",
"cs.AI",
"cs.PF",
"cs.SE",
"eess.SY"
] | true | 2025-06-26T19:10:08 | 2025-06-26T00:00:00 | 2025-06-26T19:10:08 | null | null | null | Code can be found at https://github.com/google-deepmind/regress-lm | https://arxiv.org/abs/2506.21718v1 | https://arxiv.org/pdf/2506.21718v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2506.18260v1 | 2506.18260 | 1 | Advanced For-Loop for QML algorithm search | This paper introduces an advanced framework leveraging Large Language Model-based Multi-Agent Systems (LLMMA) for the automated search and optimization of Quantum Machine Learning (QML) algorithms. Inspired by Google DeepMind's FunSearch, the proposed system works on abstract level to iteratively generates and refines ... | 116 | [
"FuTe Wong"
] | 1 | FuTe Wong | [] | cs.AI | [
"cs.AI"
] | false | 2025-06-23T03:19:36 | 2025-06-23T00:00:00 | 2025-06-23T03:19:36 | null | null | null | 7 pages, 8 figures | https://arxiv.org/abs/2506.18260v1 | https://arxiv.org/pdf/2506.18260v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2506.17453v1 | 2506.17453 | 1 | UT-GraphCast Hindcast Dataset: A Global AI Forecast Archive from UT Austin for Weather and Climate Applications | The UT GraphCast Hindcast Dataset from 1979 to 2024 is a comprehensive global weather forecast archive generated using the Google DeepMind GraphCast Operational model. Developed by researchers at The University of Texas at Austin under the WCRP umbrella, this dataset provides daily 15 day deterministic forecasts at 00U... | 103 | [
"Naveen Sudharsan",
"Manmeet Singh",
"Harsh Kamath",
"Hassan Dashtian",
"Clint Dawson",
"Zong-Liang Yang",
"Dev Niyogi"
] | 7 | Naveen Sudharsan | [] | physics.geo-ph | [
"physics.geo-ph",
"cs.LG",
"physics.ao-ph"
] | true | 2025-06-20T19:42:36 | 2025-06-20T00:00:00 | 2025-06-20T19:42:36 | null | null | null | null | https://arxiv.org/abs/2506.17453v1 | https://arxiv.org/pdf/2506.17453v1 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
2506.16750v2 | 2506.16750 | 2 | An improved example for an autoconvolution inequality | We give a nonnegative step function with 575 equally spaced intervals such that $$\frac{\|f \ast f\|_{L^{2}(\mathbb{R})}^{2}}{\|f \ast f\|_{L^{\infty}(\mathbb{R})}\|f \ast f\|_{L^{1}(\mathbb{R})}} \geq 0.901564.$$ This improves upon a recent result of Deepmind's AlphaEvolve, which found a nonnegative step function with... | 68 | [
"Christopher Boyer",
"Zane Kun Li"
] | 2 | Christopher Boyer | [] | math.CA | [
"math.CA",
"math.CO",
"math.NT"
] | true | 2025-06-20T04:58:40 | 2025-06-20T00:00:00 | 2025-08-10T15:28:21 | null | null | null | 8 pages, 3 figures; revised version incorporating referee comments | https://arxiv.org/abs/2506.16750v2 | https://arxiv.org/pdf/2506.16750v2 | (all:"Google DeepMind" OR all:DeepMind) | 2026-08-02T22:38:49 | dFKXatlrhbN1bFDIw | XYgKrIazvJEctQUl9 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.