#llm
18 articles
Latest
PLACO Framework Cuts Cost of Human-AI Teams by Selective Model Queries
Multi-stage system reduces computational expense while maintaining performance through strategic human-model interaction.
Four Papers Deploy Deep Learning and LLMs for Clinical Diagnosis Tasks
New research applies neural networks to chronic rhinosinusitis prediction, knee osteoarthritis grading, Alzheimer's progression modeling, and medical image reasoning.
OpenAI Makes GPT-5.5 Instant Default ChatGPT Model
New model reduces hallucinations in law, medicine, finance while preserving fast response times.
DriftBench Measures LLM Fidelity Loss in Iterative Ideation Tasks
New benchmark reveals how large language models drift from original constraints during multi-turn collaborative refinement.
BoostLoRA Breaks Rank Ceiling in Parameter-Efficient Fine-Tuning
New method expands adapter expressivity without increasing parameter count, addressing core efficiency-performance tradeoff.
LLMs Fail to Generate Random Numbers From Statistical Distributions
New research reveals large language models cannot faithfully sample from probability distributions, a critical gap for stochastic systems.
DeepSeek Releases V4 Model With Longer Context Window
The Chinese AI startup's new flagship processes vastly longer prompts while undercutting Western competitors on price.
OpenAI Powers Codex Agent With GPT-5.5 on NVIDIA
GPT-5.5 now runs Codex on NVIDIA GB200 infrastructure, expanding AI agents into knowledge work.
DeepSeek releases V4 with million-token context window
New model closes gap with frontier AI systems while improving efficiency over previous versions.
OpenAI Releases GPT-5.5, Narrowly Outperforms Anthropic's Claude
The new model excels at coding, research, and data analysis while racing ahead in benchmark competitions.
LLMs Learn Backwards: Scaling Hypothesis Faces Hard Limits
New analysis suggests large language models acquire intelligence in reverse order, challenging unlimited scaling assumptions.
Minimax M2.7 Arrives With New Capabilities
Chinese AI lab Minimax releases M2.7 model to compete in crowded LLM marketplace.