Tag: Models

OneFlow, a new non-autoregressive multimodal model, enables concurrent text & image generation with variable length and outperforms existing methods using ...

New research reveals how foundation models falter in cross-modal reasoning due to attention imbalance & offers a scalable solution for ...

New research explores swarm learning using small agent networks (SANs) as a potential alternative to foundation models for adapting to ...

A qualitative comparison of cognitive & generative architectures explores challenges in theory evaluation, highlighting the complexities of modeling minds & ...

Fine-tune Gemma 3 270M for custom tasks & deploy it on-device! Learn to quantize, convert, and build fast, private, offline ...

Engineering simulations face time constraints, but "surrogate models" & reduced order techniques offer fast results using machine learning. See how ...

New research explores 'off-trajectory reasoning' – how well LLMs collaborate & recover from misleading information. Findings reveal limitations & offer ...

New research explores how LLMs like ChatGPT can understand & optimize business processes (BPMN). Findings show promise for assisting designers ...

Discover Jamba Reasoning 3B, AI21's tiny but powerful language model running fast & efficiently on phones. It redefines accessibility & ...

New research introduces Encode-Think-Decode (ETD), a technique boosting LLM reasoning by iterating over key layers & employing recursive latent thoughts. ...

Discover SDAR, a new paradigm merging autoregressive efficiency with diffusion's parallel inference. This unlocks scalable sequence generation & improved reasoning ...

Discover IBM's Granite 4.0 language models on Docker Hub! Developers can now easily build & deploy generative AI apps with ...

A guide using a decision matrix to help select forecasting models, considering seasonality, trends, stationarity & data volume for improved ...

New research introduces Diffusion Routers (DR) for universal domain translation, enabling AI to translate between any pair of domains with ...

Llama.cpp now features resumable downloads for GGUF models, eliminating frustrating interruptions. Learn how this improves accessibility & explore Docker Model ...

New research reveals 'function vectors' within LMMs control spatial relationships, improving accuracy & enabling generalization to novel scenarios. A breakthrough ...

New research introduces 'Advisor Models' – learnable policies that dynamically guide black-box LLMs via natural language steering. This adaptive approach ...

Introducing MoNaCo: a new benchmark pushing LLMs with challenging questions across dozens of documents for improved reasoning & evaluation. #LLMs ...

Unlock the future of AI! Explore how evolving architectures are pushing **language models** beyond simple text generation, creating powerful tools ...

New Adaptive Reasoning Suppression (ARS) boosts Large Reasoning Language Models' efficiency by dynamically suppressing redundant steps & saving tokens, latency, ...

Fluid Benchmarking adapts evaluation items based on LLM capability, overcoming limitations of static benchmarks & providing granular insights for model ...

AI2 releases OLMoASR, open automatic speech recognition models trained from scratch on a curated dataset, democratizing access & fostering innovation.

Unlock the potential of GenerativeAI! Explore how these models are transforming creativity, automation & problem-solving across diverse industries. Discover practical ...

Unlock your model's full potential! Learn practical strategies for effective fine-tuning, boosting performance & achieving superior results with minimal effort. ...

Confused about choosing the right machine learning model? This guide simplifies model selection, offering practical strategies to maximize accuracy and ...

AI doesn't have to cost you a dime—local models are fast, private, and finally worth switching to.

Explore the differences between word & sentence embeddings! Learn when to use each approach for optimal results in your NLP ...

Google releases updated Gemini 2.5 Flash and Flash-Lite preview models with improved quality, speed, and efficiency, introducing a '-latest' alias ...

Discover how new foundation models are enabling few-shot learning for time series data, unlocking insights across industries.

Google's Test-Time Diffusion (TTD) enhances AI model accuracy during inference by injecting noise & using diffusion to refine outputs, improving ...