Enterprise AI just got a new benchmark. http://
H2O.ai’s h2oGPTe is #1 on the GAIA benchmark—again—with 79.7% accuracy, outperforming Google & Microsoft by 30+ points. GAIA tests what really matters: reasoning, research, docs, decisions.
Not just chat—real work.
RESEARCH
-

h2oGPTe Tops GAIA Benchmark with 79.7% Accuracy
By
–
-
Complicated ML Techniques Often Prove Less Useful Than Expected
By
–
a lot of the complicated stuff turned out not to be very useful or important. a few random examples: graphical models, variational autoencoders, RNN/LSTM, anything bayesian
-
NeurIPS Review Experience: LLM-Generated Papers Quality Issues
By
–
i just reviewed five papers for NeurIPS and it was an awful experience: – first paper was clearly LLM-generated. it was too short, the references didn't work, had no experiments or theory at all, and a ton of obvious mistakes. the more i read the less it made sense
– two were -
Strong Regularization Prevents RLHF Model Degradation
By
–
May your regularizer be strong, lest you RLHF to slop.
-
Community-Backed AI Research Lab Replaces Traditional Investors
By
–
It's true. We have no investors. We're entirely supported by a huge community of everyday people. We've seen government research labs, academic and corporate research labs, but this is the first time we've ever seen a community-backed research lab. Thank you <3
-
ACP Course: Building Communicating AI Agents with IBM BeeAI
By
–
New Course: ACP: Agent Communication Protocol
— Andrew Ng (@AndrewYNg) 25 juin 2025
Learn to build agents that communicate and collaborate across different frameworks using ACP in this short course built with @IBMResearch's BeeAI, and taught by @sandi_besen, AI Research Engineer & Ecosystem Lead at IBM, and… pic.twitter.com/ykhi6eHH8NNew Course: ACP: Agent Communication Protocol Learn to build agents that communicate and collaborate across different frameworks using ACP in this short course built with @IBMResearch
's BeeAI, and taught by @sandi_besen
, AI Research Engineer & Ecosystem Lead at IBM, and -
EmbaseAI: GenAI Solution for Biomedical Literature Research
By
–
Live: #EmbaseAI. A new #GenAI solution built on Embase’s trusted biomedical literature database. Ask #research questions in natural language Get fully cited. http://
spkl.io/6047A6YFP
Guided by #Elsevier’s Responsible AI Principles. -

AlphaGenome Predicts DNA Mutation Impact Comprehensively
By
–
Today we introduced AlphaGenome, a new tool that can more comprehensively predict the impact of single variants or mutations in DNA How, you ask? tldr; Our AlphaGenome model takes a long DNA sequence as input, processes that data, and predicts thousands of molecular
-

Retro AI Image Generation by SpeedResearchLab
By
–
retro shots created by speedresearchlab.https://t.co/CJBPlkEk3D pic.twitter.com/iDEJEAJgfZ
— KREA AI (@krea_ai) 25 juin 2025retro shots created by speedresearchlab. https://
krea.ai/user/speedrese
archlab
… -
Interpretability as AI subfield: leading researchers and career opportunities
By
–
interpretability is definitely its own subfield, and imo very interesting & important, there just aren't as many open roles for interpretability work here are some people that come to mind @NeelNanda5 @hendrycks @ch402 @soniajoseph_ @csinva