Dan found that the 2-bit quantization broke tool calling but upgrading to 4-bit (at 4.36 tokens/second) got that working
RESEARCH
-
LLM Coding Benchmarks Questioned Real World Performance
By
–
These kind of claims never pass the sniff test. Benchmarks can be cheated, but if it worked 0-11% of the time on real tasks (which are not part of benchmarks) nobody would ever use LLMs for coding. https://t.co/zp1qpQjf3P
— Peter Gostev (@petergostev) 19 mars 2026These kind of claims never pass the sniff test. Benchmarks can be cheated, but if it worked 0-11% of the time on real tasks (which are not part of benchmarks) nobody would ever use LLMs for coding.
-
Comparing Tool Calling Performance Between Q4 and Q2 Models
By
–
Have you observed a meaningful difference between Q4 and Q2 either when it comes to tool calling? Would love to see how you measure that
-
World Models: Unpacking AI Research Initiatives
By
–
If you've been curious about world models, read this. Got an early preview of the blog and it does a thorough job of unpacking the ill tailored tapestry of world model initiatives.
-
Frontier Models Rely on Memorization Over Generalizable Knowledge
By
–
This is more evidence that current frontier models remain completely reliant on content-level memorization, as opposed to higher-level generalizable knowledge (such as metalearning knowledge, problem-solving strategies…) https://t.co/QNqanOttqd
— François Chollet (@fchollet) 19 mars 2026This is more evidence that current frontier models remain completely reliant on content-level memorization, as opposed to higher-level generalizable knowledge (such as metalearning knowledge, problem-solving strategies…)
-
OpenAI Monitors 99.9% of Internal Traffic to Detect Anomalies
By
–
Sharing some of the work I've been doing at OpenAI: we now monitor 99.9% of internal coding traffic for misalignment using our most powerful models, reviewing full trajectories to catch suspicious behavior, escalate serious cases quickly, and strengthen our safeguards over time. [Translated from EN to English]
-
Current AI Librarian Cannot Pioneer Unknown Frontiers
By
–
Current AI is a librarian of existing knowledge. Science requires an explorer of the unknown. You don't win a Nobel Prize by staying in the library.
-
Axelera AI’s Hardware Powers HammerHAI Supercomputer for European AI
By
–
Congratulations to Hewlett Packard Enterprise and the entire HammerHAI consortium on this milestone. EuroHPC Joint Undertaking (EuroHPC JU) has signed a contract to deploy HammerHAI, an AI-optimized supercomputer at the High-Performance Computing Center Stuttgart, as part of the EU's AI Factories initiative. Delivering more than 15 exaflops of peak AI inference performance, it will give European startups, SMEs, and researchers access to world-class AI compute, free of charge. We are proud that Axelera AI's inference hardware is part of this system, and thrilled to see this kind of collaboration driving Europe's sovereign AI infrastructure forward. eu1.hubs.ly/H0sPY4h0 #AxeleraAI #HammerHAI #EuroHPC #AIFactories #EdgeAI #EuropeanAI #Semiconductors
→ View original post on X — @axeleraai, 2026-03-19 17:02 UTC
-
RAI Institute’s UMV Vehicle Achieves Advanced Movement Control with AI
By
–
Absolutely incredible work. This is a lot more impressive than dancing humanoids (and it's from the folks that pioneered the dancing humanoid). RAI Institute (@rai_inst) It was great to see our name amongst the other “AI Native” companies during @Nvidia’s #GTC keynote. NVIDIA Isaac™ Lab helps us train reinforcement learning policies that enable the UMV to drive, jump, flip, and hop like a pro! — https://nitter.net/rai_inst/status/2034646341763133873#m
→ View original post on X — @willknight, 2026-03-19 16:46 UTC
-
Humanoid Tennis Player Robot Demonstrates Advanced Physical AI
By
–
Your humanoid tennis player is here! https://
youtu.be/rDvhxKjwQrg?si
=ujSrDFs3-2xbj-CV
… via @YouTube #tennis #sports #humanoidtech #humanoid #robot #Robotics #AI #TechRevolution #TechInnovation #ArtificialInteligence #PhysicalAI @PawlowskiMario @chidambara09 @Ym78200 @CurieuxExplorer @efipm @bigfundu