Research & Benchmarks - Page 5 of 24
Academic AI research, performance benchmarks, scientific breakthroughs, and peer-reviewed studies advancing artificial intelligence frontiers.
475 articles
View complete article list
OpenClaw founder runs 100 AI agents for USD 1.3M/month code, review PRs, find bugs
Peter Steinberger, the mind behind the open‑source project OpenClaw, has built a tiny team that leans heavily on AI.
New benchmark shows AI video generators look realistic but lack reasoning
Modern generators such as Sora 2, Seedance 2.0 and Veo 3.1 can now churn out clips that look almost photorealistic. But a new benchmark from Tsinghua University shows that looking good isn’t the same as getting it right.
Researchers train AI model achieving near-full performance using 12.5% of experts
The Allen Institute for AI and UC Berkeley have unveiled EM O, a mixture‑of‑experts language model that builds internal modules around content domains instead of pure syntax.
RecursiveMAS cuts multi-agent inference time 2.4×, slashes token use 75%
Why does this matter? Researchers have introduced RecursiveMAS, a framework that reshapes how multiple language models collaborate.
ArXiv to ban authors of papers with unchecked LLM‑generated content
ArXiv is tightening the screws on AI‑generated text. Starting now, any preprint that shows “incontrovertible evidence” the authors skipped a sanity check on large‑language‑model output will trigger a one‑year ban from the repository.
Building an MCP-Routed AI Agent: Dynamic Tool Exposure via Keywords, Tags, and Constraints
Here’s the thing: building an AI agent that only shows the tools it truly needs isn’t a fantasy—it’s the core of this tutorial.
BenchJack proposes secure-by-design AI benchmark audit with eight flaw taxonomy
Why does the reliability of AI agent benchmarks matter now? Because they steer everything from research funding to real‑world deployments.
12‑Metric AI Agent Eval Harness Built in 9‑14 Days Across 100+ Deployments
“How do you know your agent isn’t hallucinating patient symptoms?” That question haunted a team that already had unit tests, integration tests and a model that shone on demo data—yet lacked any way to gauge hallucination rates, context faithfulness,...
Google DeepMind adds Gemini-powered cursor to Chrome for visual queries
The mouse pointer has been the quiet workhorse of personal computing for more than half a century—tracking position, registering clicks, and otherwise staying invisible. Google DeepMind researchers say that’s about to change.
BaLoRA adds Bayesian uncertainty to low‑rank adaptation, but lags fine‑tuning
LoRA has become the go‑to method for trimming the cost of fine‑tuning massive pretrained models.
Community review tools guide novices in AI research, study finds
We launched Parameter Golf to see how a tightly bounded problem would stir the machine‑learning community.
Tilde Research's Aurora optimizer beats Muon and NorMuon at 340M scale
Here's the thing: Tilde Research just dropped Aurora, a new optimizer that patches a hidden flaw in Muon.
OpenAI unveils Daybreak to secure Codex, with industry and government rollout
OpenAI just rolled out Daybreak, a cybersecurity program that leans on its latest AI models, the Codex Security agent, and a growing roster of security partners.
New embeddings prioritize preferential similarity over semantics for clustering
Why does this matter? Because platforms that let people voice opinions in full sentences are increasingly common, yet the algorithms that group those inputs still rely on representations tuned for meaning, not for agreement.
Baidu's Ernie 5.1 Cuts 94% Pre‑Training Costs Using Once‑For‑All Framework
Baidu has rolled out Ernie 5.1, a distilled version of its earlier Ernie 5.0. While the new model runs on roughly a third of the total parameters and uses about half the active parameters per query, Baidu says pre‑training costs dropped to just six...
Hermes Agent tops use as Nous Research’s self‑improving model leads OpenRouter
Why does this matter? As of May 10, 2026, Hermes Agent—an open‑source model from Nous Research—has claimed the top spot on OpenRouter’s global daily app and agent rankings.
Palisade Research: Open‑weight AI like Qwen boost autonomous hacking
Palisade Research has put AI agents through a practical test that reads like a cyber‑war scenario.
Study proposes method to curb AI reward hacking in safety tests
Why does this matter? As AI systems grow more capable, the gap between what they can do and what humans can verify widens.
Build Python Vector Search with Cosine Similarity for Scale‑Invariant Matching
You've probably typed a query into a search bar and got results that matched your words but missed the point. That's the gap most users feel: exact‑match engines versus meaning‑aware retrieval.
AI success shifts from 95% accuracy to latency, cost, and reliability
Remember the days when a data scientist’s pride came from squeezing that extra 0.7 % accuracy out of an XGBoost model after an overnight grid search?
📚 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.