when rag fails: four alternatives
retrieval-augmented generation breaks in predictable ways at scale, and engineers are turning to long-context prompting, memory compression, structured retrieval, and graph-based reasoning instead.
aitag machine learning
summaries filed under machine learning
pytorch adds cross-repo ci relay for backend testing
pytorch introduces a cross-repository ci relay that automatically triggers and tracks downstream ci, showing results on a single dashboard.
aiproper positive-only learning solved with new condition
a new combinatorial condition called uniform exterior separability characterizes when a concept class can be properly learned from positive-only samples.
aidarts unifies zero-shot time series foundation models
darts introduces a unified interface for foundation models like chronos-2 and timesfm 2.5, simplifying zero-shot forecasting in python.
aiagentic workflows for data science pipelines
five concrete agentic workflows automate major stages of a data science pipeline, from exploratory data analysis to feature engineering, reducing manual effort.
aifine-tune language models on apple silicon with mlx
fine-tune open language models locally on your mac using mlx, with no cloud gpus or costs required.
aigemini creates google sheets from prompts
use gemini to build spreadsheets, generate formulas, and analyze data with natural language commands.
aiwhich tokens hybrid models predict better
a study comparing token-level predictions of a hybrid model and a transformer reveals where each architecture excels.
aineuro-symbolic framework grounds amr predictions in biology
kg-trace combines neural networks with a knowledge graph to make antimicrobial resistance predictions that align with known biological pathways.
aitokenspeed-kernel: clean apis for multi-silicon llm inference
tokenspeed-kernel introduces a layered api and registry system to decouple llm inference runtimes from hardware-specific kernels, enabling portable and high-performance multi-silicon support.
aispin up a vllm server on hugging face jobs in one command
launch a private, openai-compatible llm endpoint on hugging face infrastructure with a single command, paying per second with no server setup.
aifive open source omni ai models for multimodal tasks
a practical look at open source models that handle text, images, audio, and video for real-world ai applications.
aion-device neural architecture search for sensor data
a lightweight neural architecture search runs directly on embedded devices to adapt tiny neural networks to new users and sensor data in real time.
aiai architect roadmap 2026
a step-by-step guide to the design, decision-making, and leadership skills needed to become an ai architect in 2026.
aihidden scale escapes per-loop loss in looped language models
dense per-loop cross-entropy does not control all recurrent state variables, leaving a readout blind spot that allows hidden-state scale to grow unchecked.
aiminimax pac bounds for learning in exogenous contextual mdps
new sample complexity bounds for reinforcement learning with i.i.d. contexts, showing how many oracle calls are needed before and during policy execution.
aispectral phase diagram for few-shot learning
a new saturation index helps decide when to stop labeling data for binary classification.
aifederated causal discovery and inference survey
a systematic review of federated causal discovery and inference methods, covering design choices, privacy, and open challenges.
aiopen far-field asr benchmark reveals real-world gap
the ffasr leaderboard is the first open benchmark evaluating asr models under realistic far-field acoustic conditions, showing a large performance gap compared to near-field tests.
ainemo automodel speeds moe fine-tuning 3.7x over transformers v5
nvidia nemo automodel builds on huggingface transformers v5 to deliver 3.4-3.7x higher training throughput and up to 32% less gpu memory for mixture-of-experts models with a single import change.
ai