The Archive

AngelSpec unified training framework adaptively switches between multi-token prediction and block-parallel diffusion for speculative LLM decoding across heterogeneous workloads.

Online learning algorithms for distributed constraint optimization applied to large-scale satellite scheduling via decomposition and iterative pricing methods.

Agentic workflow automates translation of quantum protocols to hardware experiments on Pasqal neutral-atom QPUs with researcher-in-the-loop validation.

Zero-shot sEMG movement classification via Compositional Prototype Interpolation and Synthetic Adaptation for prosthesis control without re-training per combination.

Joint agent-speculator RL aligns next tool-call prediction with deployed agent behavior by unifying speculator and agent in single model, reducing latency.

Evaluation of adversarial robustness in five Arabic language models under multi-granularity attacks exposes security vulnerabilities in non-English LLMs.

Physics-informed spectral deep operator network (SpectONet) for Euler-Bernoulli beam vibration using nonuniform sensor placement.

MC-ALFCG algorithm for stochastic nonconvex composite optimization over Markovian-sampled gradients with variance reduction.

WorkSurface-Bench evaluates enterprise agents on multi-source knowledge routing across documents, tables, and graphs with 1,151 tasks.

WALoMA: multitask 6G wireless foundation model using masked autoencoders to reduce dependence on labeled channel data.

Rule-based Visiting Circulation model for trip destination prediction in unseen metropolitan areas using behavioral patterns.

Method to detect CSAM-generating LoRAs from weight fingerprints (singular vectors) without generating outputs, enabling safer moderation.

Concept bottleneck model for echocardiography shows loss invariance determines learned concepts; validates volume grounding without residual paths.

Analysis of how image quality degradation (noise, darkness) affects neural network robustness in autonomous driving perception systems.

Since launching last year, the startup today has more than 8 million people using the open-source or hosted version of its models, and now generates annual recurring revenue of $21 million.

Multi-UAV cooperative control framework combining shared voxel-map occupancy model with multi-agent Soft Actor-Critic decentralized policy.

LLM-generated sustainability-framed recommendation explanations as behavioral nudges; tests how different framings influence user choice.

Dynamic Constraint Learning framework learns local surrogates for constrained optimization when constraint functions are unknown.

HYSET evaluates tool sets jointly for LLM agents via hyperedge prediction, improving on sequential tool retrieval.

First federated learning study on SpeechLLM-based end-to-end ASR with communication-efficient optimization for distributed training.

Google DeepMind releases Gemini Robotics 2, extending multimodal foundation model capabilities to whole-body robot control and coordination.

Recursive’s $400 million outlay represents the bulk of the company’s fundraising to date.

Deep learning models reconstruct urban air pollution fields from sparse monitoring stations in Paris.

Google Search AI Mode assists with offline activities like concert ticketing and event planning.

Google Search AI features generate menu and tablescape suggestions for dinner party planning.

Cognivia LLM-based copilot for Cognitive Behavioral Therapy with domain-specific cognitive distortion annotations.

Rashomon Alignment measures functional model similarity geometrically across input space, beyond data-distribution-dependent metrics.

DynaBridge multimodal framework fuses acoustic, visual, textual data with LLM-generated summaries for DASS mental health assessment.

DecoEvo decouples scoring to co-evolve LLM solver and rubric generator in text space, addressing evaluation bottlenecks.