Search: “models”

Fine-tuned a distilbert with ml-intern today for the first time. The procedure is really straightforward, almost unexpectedly so. - Found a …

model-releasesclem-delangue--x

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

Research

Fine-Tuning Without Forgetting via Loss-Adaptive Learning Rates

DGX agent

arXiv:2605.20005v1 Announce Type: new Abstract: Fine-tuning large language models on new data improves task performance but degrades capabilities learned during pretraining, a phenomenon known as cata

researcharxiv-cs-lg

20 May 2026

First-Passage Prediction of Grokking Delay: ACalibrated Law under AdamW with Causal Validation

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

FLUIDSPLAT: Reconstructing Physical Fields from Sparse Sensors via Gaussian Primitives

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

FLUXtrapolation: A benchmark on extrapolating ecosystem fluxes

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

Forward launches Predict to verify network changes before they reach production

model-releasessiliconangle

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

From Llama to Cria: Scaling Down Neural Networks via Neuron-Level Spectral Structural Importance Evaluation

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

From SGD to Muon: Adaptive Optimization via Schatten-p Norms

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

GoLongRL: Capability-Oriented Long Context Reinforcement Learning with Multitask Alignment

model-releasesarxiv-cs-cl

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

Google I/O, Gemini Spark, Antigravity

model-releasessimon-willison

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

Local Ai

GRASP: Deterministic argument ranking in interaction graphs

arXiv:2605.19141v1 Announce Type: cross Abstract: Large language models are increasingly deployed as automated judges to evaluate the strength of arguments. As this role expands, their legitimacy depe

local-aiarxiv-cs-ai

Local Ai

GRLoc: Geometric Representation Regression for Visual Localization

arXiv:2511.13864v2 Announce Type: replace Abstract: Absolute Pose Regression (APR) has emerged as a compelling paradigm for visual localization. However, APR models typically operate as black boxes, d

local-aiarxiv-cs-cv

How Ramp engineers accelerate code review with Codex

Ramp engineers use OpenAI's Codex to streamline and accelerate their code review process, leveraging AI-assisted code understanding and analysis. The implementation likely demonstrates how Codex helps

model-releasesopenai

Industry

I’m not a doomer an AI at all. I think the nature of work, particularly entry level jobs will change. Ai will make business more complicated…

I’m not a doomer an AI at all. I think the nature of work, particularly entry level jobs will change. Ai will make business more complicated and competitive. Not less. Which means more layers where hu

industryclem-delangue--x

Tutorials

INAR-VL: Input-Aware Routing for Edge-Cloud Vision-Language Inference

arXiv:2605.18853v1 Announce Type: cross Abstract: Edge deployment of Vision-Language Models (VLMs) faces a tradeoff between latency and accuracy: cloud execution provides high-quality predictions but

tutorialsarxiv-cs-cv

Investigating Cross-Modal Skill Injection: Scenarios, Methods, and Hyperparameters

arXiv:2605.19523v1 Announce Type: cross Abstract: Vision-Language Models (VLMs) have demonstrated remarkable proficiency in general multi-modal understanding; yet they struggle to efficiently acquire

researcharxiv-cs-ai

KappaPlace: Learning Hyperspherical Uncertainty for Visual Place Recognition via Prototype-Anchored Supervision

arXiv:2605.19435v1 Announce Type: cross Abstract: Visual Place Recognition (VPR) is critical for autonomous navigation, yet state-of-the-art methods lack well-calibrated uncertainty estimation. Standa

Learning-Accelerated Optimization-based Trajectory Planning for Cooperative Aerial-Ground Handover Missions

arXiv:2605.19562v1 Announce Type: cross Abstract: This paper presents a learning-augmented trajectory planning framework for cooperative unmanned aerial vehicle (UAV) and unmanned ground vehicle (UGV)

Less is More: Efficient Black-box Attribution via Minimal Interpretable Subset Selection

arXiv:2504.00470v2 Announce Type: replace-cross Abstract: To develop a trustworthy AI system, which aim to identify the input regions that most influence the models decisions. The primary task of exis

researcharxiv-cs-cv

LLM Benchmark Datasets Should Be Contamination-Resistant

arXiv:2605.19999v1 Announce Type: cross Abstract: Benchmark datasets are critical for reproducible, reliable, and discriminative evaluation of LLMs. However, recent studies reveal that many benchmark

Matern Noise for Triangulation-Agnostic Flow Matching on Meshes

arXiv:2605.19305v1 Announce Type: cross Abstract: This paper tackles the task of learning to generate signals over triangle meshes in a triangulation-agnostic manner, meaning the trained model can be

researcharxiv-cs-cv

Applications

Multi-Token Residual Prediction

arXiv:2605.18817v1 Announce Type: new Abstract: Diffusion Language Models (DLMs) generate text by iteratively denoising masked token sequences, offering a tradeoff between parallelism and quality comp

applicationsarxiv-cs-lg

Nonlinearity as Rank: Generative Low-Rank Adapter with Radial Basis Functions

arXiv:2602.05709v2 Announce Type: replace Abstract: Low-rank adaptation (LoRA) approximates the update of a pretrained weight matrix using the product of two low-rank matrices. However, standard LoRA

PlantTraitNet: An Uncertainty-Aware Multimodal Framework for Global-Scale Plant Trait Inference from Citizen Science Data

arXiv:2511.06943v3 Announce Type: replace-cross Abstract: Global plant maps of plant traits, such as leaf nitrogen or plant height, are essential for understanding ecosystem processes, including the c

Position: The Turing-Completeness of Real-World Autoregressive Transformers Relies Heavily on Context Management

arXiv:2605.19514v1 Announce Type: new Abstract: Many works make the eye-catching claim that Transformers are Turing-complete. However, the literature often conflates two distinct settings: (i) a fixed

Safety

Prediction Is Not Physics: Learning and Evaluating Conserved Quantities in Neural Simulators

arXiv:2605.18883v1 Announce Type: cross Abstract: A diffusion model trained on Hamiltonian trajectories can achieve rollout MSE near 10^{-3}, but the standard deviation of its energy over time is betw

safetyarxiv-cs-ai

Preferences Order, Ratings Anchor: From Fused Expert Aesthetic Ground Truth to Self-Distillation

arXiv:2605.19776v1 Announce Type: new Abstract: Pairwise preferences and pointwise ratings are the two dominant annotation protocols in image aesthetic assessment (IAA), yet existing benchmarks adopt

Program Evaluation with Remotely Sensed Outcomes

arXiv:2411.10959v4 Announce Type: replace-cross Abstract: We study causal inference in experiments and quasi-experiments, where the economic outcome is imperfectly measured by a remotely sensed variab

Quantum Adversarial Machine Learning: From Classical Adaptations to Quantum-Native Methods

arXiv:2605.18821v1 Announce Type: new Abstract: Machine learning has revolutionized numerous industrial domains. Despite recent advances, machine learning models remain vulnerable to adversarial threa

researcharxiv-cs-lg

Quantum Machine Learning for Cyber-Physical Anomaly Detection in Unmanned Aerial Vehicles: A Leakage-Free Evaluation with Proxy-Audited Feature Sets

arXiv:2605.19233v1 Announce Type: cross Abstract: Unmanned aerial vehicles (UAVs) are cyber-physical systems whose attack surface spans networked avionics and on-board sensor fusion: a compromised GPS

ReacTOD: Bounded Neuro-Symbolic Agentic NLU for Zero-Shot Dialogue State Tracking

arXiv:2605.19077v1 Announce Type: cross Abstract: Task-oriented dialogue systems -- handling transactions, reservations, and service requests -- require predictable behavior, yet the moderately-sized

RecoAtlas: From Semantic Plausibility to Set-Level Utility in LLM Recommendation Agents

arXiv:2605.18805v1 Announce Type: cross Abstract: LLM recommendation agents increasingly produce structured recommendation reports: sets of items accompanied by natural-language justifications. Yet ex

Replacement Learning: Training Neural Networks with Fewer Parameters

arXiv:2605.19533v1 Announce Type: new Abstract: End-to-end training with full-depth backpropagation remains the dominant paradigm for optimizing deep neural networks, but its efficiency deteriorates a

SCAFDS: Edge-Feature Graph Attention for Interbank Fraud Detection with Attribution-Grounded SAR Generation

arXiv:2605.18913v1 Announce Type: cross Abstract: The U.S. financial system processes approximately 1.3 million interbank transactions daily, yet no system in the reviewed literature models fraud prop

researcharxiv-cs-ai

SEAL: Semantic Aware Image Watermarking

arXiv:2503.12172v4 Announce Type: replace-cross Abstract: Generative models have rapidly evolved to generate realistic outputs. However, their synthetic outputs increasingly challenge the clear distin

researcharxiv-cs-cv

Sequential Consensus for Multi-Agent LLM Debates: A Wald-SPRT compute governor with calibration-based failure detection

arXiv:2605.19193v1 Announce Type: new Abstract: Multi-agent LLM debate improves factuality and reasoning, but most recipes pick a fixed round count, over-spending on easy items and under-spending on h

Sonar-TS: Search-Then-Verify Natural Language Querying for Time Series Databases

arXiv:2602.17001v2 Announce Type: replace Abstract: Natural Language Querying for Time Series Databases (NLQ4TSDB) aims to assist non-expert users retrieve meaningful events, intervals, and summaries

Applications

Taming Real-World Space-Time Video Super-Resolution with One-Step Diffusion

arXiv:2601.20308v2 Announce Type: replace Abstract: Diffusion models have demonstrated exceptional success in video super-resolution (VSR), exhibiting powerful capabilities for generating fine-grained

applicationsarxiv-cs-cv

Safety

Text-to-SPARQL Generation with Reinforcement Learning: A GRPO-based Approach on DBLP

arXiv:2605.20066v1 Announce Type: new Abstract: Knowledge graph question answering seeks to translate natural language questions into executable queries over knowledge graphs, but existing approaches

safetyarxiv-cs-cl

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 …

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions l

model-releasesopenai--x

Towards Trust Calibration in Socially Interactive Agents: Investigating Gendered Multimodal Behaviors Generation with LLMs

model-releasesarxiv-cs-cl

arXiv:2605.19798v1 Announce Type: new Abstract: As Socially Interactive Agents (SIAs) become increasingly integrated into daily life, the ability to calibrate user trust to an agent's actual capabilit

Training Neural Networks with Optimal Double-Bayesian Learning

arXiv:2605.20009v1 Announce Type: cross Abstract: Backpropagation with gradient descent is a common optimization strategy employed by most neural network architectures in machine learning. However, fi

Trajectory Planning and Control near the Limits: an Open Experimental Benchmark on the RoboRacer Platform

model-releasesarxiv-cs-ro

arXiv:2605.19881v1 Announce Type: new Abstract: We present a modular framework to benchmark new and existing methods for trajectory planning and control in high-acceleration maneuvers that push autono

We partnered with artists, designers, and builders to create new AI tools that solve real problems in their creative workflows. Here’s what’…

model-releasesgoogle-ai--x

We partnered with artists, designers, and builders to create new AI tools that solve real problems in their creative workflows. Here’s what’s new: — Introducing Google Pics in @GoogleWorkspace: A bran

We're excited to be an official shoutout at the Google I/O Developer Keynote 🔥 @llama_index is building the document infrastructure for AI …

model-releasesjerry-liu--x

We're excited to be an official shoutout at the Google I/O Developer Keynote 🔥 @llama_index is building the document infrastructure for AI agents, and we plan to integrate even more heavily with both

XNote: Benchmarking Automated Community Notes Generation for Image-based Contextual Deception

model-releasesarxiv-cs-cl

arXiv:2603.22453v2 Announce Type: replace Abstract: Community Notes have emerged as an effective crowd-sourced mechanism for combating online deception on social media platforms. However, its reliance

A Comprehensive Survey of Action Quality Assessment: Method and Benchmark

arXiv:2412.11149v2 Announce Type: replace Abstract: Action Quality Assessment (AQA) aims to automatically evaluate how well human actions are performed and has been widely applied in sports analysis,

19 May 2026

A No-Defense Defense Against Gradient-Based Adversarial Attacks on ML-NIDS: Is Less More?