Paper: https://arxiv.org/abs/2604.24763 Check out http://AlphaSignal.ai for a daily summary of the top AI models, repositories, and research papers. Read by 280,000+ developers.
RESEARCH
-
METR’s Graph: 50% Success, Not 100
By
–
Hot take on METR’s new graph that so many people are flipping about today. • Claude Code is a real advance; Mythos probably builds on some of what is learned there. But… • If you read the graph carefully, it is about achieving *50%* success. Not 100 or 99 or even 90. The
-
RouteMoA: Efficient AI Agent Collaboration via Intelligent Model Routing
By
–
What if you could boost AI agent collaboration while slashing costs by nearly 90%? Researchers from SJTU, CUHK, Tencent, and NTU present RouteMoA. Instead of running every model first, a lightweight scorer predicts each model’s potential from the query alone, then a mix of
-
Strategic Decisions for Future AI Influence
By
–
Without a trace of irony, at all: one of the most important strategic decisions for every publication, think tank, and etc is “What decisions are we taking today to influence future training runs and inference sessions?”
-
ERNIE 5.1 achieves near SOTA with only 6% pre-training cost
By
–
Hold on, Chinas ERNIE 5.1 is almost SOTA but using only around 6% of the pre-training cost of comparable models?? ERNIE 5.0’s pre-training foundation: Baidu says ERNIE 5.1 achieves stronger search, reasoning, knowledge Q&A, creative writing, and agentic capabilities while using
-
13B Language Model Mimicking 1930
By
–
Researchers trained a 13B language model that thinks it's 1930.
— AlphaSignal AI (@AlphaSignalAI) 9 mai 2026
It speaks like someone from a century ago, with no awareness of computers, World War II, or anything after.
The team behind it, talkie, gathered 260B tokens from public-domain books, newspapers, patents, and… pic.twitter.com/JDbjmh1E3bResearchers trained a 13-billion-parameter language model that believes it is 1930. It communicates like someone from a century ago, unaware of computers, World War II, or any events that followed. The team behind it, Talkie, collected 260 billion tokens from public-domain books, newspapers, and patents.
-
New Research Paper Examines Capabilities of Autonomous AI Agents
By
–
Harvard, MIT, Stanford & Carnegie Mellon just released one of the most disturbing AI papers of 2026. “Agents of Chaos” http://
arxiv.org/pdf/2602.20021 Autonomous AI agents with real tools:
→ Email
→ File systems
→ Persistent memory
→ Shell access http://
drdebashisdutta.com -
Context Awareness Unlocks Smarter Research
By
–
The bigger unlock is context awareness. You can bring:
→ Multiple tabs
→ Documents
→ Images Into one query. That means the system understands your entire research flow, not isolated searches. This is what Chrome’s AI Mode is aiming for, and it changes how we learn and -
Lighter and Faster Language Models
By
–
Sparser, Faster, Lighter Transformer-Based Language Models Paper:
-
New RAG Approach ‘Blockify’ Improves Search Efficiency
By
–
Naive RAG vs. Blockify!
— Akshay 🚀 (@akshay_pachaar) 9 mai 2026
There's a new RAG approach that:
– cuts corpus size by 40x.
– reduces tokens per query by 3x.
– improves vector search relevance by 2.3x.
Blockify GitHub: https://t.co/R0LSWC1V8a https://t.co/xNuoXSBj7E pic.twitter.com/ibwgORIHIbNaive RAG vs. Blockify! There's a new RAG approach that: – cuts corpus size by 40x.
– reduces tokens per query by 3x.
– improves vector search relevance by 2.3x. Blockify GitHub: https://
github.com/iternal-techno
logies-partners/blockify-agentic-data-optimization
…