AI News Archive - Browse Page 121 of 262
Browse AI news articles covering LLMs, tools, research, and industry trends
Language Models Forecast Research Success Using 11,488 Comparative Idea Pairs
Forget raw intelligence. Predicting a good research idea is a job for a well-trained referee.
OpenAI’s Q1 2026 adjusted margin slips to –122%, burning USD 1.22 per USD 1 earned
OpenAI is losing $1.22 for every dollar it earns, even after excluding stock-based compensation.
VSAS‑Bench Introduces Standardized Real‑Time Evaluation for Visual Assistants
Building a visual assistant that keeps up with reality is hard. Measuring it is harder. Most tests are a slideshow. The real world is a live feed.
F_Call_Analysis_Planner forwards Parent_Instruction to generate Selection_Rule
The best part of any system is the small, stupid piece that does one job perfectly. This one is called F_Call_Analysis_Planner.
Temporal Contrastive Transformer embeddings boost financial crime detection
The promise of self-supervised learning in financial crime detection rests on a single, powerful idea: that a model can discover behavioral patterns...
Quantum ML Hits Data Input Bottleneck: Processors Can't Read Images, Text
A quantum processor can’t look at a cat photo. It can’t read this sentence. That’s the problem.
Experimental MLX Delegate Enables PyTorch Models on Apple Silicon GPUs
For Mac developers, PyTorch has always demanded a choice: sacrifice GPU performance or abandon your established tools. That compromise just narrowed.
OSCToM uses RL to generate adversarial scenarios testing high-order Theory of Mind
Most AI can't handle a simple lie. Getting one to grasp a complex, layered misunderstanding where someone acts on a belief the AI knows is false has...
Gemma 4 Executes Sequential Tool Calls to Inspect Folder and Compute Results
Look at the files. Compute the total size. Two simple instructions, but for a small language model they demand sequential reasoning.
Alibaba's Qwen3.7-Max runs 35 hrs, self‑monitors reward‑hacking, supports Claude Code
Alibaba’s Qwen 3.7-Max, a large language model marketed as a “versatile agent foundation,” completed a 35-hour test on an unfamiliar processor last...
Researchers use triplet loss to train high-quality Horn logic embeddings
Machine logic snaps. Teaching it to flex is the real challenge. Consider Horn logic, a rule-based system where conclusions hinge on perfect chains of...
Cohere launches Command A+ 218B sparse MoE, runs on two H100 GPUs using QAD
Cohere’s latest move is a study in efficiency without compromise. Command A+ packs 218 billion parameters into a sparse mixture of experts, and it...
Gemini 3.5 Flash Shows Fast Responses in Free Account Tests
Speed sells AI. It's a neat trick. A common lie, too. Most models pause. They take a long, digital sigh before answering.
Positive-IC 46.4% indicates negative bias; |IC| just under 0.02 after two runs
A 46.4% positive-IC ratio is a confession: the signal tilts negative more often than not.
AI Leap and OpenAI integrate global AI expertise into Estonia's schools
AI Leap and OpenAI are bringing their tech to Estonian classrooms. Announced April 18, the partnership has a dual goal: integrating tools like Codex...
PyTorch Docathon 2026 merges 150+ PRs, adds API and ExecuTorch docs
Good documentation is infrastructure. It’s the silent scaffold that turns a powerful framework into a usable tool, and the PyTorch Docathon 2026 just...
Pricing Change Alters Complaint Language, Skews Classifier Accuracy
A pricing change can break your model before you even see the data. It reshapes how customers complain, and the language you trained your classifier...
Polyend's Endless AI pedal opens gallery for third‑party Playground effects
Polyend's Endless AI guitar pedal just went from being a clever toy to a legitimate platform.
OpenAI's AI achieves 'milestone in AI mathematics,' say Gowers, Shankar
An algorithm has found a proof for a problem that once required human intuition. Mathematician Timothy Gowers calls it "a milestone in AI...
Cohere open-sources Command A+ with 128k token window, 48‑language support
Cohere has thrown open the doors to its most capable model yet. Command A+ arrives with a 128,000-token context window, support for 48 languages, and...