Knowledge catalogue
arXiv:2605.18999v1 Announce Type: new Abstract: Muon and related normalized optimizers decouple the choice of update direction from the choice of step scale, but their practical performance remains se
arXiv:2605.18993v1 Announce Type: cross Abstract: Task vector composition has emerged as a promising paradigm for editing pre-trained models, enabling model merging through addition and unlearning thr
arXiv:2605.19779v1 Announce Type: new Abstract: We adapt split conformal prediction and adaptive conformal inference (ACI) to continuous AI agent evaluation, providing distribution-free coverage guara
arXiv:2605.18871v1 Announce Type: cross Abstract: When Large Language Models produce structured outputs such as travel plans, code solutions, or multi-step proofs, individual reasoning steps may appea
arXiv:2605.19029v1 Announce Type: new Abstract: Reliable robotic manipulation requires control policies that can accurately represent and adapt to uncertainty arising from contact-rich interactions. M
arXiv:2602.23622v2 Announce Type: replace-cross Abstract: Significant progress has been made in the field of Instruction-based Image Editing Models (IIEMs). However, while these models demonstrate pla
arXiv:2605.18915v1 Announce Type: cross Abstract: Multimodal Large Language Models (MLLMs) are vulnerable to jailbreak attacks, which can elicit harmful responses from MLLMs. Many MLLMs support multi-
arXiv:2605.19688v1 Announce Type: new Abstract: Document manipulation localization models achieve strong performance on public benchmarks yet fail to generalize to operational document workflows. We i
arXiv:2605.20049v1 Announce Type: cross Abstract: As autonomous coding agents see rapid adoption, their evaluation has primarily focused on task completion rates holding the target codebase fixed. Thi
arXiv:2605.19527v1 Announce Type: new Abstract: Occluded person re-identification focuses on matching partially visible pedestrians across multiple camera views. However, occlusions disrupt body-regio
arXiv:2605.18767v1 Announce Type: cross Abstract: Multi-hop question answering requires aggregating information from multiple documents, a critical capability for knowledge-intensive applications. A f
arXiv:2605.18904v1 Announce Type: cross Abstract: Model merging enables the reuse of fine-tuned models without joint training or access to original data. Dynamic merging further improves flexibility b
arXiv:2605.18815v1 Announce Type: new Abstract: Modern large language model (LLM) training is inherently dynamic: resource fluctuations, RLHF phase shifts, and cluster elasticity continually reshape t
arXiv:2605.19130v1 Announce Type: cross Abstract: Children acquire language grounding with remarkable robustness from limited visuo-linguistic input in ways that surpass today's best large multimodal
arXiv:2605.19559v1 Announce Type: cross Abstract: The rapid development of Multimodal Large Language Models (MLLMs) has led to growing interest in egocentric video understanding, specifically the abil
arXiv:2605.19004v1 Announce Type: new Abstract: Accurately forecasting human trajectories from an egocentric perspective plays a central role in applications such as humanoid robotics, wearable sensin
arXiv:2512.23461v2 Announce Type: replace-cross Abstract: Reward models (RMs) are essential in reinforcement learning from human feedback (RLHF) to align large language models (LLMs) with human values
arXiv:2605.19093v1 Announce Type: new Abstract: System prompts are a central control mechanism in modern AI systems, shaping behavior across conversations, tasks, and user populations. Yet they are di
arXiv:2605.18820v1 Announce Type: cross Abstract: Superposition allows Transformers to reason in depth, carrying an entire reasoning frontier in parallel through a bounded-depth forward pass instead o
arXiv:2602.04381v2 Announce Type: replace-cross Abstract: Real-time polyp segmentation is essential for early colorectal cancer detection, yet clinical deployment remains blocked by GPU dependency. We
arXiv:2605.19743v1 Announce Type: new Abstract: Large Language Model (LLM) agents are increasingly applied to engineering design tasks, yet existing evaluation frameworks do not adequately address mul
arXiv:2410.18856v4 Announce Type: replace Abstract: Frontier large language models (LLMs), such as GPT-5, Claude 4.5, Gemini 3, Llama 4, and DeepSeek-R1, represent a transformative class of AI tools c
arXiv:2605.18872v1 Announce Type: cross Abstract: Robotic assembly in architectural construction faces a persistent bottleneck: existing planners are either highly specialized, requiring prohibitive r
arXiv:2605.18867v1 Announce Type: cross Abstract: Test-time model evolution offers a promising way for deployed models to improve from unlabeled test-time experience, yet most existing methods depend
arXiv:2605.18937v1 Announce Type: new Abstract: Patient-managed Personal Health Records (PHRs) promises to empower patients to better understand their health; but information in the record is complex,
arXiv:2605.19506v1 Announce Type: new Abstract: First-person dynamic spatial reasoning requires models to track continuous motion and precise geometric structure, but the quadratic attention cost of T
arXiv:2605.19283v1 Announce Type: cross Abstract: Sequential prediction is challenging in regimes of delayed disambiguation, where early observations are ambiguous and multiple latent explanations rem
arXiv:2605.19826v1 Announce Type: new Abstract: Operators of safety-critical industrial processes increasingly rely on digital twins to screen control interventions, but such simulators rarely carry c
arXiv:2605.19478v1 Announce Type: cross Abstract: Existing ViT backdoor attacks based on backbone-overwriting full-tuning are computationally expensive and inflict performance degradation. This has fo
arXiv:2605.19916v1 Announce Type: cross Abstract: We introduce Contrastive FUSE, a fast and unified framework for scalable node representation learning in graphs with partially available pairwise node
arXiv:2605.18908v1 Announce Type: cross Abstract: Deep neural networks (DNNs) remain critically vulnerable to backdoor attacks. Existing post-training detectors often require clean or surrogate data,
arXiv:2512.08237v3 Announce Type: replace Abstract: The advancement of vision-only Bird's-Eye-View (BEV) perception, a core paradigm for cost-effective autonomous driving, is hindered by the long-stan
arXiv:2507.03122v2 Announce Type: replace-cross Abstract: This study investigates the feasibility and performance of federated learning (FL) for multi-label ICD code classification using clinical note
arXiv:2605.18936v1 Announce Type: cross Abstract: Social media text data are often used to train Machine Learning (ML) models to identify users exhibiting high-risk mental health behaviors. However, s
arXiv:2602.04555v2 Announce Type: replace Abstract: Learning from a stream of tasks usually pits plasticity against stability: acquiring new knowledge often causes catastrophic forgetting of past info
arXiv:2605.18824v1 Announce Type: cross Abstract: Evaluation of foundation models often rely on aggregate scores from benchmarks that lack comprehensive coverage and metadata for a fine-grained evalua
Fine-tuned a distilbert with ml-intern today for the first time. The procedure is really straightforward, almost unexpectedly so. - Found a few datasets relevant to the task (prompt injection detectio
arXiv:2507.10614v2 Announce Type: replace-cross Abstract: The integration of large language models (LLMs) into automated algorithm design has shown promising potential. A prevalent approach embeds LLM
arXiv:2605.19846v1 Announce Type: cross Abstract: Vision-Language Models (VLMs) have demonstrated remarkable capabilities in general video understanding, yet they often struggle with the fine-grained
arXiv:2605.18845v1 Announce Type: cross Abstract: We give the first quantitative prediction of grokking delay under AdamW. Treating the delay as a first-passage time, we derive a closed-form law T_gro
arXiv:2605.18866v1 Announce Type: cross Abstract: Reconstructing continuous flow fields from sparse surface-mounted sensors is central to aerodynamic design, flow control, and digital-twin instrumenta
arXiv:2605.19812v1 Announce Type: cross Abstract: We introduce FLUXtrapolation, a benchmark for extrapolating ecosystem fluxes under progressively harder distribution shifts. Ecosystem fluxes are cent
for anyone curious, this was the result of many experiments bouncing around but this version uses... initial ideation: chatgpt for ideas/@replit for quick build final repo build: opus 4.7 prompting cl
For centuries, the scientific method has been our best tool for progress. But today, there’s so much data out there that it’s impossible for any one researcher to connect all the dots. We want to fix
Network verification company Forward Inc. today launched Forward Predict, a new capability that lets network teams test proposed changes against a digital twin of their production network before deplo
arXiv:2605.18860v1 Announce Type: cross Abstract: This paper proposes a neuron pruning framework based on neuron-level spectral structural importance evaluation. Given a trained neural network, we rec
arXiv:2605.19824v1 Announce Type: new Abstract: Recent attempts to support high-level scene interpretation and planning in Autonomous Vehicles (AVs) using ensembles of Large Language Models (LLMs) and
arXiv:2605.19781v1 Announce Type: new Abstract: Modern optimizers, like Muon, impose matrix-wise geometry constraints on their updates. These matrix-wise constraints can be unified under Linear Minimi
arXiv:2605.19263v1 Announce Type: new Abstract: Physics-informed neural networks (PINNs) offer a mesh-free framework for solving partial differential equations (PDEs), yet training often suffers from
arXiv:2605.18865v1 Announce Type: cross Abstract: Self-attention serves as the core foundation of large-scale transformer pretraining, but its quadratic token interaction cost makes inference expensiv
fun read on real tradeoffs & design decisions we debated when designing Engine for the data scale that customers produce one common thread is that we’re pretty strong supporters of just giving the age
Google introduced Gemini 3.5, its latest family of models combining frontier intelligence with action capabilities, representing a major leap forward in building more capable, intelligent agents. The
arXiv:2605.19813v1 Announce Type: new Abstract: We prove a general lower bound for differentially private federated learning protocols with arbitrary public-transcript interactions. The protocol may u
arXiv:2407.17200v3 Announce Type: replace-cross Abstract: Many real-world decision problems require solving, again and again, combinatorial optimization instances drawn from a common distribution. A r
arXiv:2605.20006v1 Announce Type: new Abstract: Geospatial reasoning requires solving image-grounded problems over the complex spatial structure of a scene. However, developing this capability is hind
arXiv:2605.19577v1 Announce Type: new Abstract: We present GoLongRL, a fully open-source, capability-oriented post-training recipe for long-context reinforcement learning with verifiable rewards (RLVR
It's hard to find much to write about Google I/O this year because I have a policy of not writing about anything that I can't try out myself, and a lot of the big announcements are 'coming soon'. I ac
Anu Adegbola / Search Engine Land: Google says it is testing new ad formats in search results and AI Mode, including Conversational Discovery ads, Highlighted Answers, and AI-powered Shopping ads — Go
Google's AI-powered Search era apparently also extends to its ads. Now, when you search for a product, Google's Gemini AI chatbot will surface relevant items and generate a 'custom explainer' about wh
Got to play with a little of this before launch as well. My experience as a social scientist was that it was more bioscience focused right now, but I think Google has been the leading lab in releasing