LLMs & Generative AI - Page 10 of 48
Latest breakthroughs in large language models and generative AI shaping the future of artificial intelligence and machine learning.
950 articles
View complete article list
GraphDC Uses Divide‑and‑Conquer Agents to Scale Graph Reasoning
LLMs have made strides on many math‑heavy prompts, yet they still stumble when faced with graph‑centric problems.
RateQuant reveals mixed-precision KV cache pitfall: β decay rates span 3.6‑5.3
Why does KV‑cache memory matter for large language models? Because every token generated adds a new key‑value pair, and the cache expands linearly with sequence length, quickly becoming the dominant memory cost in serving.
Top 10 2026 LLM Papers Highlight Pass@k Efficiency for Reasoning Models
Why does this matter now? Because 2026 marks a shift from sheer size to purpose. Large language models are being judged on safety, controllability and real‑world utility rather than just parameters.
Generative AI fuels industrial-scale record 2025 data breaches, ITRC reports
Generative AI and autonomous agents are turning identity theft into a near‑industrial operation. While the tech is impressive, its misuse is stark: Experian reports that 40 percent of the data breaches investigated last year involved AI.
Strain drives exponential error growth; vorticity only linear impact
Flow matching builds data by stepping through a learned velocity field, and each integration step—measured as number of function evaluations (NFE)—adds directly to inference cost.
LKV learns head-wise budgets and token selection for LLM KV cache eviction
Why does long‑context inference still choke on memory? In large language models the key‑value cache expands linearly with each processed token, forcing developers to prune information aggressively.
LLM Summarizers Omit Identification, Distinguish Observed vs Inferred Claims
Reading the raw transcript reveals a troubling pattern: two sections trace back to a single ambiguous sentence, one line was invented outright, and three more simply echo what the model expects a meeting summary to contain.
NVIDIA's Star Elastic bundles 30B, 23B, 12B models; 23B hits 85.63 on AIME-2025
Training a family of large language models has always been a cost‑heavy exercise. Every variant—whether 8 B, 30 B, or 70 B—needs its own full training run, its own storage footprint and its own deployment stack.
Understanding 'Compute': The Core Power Driving Modern AI Models
Artificial intelligence is reshaping everything, and it’s doing so with a brand‑new vocabulary. Spend five minutes scanning any AI‑focused feed and you’ll hit LLMs, RAG, RLHF and a slew of acronyms that can make even seasoned engineers pause.
Fields Medalist: ChatGPT 5.5 Pro produced PhD-level math proof in under an hour
British mathematician Timothy Gowers, a Fields Medalist who holds the Combinatorics Chair at the Collège de France and a fellowship at Trinity College, Cambridge, recently posted a blog entry that puts a new spin on AI‑assisted research.
Key Topics for LLM Engineers: Using Instruction Data to Align Models
Why does this matter for anyone crossing into large‑language‑model work? Because the jump from computer‑vision pipelines to LLM engineering isn’t just a change of data type—it forces you to relearn everything from how text becomes numbers to how a...
Semantic memory query retrieves Friday deployment approval for user-123
Memory isn’t a luxury for an AI agent; it’s a prerequisite for anything beyond a single prompt.
OpenAI launches Realtime‑Translate for 70+ languages and Realtime‑Whisper transcription
Voice agents have long been costly to run and tricky to orchestrate. The problem isn’t the models’ ability to converse; it’s the context ceilings that force engineers to build session resets, compress state, and reconstruct layers for every...
Google's Chrome 4GB on-device AI model unchanged, but explanation lacking
Google’s Chrome now ships a 4 GB Gemini Nano model for on‑device AI, and a handful of users have suddenly seen their browsers ask for extra storage.
RVPO boosts HealthBench score to 0.261, beating GDPO’s 0.215 at 14B (p < 0.001)
Why does this matter? Current critic‑less RLHF pipelines add up multiple reward signals with a simple average.
Adding Tools and Memory Expands AI Agent Threat Surface, Study Finds
Most AI security work still treats large language models like a simple chatbot: you type, it replies. That view made sense when the only interface was text.
Grammar-Constrained Decoding Adds 495 Bash Tasks, Some Regressions Seen
Bash sits at the core of many AI‑driven workflows, letting a model emit grep, curl, tar, or full pipelines that actually touch files, open network sockets, or stitch tools together.
SAT ensures improvement and plug‑and‑play upgrades in multi‑LLM training
Large language models keep getting bigger, but the price tag isn’t getting any smaller. Researchers have started swapping a single, massive model for a squad of leaner LLMs that can, in theory, hit the same marks—or even beat them. The catch?
Transfer Learning Boosts Efficiency of Physics-Informed Neural Networks
Why does this matter? Because scientific machine learning often stalls when data are scarce and physics constraints are hard‑coded.
GPT‑5.5 and GPT‑5.5‑Cyber Scale Trusted Access for Cyber Defense
Why does this matter now? For the first time, a single family of models is being positioned to serve every layer of the cyber‑defense ecosystem. While the tech is impressive, the rollout is deliberate.
📚 Featured Resources & Reviews
🎓
No Code MBA Course Review
Learn to build AI-powered apps without coding. Our comprehensive review of No Code MBA's course.
🔧
AI Tools & Resources
Curated collection of AI tools, courses, and frameworks to accelerate your AI journey.
📬
Weekly AI Digest
Get the week's most important AI news delivered to your inbox every week.