1221 articles · Page 42 / 51
BigCodeBench evaluates AI coding productivity using 1,000+ real-world libraries, challenging models with complex tasks beyond simple algorithm puzzles.
Compare DeepSpeed and FSDP performance based on model size and explore strategies using Hugging Face Accelerate.
Learn how Dell Enterprise Hub reduces AI operational costs by up to 75% using optimized on-premise infrastructure and open-source models.
Optimizing LLM performance using distilabel and Argilla 2.0 through high-quality synthetic data.
Frontier AI Safety Frameworks, critical capability levels, and regulatory compliance for autonomous agents.
Gemini 2.5 DT achieves gold-medal-level reasoning at ICPC, signaling a shift toward autonomous agentic coding systems.
Gemini class AI accelerates cosmic simulations, transforming astronomical research through thinking tokens and multimodal data.
Aeneas by Google DeepMind uses multimodal networks to restore ancient inscriptions and predict their historical context.
Genie 3 by Google DeepMind enables real-time interactive world models, advancing the future of robotics and simulations.
Google Perch 2.0 uses bioacoustics and multimodal reasoning to monitor ecosystems, providing vital data for ESG and TNFD.
Hebrew AI leaderboard shows local models like DictaLM-3.0 outperforming GPT-4o by addressing complex linguistic structures.
Hugging Face partners with Artificial Analysis to provide real-time metrics on AI model throughput, latency, and cost.
Hugging Face Open CoT Leaderboard evaluates AI reasoning transparency and logic using the Marginal Accuracy Gain metric.
Explore how Intel Gaudi's Assisted Generation and speculative decoding optimize LLM inference speed by up to 3 times.
LAVE evaluates semantic accuracy in Document VQA using LLMs, offering better correlation with human judgment.
LeRobot v0.4.0 introduces Dataset v3.0 for standardization and optimizes inference to enable robotics on edge devices.
Explore LiveCodeBench, a benchmark measuring AI's genuine coding and self-repair skills through time-segmented evaluation.
T2I leaderboards use Elo ratings and DiT architecture to objectively evaluate and rank image generation models.
Explore how Hugging Face TGI optimizes GPU resources by serving multiple LoRA adapters simultaneously using SGMV kernels.
Explore preference optimization and reward models for reducing hallucinations and improving zero-shot reasoning in VLMs.
Analyzing the WhisperPair vulnerability in Google Fast Pair that allows unauthorized eavesdropping on Bluetooth audio devices.
Explore how power grids and infrastructure define the 2026 global order amidst GPT 5.2 and the rise of sovereign AI.
How AI models like GPT 5.2 and MTSViT transform ecosystem monitoring and climate crisis response in 2026.
GPT 5.2 and Gemini 3 lead a mathematical revolution using neuro-symbolic systems and formal verification to achieve reliable AGI.